Strategies for reducing

losses in distribution
networks
Goran Strbac, Predrag Djapic,
Danny Pudjianto, Ioannis Konstantelos, Roberto Moreira

Imperial College London

February 2018
Acknowledgments
The authors are grateful for very valuable inputs and advice received from Steve Mould and Evert van
Schalkwyk from UK Power Networks.

2
 Content
Executive Summary .............................................................................................................. 5
1 Quantification of network losses ................................................................................... 11
1.1 Loss Operation & Investment Model ..................................................................... 11
1.2 Losses Heat Maps ................................................................................................ 12
1.3 Analysis of losses ................................................................................................. 14
1.4 Impact of sample rate on losses modelling ........................................................... 18
1.5 Summary .............................................................................................................. 18
2 Distribution of losses across network segments ........................................................... 20
2.1 Overall network-level analysis ............................................................................... 20
2.2 LV network analysis .............................................................................................. 21
2.3 HV feeders with high losses .................................................................................. 21
2.4 Cluster-based network analysis ............................................................................ 24
2.5 Summary .............................................................................................................. 28
3 Identification of potential operational strategies for loss reduction ................................ 29
3.1 Optimisation of Normally Open Point locations ..................................................... 29
3.2 Power factor correction ......................................................................................... 31
3.3 Voltage control driven loss reduction .................................................................... 37
3.4 Balancing load across phases .............................................................................. 39
3.5 Harmonics ............................................................................................................ 40
3.5.1 Impact of harmonics on transformer’s and Joule losses ................................. 40
3.5.2 LV Network Harmonics .................................................................................. 40
3.5.3 Long HV Feeders, Harmonic Resonance ....................................................... 41
3.6 Upgrading single phase spur to three phase ......................................................... 42
3.7 Impact on transmission losses .............................................................................. 43
3.8 Matching Demand with PV Output ........................................................................ 44
3.9 Switching off transformers..................................................................................... 45
3.10 Summary .............................................................................................................. 46
4 Application of smart technologies for reduction of losses ............................................. 47
4.1 Case Study No 1 - Flexible Plug and Play (FPP) Project....................................... 47
4.1.1 Loss performance of different control approaches for QB .............................. 48
4.1.2 Distributed Generation (DG) reactive power capability for losses reduction ... 50
4.2 Case study No 2: Soft open points for loss minimisation ....................................... 53
4.2.1 Introduction .................................................................................................... 53
4.2.2 Case Study 1: Eastbourne Terrace ................................................................ 53
4.2.3 Case Study 2: Boyce’s Street ........................................................................ 54

3
 4.2.4 Case Study 3: Prudential North ...................................................................... 55
4.2.5 Impact of SOP efficiency ................................................................................ 56
4.3 Case study No 3: Role and value of energy storage systems in minimizing
distribution network losses............................................................................................... 56
4.3.1 Introduction .................................................................................................... 56
4.3.2 Modelling considerations ............................................................................... 57
4.3.3 Results........................................................................................................... 59
4.4 Summary .............................................................................................................. 63
5 Identification of efficient loss reduction investment strategies....................................... 65
5.1 Eco-design: low-loss transformers ........................................................................ 65
5.2 Amorphous Steel Transformers ............................................................................ 66
5.3 Conductor Sizes Rationalisation ........................................................................... 67
5.4 Voltage rationalisation ........................................................................................... 73
5.4.1 HV networks .................................................................................................. 73
5.4.2 EHV networks ................................................................................................ 74
5.5 Smart distribution transformer ............................................................................... 74
5.6 Scott Connected Transformers ............................................................................. 75
5.7 Impact of Distribution Transformer Density ........................................................... 76
5.8 Impact of tapering ................................................................................................. 78
5.9 Summary .............................................................................................................. 79
6 Conclusions ................................................................................................................. 80
7 References .................................................................................................................. 86
8 Acronyms ..................................................................................................................... 87

4
Executive Summary

The key objectives of the modelling and analysis carried out by Imperial College, presented
and analysed in this report, are:
 Provide new insights and understanding of network losses in UK Power Networks’
distribution networks by quantifying losses across different network segments. This covers
each of their three licence areas and considers the impact and significance of different
losses drivers relative to each area;
 Assess the effectiveness of alternative loss management strategies through the
consideration of network reconfiguration, power factor compensation, phase unbalance
management and power harmonic reduction;
 Investigate the opportunities for the application of novel smart grid technologies, already
deployed by UK Power Networks’ Innovation team, to manage network losses;
 Identify areas with high losses and determine the optimal approach to manage losses
within these areas. Quantify the potential losses reduction in these areas and use these
findings to inform UK Power Networks Losses Management Strategy;
 Investigate efficient loss reduction investment strategies including the application of low-
loss transformers, investment in high-capacity cables and overhead lines including service
cables, converting single-phase and Scott connected networks to three-phase supplies
and the impact of (removing) tapering;
 Support dissemination and communication of learnings across DNO community.
Modelling Framework:
Using the established Load Related Expenditure (LRE) model as a foundation, a new
modelling tool - Loss Operation & Investment Model (LOIM), has been developed and applied
throughout this project. The extent of the model covers all three licence areas served by UK
Power Networks from low voltage networks to grid supply points and has been used to
calculate losses within these areas. This is in stark contrast to the previous analysis of network
losses that has traditionally been based on the application of representative distribution
networks1. The LOIM has been used to generate Losses Heat Maps for each of UK Power
Networks’ areas in order to identify the regions in which network losses are most significant.
The effectiveness of various network losses-reduction techniques in different UK Power
Networks areas were analysed in detail to provide core insights regarding the business cases
for alternative losses mitigation strategies and losses-driven network infrastructure
investment.
Quantification of network losses
The analysis carried out highlighted that more than 75% of network losses are associated with
LV networks, HV networks and distribution transformers. Overall:
 36-47% of the total losses are in LV networks
 9-13% of losses are associated with distribution transformer load related losses
 7-10% of losses are associated with distribution transformer no-load losses
 17-27% are in HV networks

1 Imperial College London and Sohn Associates, Management of electricity distribution network
losses, supported by UKPN and WPD, 2014

5
 17-24% of total losses are in primary and grid transformers, and EHV and 132 kV
networks.
Understanding the contribution of different network sections to the total losses will be important
when identifying loss management strategies, assessing corresponding cost effectiveness
and determining the potential impact of those strategies.
Distribution of losses across network segments
Asset utilisation and circuit lengths are major losses drivers and hence their impacts have
been investigated and analysed across each region. UK Power Networks operate a wide
range of network types. These range from rural areas, such as parts of Norfolk and Suffolk, to
very densely populated urban areas like London. The corresponding peak demand density
varies from a very low 0.05 MW/km2, to a relatively high density of 137 MW/km2. In this context,
average utilisations of distribution transformers of 51% and 38% are observed in LPN and
EPN areas respectively.
Furthermore, the proportion of transformers which have a utilisation factor in excess of 70%
in LPN is 20%, while in EPN this figure is only 4%.
Detailed power flow modelling revealed that HV feeders in LPN deliver an average of 50%
more energy than feeders in EPN, while circuits in LPN are typically about 60% shorter than
in the EPN region. In this context, the analysis demonstrated that losses in LPN are primarily
driven by high network utilisation, while in EPN, losses are driven by long feeder lengths.
Overall, the LV network losses are comparable in both areas despite LPN LV networks having
significantly shorter lengths but higher loading. Conversely, losses in the HV networks are
greater in the EPN region.
The analysis demonstrated that the magnitudes of losses vary significantly across each
network type. Modelling quantified losses for more than 4,000 HV feeders, demonstrating a
relatively small number of HV feeders are characterised with high losses. About 70% of the
total losses are in 20% of the feeders. This clearly demonstrates that loss reduction initiatives
in HV networks should target a relatively small proportion of the feeders characterised by these
high losses. Undertaking a targeted approach will maximise the cost efficiency of this activity.
An unequal distribution of losses was noted in the LV network with more than 50% of losses
noted to occur in only 20% of LV feeders.
Based on advanced neural networks methodology, UK Power Networks’ HV feeders and LV
networks were classified into 22 clusters. These clusters were determined according to the
number of customers and their load characteristics, network length, rating, type and
construction. Average parameters for each cluster were quantified and corresponding
representative networks created. These included a range of rural and urban networks, and the
related loss performance for each was assessed.
As a significant amount of losses are associated with a small number of very specific feeders,
it should be noted that use of generic feeders with average parameters may not provide
appropriate evidence to inform the development of effective losses reduction strategies.
Identification of potential operational strategies for loss reduction
A number of key losses drivers were identified and analysed. Learning from this analysis can
be used to inform the development of future losses reduction strategies. These include
changes in network operational topology, improvement of power factor, changes in load
profile, controlling phase imbalance and harmonic distortion.
Key results of conducted case studies are as follows:

6
 Analysis demonstrated that Normally Open Point (NOP) reconfiguration could reduce HV
feeder losses by up to 15% in specific areas. The economic case for this operational
strategy, as a result, appears to be strong.
 For the three UK Power Networks licence areas feeders are ranked by the possible
reduction in losses driven by power factor improvement. The potential for loss reduction is
assessed assuming power factor improvement from 0.85 to 0.95. This would lead to
reduction in losses on each feeder between 11% and 14%. It is interesting that the
modelling demonstrated that improving power factor in only one third of HV feeders could
achieve 90% of potential losses reduction. Hence, the list of 30 highest ranked HV feeders
in each licence area is created and measurements of the actual power factor in future trials
are proposed to be carried out.
 It was noted that phase imbalance increases losses non-linearly. For example, phase
imbalance ranging from 10% to 30% would increase losses by 5% to 45% respectively.
As a consequence, we identified a list of 30 LV networks that would deliver the highest
benefits for imbalance improvement, based on the networks’ electrical characteristics.
 Implementing voltage management across UK Power Networks’ three licence areas could
potentially reduce losses by around 5%. Further investigation is required to understand
the voltage dependency of customer loads. Measurements are recommended to enhance
the understanding of voltage dependency in real time. This information will aid the
formation of future loss mitigation strategies. Performing actual measurements of voltage
dependency of demand in different segments of the network should provide key
information related to the potential development of corresponding loss mitigation
strategies.
 Harmonic distortion is limited though network design standards, which ensure that the
impact of harmonic currents on networks are limted. The impact of voltage harmonics on
transformer no-load losses is linearly dependant on the total harmonic distortion (THD),
and hence, the impact on losses in this domain is more significant. Eco design
transformers’ iron losses are lower than previous transofmer specifications. The net effect
of this should mean that the impact of harmonic distortion on no-load losses will decrease
over time.
Application of smart-grid technologies for reduction of network losses

 Modelling demonstrated that the use of UK Power Networks’ Quadrature Booster, beyond
the network constraint management utilised by their Flexible Plug and Play (FPP) project2,
could deliver savings in the local network losses from about 11% in the case of high
demand and high distributed generation (DG) growth, up to 25% for low demand and low
DG growth.
 Furthermore, modelling demonstrated that optimally controlling the power factor of
distributed generators in the FPP project area could potentially reduce 33kV network
losses by 13%.
 Smarter Network Storage (SNS)3 installed in Leighton Buzzard to manage peak demand
and postpone network reinforcement (in addition to delivering system balancing services),
could potentially reduce losses in supplying circuits by about 15%.

2 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Flexible-Plug-and-
Play-(FPP)/
3 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Smarter-Network-

Storage-(SNS)/

7
 Modelling demonstrated that Soft Open Points (SOPs)4, installed for the management of
constraints in LV feeders, could potentially reduce losses in the corresponding LV network
and distribution transformers by about 10%-15%.
 Potentially further reduction in losses could be achieved by optimizing NOP positions in
real time to take into account changes in demand and generation.
 The former Department of Energy and Climate Change (DECC) indicated that smart
meters, combined with home display units, could reduce energy consumption by 2.8%5.
Analysis showed that correspondingly, distribution network losses would reduce by 5.5%
due to the decrease in consumption.
 Furthemore, analysis demonstrated that demand side response, which could potentially
shift 2.5% load from peak to off-peak period, would lead to a reduction of losses by about
3%.
Identification of efficient loss reduction investment strategies

 UK Power Networks could save 17GWh per annum by replacing all Health Index 4 and 5
distribution transformers with Ecodesign units. Given the current rate of replacement,
savings could reach up to 3.2 GWh per year.
 Loss reduction benefits alone are not sufficient to justify the upgrade of existing
underground cables. Howerver, when thermal constaints drive network reinforcement ,
installing cables of higher capacity would significantly reduce losses. In this context,
analysis carried out to determine the benefits in loss reduction by adopting a minimum
feeder cross-section area of 185 mm². This would reduce LPN HV feeder losses by 10%.
The corresponding values for EPN and SPN are 40% and 32% respectively. Removing
tapering could potentially decrease losses by up to 25%. For LV networks, the benefits of
applying larger cables would be very significant, ranging from 52% to 63%, depending on
the area.
 Using 30-minute samples tends to understate network losses, particularly in service cables
that supply one customer only. To inform this process, 5,000 five-second samples from
the Low Carbon London (LCL)6 project were used comparatively. This modelling
demonstrated that applying higher sampling rates increases calculated losses by a factor
of 1.9 compared with the losses estimated using half-hourly profiles (the range is from 1.2
to 5.8). This further reinforces the case for significantly increasing the standard capacity
of service cables.
 If single-phase HV spurs are converted to three phase, losses could potentially be reduced
by up to 80% in the corresponding network.

Benefits of loss reduction strategies

Based on the analysis carried out, the capitalised value of the benefits associated with
alternative loss reduction strategies are summarised in Table 1 below. The annual capitalised
benefit is calculated by applying a discount rate of 3.5%.

4 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Flexible-Urban-
Networks-Low-Voltage/
5 https://publications.parliament.uk/pa/cm201617/cmselect/cmsctech/161/161.pdf
6 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Low-Carbon-London-

(LCL)/

8
 Table 1 - Capitalised value of the benefits associated with alternative loss reduction strategies

Strategy Capitalised value Comment

NOP optimisation £5.4-8.9m LPN area
HV smart £2.6-4.3m LPN area
switches
Multiple power SPN £48-80k Minimum for 30 ‘best’ HV feeders
factor correction EPN £56-92k per each licence area if power
per HV feeder LPN £53-88k factor is reduced from 0.85 to
0.95; the power factor is not yet
measured and hence potential
value might be lower of higher
Single point SPN £25-41k Minimum for 30 ‘best’ feeders per
power factor EPN £30-49k each licence area if single point
compensation per LPN £28-46k compensation is installed.
HV feeder Potential value depends on actual
power factor
Voltage control LV £1.5-2.4k per site Maximum expected value for
HV £9.2-15.2k per voltage dependent loads (constant
feeder power and constant impedance);
for mixes different types of loads,
i.e. constant power, constant
current and constant impedance
based loads, the benefits of
voltage control are marginal.

LV load balancing £0.9-13.6k per site LPN LV network

LV harmonics £200-300 per site
Primary Negligible For typical transformer load and
transformer de- no-load losses, the benefit is
energisation negligible; in the event of high no-
during low load load losses relative to load
conditions losses7, the potential benefit could
be £49-81k per substation
Eco-design £4-7.4k per transformer Average savings per transformer
transformers (392 transformers considered)
Amorphous £0.9-1.4k per PMT Average savings per PMT
transformers transformer (15 pole mounted
transformers (PMTs) considered)
Conductors LPN LV £63-104m All conductors lower than Al 185
rationalisation LPN HV £1.1-1.8m mm² are replaced with Al 185 mm²
EPN LV £114-188m conductors.
EPN HV £24-39m LPN HV network already uses
SPN LV £87-144m relatively higher conductor sizes
SPN HV £15-25m and hence benefit is relatively
lower than in EPN and SPN.

7 High no-load losses imply older transformers, which based on life expectancy, could reduce the
indicative value of the capitalised benefits as these might be replaced, based on condition, before the
full benefits are achieved.

9
 Strategy Capitalised value Comment
SPN HV voltage Min 11 kV £7.3-12m - SPN HV voltages 2.2, 3.3 and
rationalisation Min 20 kV £59-97m 6.6 kV are upgraded to 11 kV,
2,300 km of conductors
- All HV voltages are upgraded
to 20 kV, 17,700 km of
conductors
- Impact of transformers is not
taken into account
LPN EHV voltage £12-19m LPN 33 kV network is upgraded to
rationalisation 132 kV, 6,100 km of conductors;
Impact of transformers is not
taken into account
Smart distribution £7.4-12.3k per Minimum benefit per site
transformer8 secondary site [considering EPN 30 ‘best’ sites
for voltage control on LV network
(4% loss reduction), HV network
(5% loss reduction), power factor
improvement (8% loss reduction)
and phase imbalance reduction
(5% loss reduction)]
Scott connected £10.4-17.3k per site SPN LV networks supplied from
transformers 307 Scott connected transformers

Impact on Average savings on Savings are due to reduced active

transmission National Grid’s power on UK Power Networks
system networks of up to 5.5% regions. Control of reactive power
could be achieved could potentially generate
additional savings.

8Typically distribution transformers are equpted by off-load tap changers to adjust for a seasonal
variation in expected voltage range. Smart distribution transformers could control voltage during
operatioin in order to, for example, reduce losses.

10
1 Quantification of network losses

Comprehensive case studies have been carried out to quantify the impact that various losses-
drivers have on overall network losses. These drivers were ranked in terms of their
corresponding impact. On the basis of the Load Related Expenditure (LRE) model concept a
new modelling tool, Loss Operation & Investment Model (LOIM), has been developed and
applied for the first time, to the detailed quantification of losses in distribution networks within
the licence areas of UK Power Networks. Furthermore, LOIM has been applied to generate
Losses Heat Maps for UK Power Network’s EPN and LPN network licence areas. The
effectiveness of various loss-reduction techniques in different areas were analysed in detail.
The LOIM tool was also utilised to assess losses performance in different network types and
configurations in order to provide core evidence for creating representative networks based
on the parameters that drive their corresponding losses performance.
1.1 Loss Operation & Investment Model
The LOIM was developed to quantify losses in UK Power Networks’ licence areas. This model
was used to assess the impact of different demand and network parameters that are known
to influence losses (e.g. power factor, network phase imbalance, harmonics etc.), and assess
the loss-improvement impact of alternative loss-mitigation techniques (e.g. optimisation of
normally-open points) and investment strategies (size of cables, low loss technologies). The
main components of the LOIM model are presented in the flow diagram depicted in Figure 1.1.

Alternative network
operation loss-reduction
strategies

Loss Loss-related
Operation & networks
Selection of network Investment performance and
losses drivers Model benefit of alternative
loss-reduction
(LOIM) strategies

Alternative network
investment loss-reduction
strategies

Figure 1.1. Loss operation and investment model

The LOIM contains a detailed network model of the entire UK Power Networks area, in
contrast to previous modelling carried that was based on representative distribution networks
only.
The magnitude of losses in a given network depends on load and voltage profiles, demand
power factors, demand phase imbalance, and harmonics. In this project, network losses are
quantified and presented using detailed spatial resolutions in the form of loss heat maps,
described in the following section.

11
1.2 Losses Heat Maps
Losses heat maps were developed to identify regions of each of UK Power Networks’ licence
areas in which magnitude of losses are highest. Relevant data is obtained from the studies
carried out using the LOIM tool. Each licence area is split into squares of 500 x 500 metres.
Every Distribution Transformer (DT) in the licence area is associated with a particular square.
It should be noted that more than one DT could be associated with a particular square in which
case losses are the sum of losses associated with each DT. The magnitude of annual losses
is quantified for each 500 x 500 metre square across UK Power Networks’ licence areas. The
losses heat maps generated are shown in Figure 1.2, Figure 1.3 and Figure 1.4.

Figure 1.2: LPN and EPN service cables, low and high voltage networks and distribution transformer losses
density in MWh/year.km2

Some basic information related to network statistics for the area under consideration is
presented in Table 2.

12
 Table 2. LPN and EPN networks characteristics

LV and HV network LPN EPN

statistics
LV length, ‘000 km 22.6 48.7
DTs, ‘000 17.5 67.5
HV length, ‘000 km 12.0 38.5
Losses, % 3.0-4.8 3.6-5.4

Data related to the LPN and EPN LV and HV network characteristics in each area was
analysed. Figure 1.2 presents overall LV and HV loss densities per square kilometre (km2)
which is associated with unit area (500 x 500 m2). It can be concluded that higher loss
densities tend to be associated with urban and more densely populated areas or regions with
a significant amount of non-residential demand.

Figure 1.3: LPN and EPN service cables, low and high voltage networks and distribution transformer losses in
percentage terms

However, LV and HV loss expressed in relative terms (percentages) are more evenly
distributed, as can be observed in Figure 1.3. Relatively high level of percentage losses are
observed throughout the network. High loss percentages indicate areas of relatively low
network efficiency, which is also observed throughout the network. LV network percentage
losses, however, tend to be high in areas with a high level of loss density, as depicted in Figure
1.4.

13
 Figure 1.4: LPN and EPN low voltage network losses in percentage terms

It is clear that loss heat map representation of absolute and relative loss distribution and
intensity levels in the network provide a valuable visual aid in identifying key critical regions in
the system. For the first time, the analysis and quantification of network losses is based on a
detailed network model of the entire UK Power Networks areas, as opposed to previous
representations based on representative distribution networks only.
The next sections elaborate on various loss drivers, and how these influence losses. These
drivers require different interventions to reduce losses, and heat maps serve as a high-level
guide to steer network operators’ loss mitigation activities.

1.3 Analysis of losses

The LPN and EPN network data were used to assess losses in different network segments.
The results of this analysis are shown in Figure 1.5. As presented in Figure 1.5 (top right)
losses attributable to service cables and LV networks are larger in LPN, due to higher
utilisation levels. Total losses attributable to HV networks are higher in EPN than in LPN due
to longer network length. It can be observed that load-related losses in DTs are higher in LPN
while no-load losses are larger in EPN. Hence, transformer sizing could be enhanced in LPN,
while addressing fixed losses in EPN would yield the best benefits. Overall, the total absolute
losses are a bit higher in EPN, due to the greater HV losses.

14
In relative terms, however, HV network losses tend to be lower in the LPN area (0.69% min,
0.96% max), as compared to EPN (1.01% min, 1.76% max), as it can be observed in Figure
1.5 (top left). Relative LV network losses, including service cable losses, on the contrary tend
to be higher in LPN (LPN: 1.57% min, 2.73% max; EPN: 1.71% min, 2.39% max). The share
of LV losses tend to be higher in the LPN when compared to the EPN area (Figure 1.5).
EHV/HV, EHV, 132/EHV and 132 kV network losses represents up to 25% (in min case) or
less than 20% (in max case) of the overall losses. Given that more than 75% of total losses
are in LV and HV networks, focus of the analysis is on these segments of the network.
DT associated losses have a similar profile, both in absolute and relative terms (Figure 1.5),
with the dominance towards higher load related losses in LPN (LPN: 0.51% min, 0.71% max;
EPN: 0.43% min, 0.61% max), and higher no-load related losses in EPN (LPN: 0.27% min,
0.38% max; EPN: 0.48% min, 0.67% max).

7% 1,600
6% 1,400

Losses, GWh/year
5% 1,200
Losses, %

1,000
4%
800
3%
600
2% 400
1% 200
0% 0
Min Max Min Max Min Max Min Max
LPN EPN LPN EPN

100%
132 kV network
90%
80% 132/EHV transformers
70% EHV network
Losses, %

60%
50% EHV/HV transformers
40% High voltage network
30%
Distribution transformer no-load losses
20%
10% Distribtution transformer load losses
0% Low voltage network
Min Max Min Max
Service cable
LPN EPN

Figure 1.5: Comparison of LPN and EPN network losses

In the LPN licence area, annual losses are between 1,000-1,400 GWh/year. In EPN, annual
losses are between 1,100-1,500 GWh/year.
Further analysis was undertaken to understand how HV feeder length and utilisation affect
losses. Figure 1.6 shows that EPN feeders are longer, but that feeder losses in LPN are higher.
Hence, this analysis reveals that losses are predominantly driven by feeder lengths in EPN,
and that intensive asset utilisation drives losses in LPN.

15
 EPN LPN LPN EPN
80 70
70 60

Energy (GWh/year)
60 50
Length (m)

50
40
40
30
30
20 20
10 10
0 0
0 0.5 1 0 0.5 1
Percentage of feeders Percentage of feeders

Figure 1.6: Comparison of LPN and EPN HV losses for length (left) and energy (right)

With regard to DT peak utilisation rates (Figure 1.7): in the LPN area 19% of transformers
have a peak utilisation of 70% or more, while in the EPN area the share comes down to only
about 4%. This reinforces previous observations made on DT associated load and no-load
loss rates. Some further conclusions can also be drawn with respect to overall losses, in a
more detailed elaboration broken down by population density.

100%
Average DT peak utilisation LPN
LPN 51%
Peak utilisation (%)

80%
EPN 31% EPN
60%

40%

20%

0%
0% 20% 40% 60% 80% 100%
Percentage of distribution transformers

Figure 1.7. LPN and EPN DT Peak Utilisation

Networks in both LPN and EPN regions are categorised by population density (number of
customers per km2) and the relative losses attributable to differing levels of population density
are analysed. The applied classifications are <150 customers/km2, <750 customers/km2,
<5000 customers/km2, and >5000 customers/km2. The results of the analysis of the relation
between population distribution and losses are shown in Figure 1.8 and Figure 1.9. Here the
X-axis represents customer densities in LPN and EPN. The Y-axes represent total losses for
all squares with relevant customer density shown in Figure 1.2 in GWh/year and percentage
proportion of losses per voltage level, respectively. Losses in areas with a customer density
of 750 customers/km2 or higher are significantly pronounced in the LPN area. The EPN area
is predominantly characterised by losses associated to the areas with customer densities of
less than 150 customers/km2 and a customer density of between 750-5000 customers/km2.

16
 350
300
Losses (GWh/year)

250
200
150
100
50
0
< 150/km2 < 750/km2 < > < 150/km2 < 750/km2 < >
5000/km2 5000/km2 5000/km2 5000/km2
LPN EPN
LV Network Losses DT Load Losses DT no load losses HV Network Losses

Figure 1.8: Comparison of LPN and EPN network losses with population density

Figure 1.9 shows that losses, expressed in percentage values, are fairly evely distributed
across the various regions and classes. The proportion of HV network losses, however, is
greater in lower customer density areas. In those areas, customers are typically supplied from
shorter LV networks and relatively longer HV networks. This suggests that the loss-driver is
loading, rather than length of LV circuits. Proportions of DT losses tend to be lowest in EPN
with a customer density of between 150-750 customers/km2. It can be seen that DT losses
increase with increase in customer density as well as with decrease in customer density in
which case pole mounted transformers (PMTs) would be predominantly used. Hence, the use
of amorphous steel transformers in overhead networks with PMTs could be an economically
efficient way of managing DT no-load losses.

100%
90%
80%
70%
Losses, %

60%
50%
40%
30%
20%
10%
0%
< 150/km2< 750/km2 < > < 150/km2< 750/km2 < >
5000/km2 5000/km2 5000/km2 5000/km2
LPN EPN
LV Network Losses DT Load Losses DT no load losses HV Network Losses

Figure 1.9: Comparison of LPN and EPN network losses with population density

For networks where GMT are installed, distribution transformer losses increased
proportionately in line with customer density while HV network losses decrease. Hence, the
use of low-loss transformers across higher customer density areas to reduce losses could be
more economically efficient. However, for areas with less than 150 customers per km2
distribution transformers tend to be smaller which leads to an increase in distribution

17
transformer losses. Potentially, in these areas, losses could be mitigated by achieving a better
balance between load and no-load losses.
1.4 Impact of sample rate on losses modelling
Calculated losses vary depending on the sampling rate of the load measurements used in the
calculation. Depending on load variability, a different sampling rate might produce different
calculated losses. A service cable case study is carried out to analyse the impact of high load
variability on the losses calculation. About 5,000 daily load profiles are considered, each with
a sampling rate of one measurement per five seconds. For each daily five-second load profile,
daily losses are calculated for different service cable sizes. Following this, half-hourly profiles
are calculated by averaging the five-second profiles. Calculation of losses is repeated for half-
hourly profiles. For each profile, the ratio of daily losses calculated for the two sampling rates
are calculated and shown in Figure 1.10.

4
Losses ratio

0
0 1000 2000 3000 4000 5000
Profile index

Figure 1.10. Losses multiplier representing ratio of daily losses calculated with five-second and half-hourly
profiles

It can be seen that for a few profiles the losses ratio is greater than 4. In this case, losses
calculated using half-hourly profile could be as much as a quarter below the five-second
sample value. For the considered profiles the losses ratio was between 1.2 and 5.8. The
losses ratio average is about 1.9. To improve the accuracy of service cable losses calculations
when half-hourly or hourly profiles are used, a losses factor of 1.9 or similar is recommended.
1.5 Summary
In this study, analysis and quantification of network losses is based on a detailed network
model of all three of UK Power Networks’ licence areas, from low voltage networks to grid
supply points9. This is in stark contrast to the previous analysis of network losses that was
based on the application of representative distribution networks10.
Losses heat maps are developed to identify regions of each of UK Power Networks’ licence
areas in which magnitude of losses are most significant. Relevant data is obtained form the
studies carried out using the LOIM tool. It is clear that losses heat map representations of

9 Specifically, the scope of Loss Operation & Investment Model (LOIM) is extended to enable
quantification of network losses under different scenarios and loss-reduction strategies
10 Imperial College London and Sohn Associates, Management of electricity distribution network

losses, supported by WPD and UKPN, 2014

18
absolute and relative loss distribution and intensity levels in the network provide a valuable
visual aid in identifying key critical regions in the system.
Furthermore, the analysis carried out demonstrates that more than 75% of network losses are
associated with LV networks, HV networks and distribution transformers. Overall, 36-47% of
the total losses are in LV networks, 9-13% and 7-10% of losses are associated with distribution
transformer load related losses and no-load losses respectively, 17-27% are in HV networks
and finally 17-24% of total losses are in primary transformers, grid transformers, EHV
networks, and 132 kV networks. Understanding the contribution of different network sections
to the total losses will be important when analysing the cost effectiveness and potential impact
of different loss management strategies.
Service cable losses modelling needs to account for sampling-rate load variability. It is
recommended that a losses ratio of 1.9 is used in the first instance when half-hourly profiles
are used.

19
2 Distribution of losses across network segments

The aim of this section is to understand which loss-drivers were most significant in which types
of networks in order identify loss-reduction techniques that are likely to be most effective in
different types of network.

2.1 Overall network-level analysis

Figure 2.1 shows losses for more than 4,000 UK Power Networks radial HV feeders. It can be
seen that there is a relatively small number of HV feeders that have high losses. Losses on
some HV feeders could be up to about 1,000 MWh/year.

7% 1,000
Losses (%) Losses (MWh/year)
6%

Losses (MWh/year)
800
5%
Losses (%)

4% 600
3% 400
2%
200
1%
0% 0
0 1000 2000 3000 4000 5000
Number of feeders

Figure 2.1. UK Power Networks radial HV feeder losses

Figure 2.2 shows losses for LV networks. It can be seen that relatively fewer LV networks are
characterised with high losses. This could be used for prioritisation of LV networks for potential
losses reduction. LV network losses could be up to about 85 MWh/year.

Figure 2.2. LV network losses

20
2.2 LV network analysis
The relationship between the length of LPN LV circuits and their losses was analysed, with
the results shown in Figure 2.3. The X-axis shows the total length of LV networks supplied
from a distribution site which contains one or more distribution transformers. The Y-axis shows
the annual losses per distribution site. The correlation between LPN LV network losses and
circuit length is shown below. Generally, the losses increase with length of the networks,
although there are some relatively short networks characterised with high losses, and some
long networks with low losses. The former would typically have few customers with high loads,
and the latter many customers with low loads.

Figure 2.3: Correlation of LPN LV network losses and length per site

The scatter plot is very wide and two categories of LV networks could be considered:
 High losses and short networks

 Cases 1 and 3: single-connected customer at the end of a feeder with relatively high
loading will result in relatively high losses; for comparison, uniformly distributed load
for a large number of customers would results in one third of the losses
 Case 2: small numbers of connected customers with relatively high loading (about
double of the cases 1 and 3)
 Cases 4 (a, b and c): relatively high loading and longer network length for customers
of a) same type, b) two types and c) four customer types (domestic and non-domestic;
unrestricted and multi tariff),

 Low losses and long networks

 Case 5: Relatively low loading and high number of customers.

2.3 HV feeders with high losses

The LOIM was used to quantify losses on UK Power Networks HV feeders. An illustration of
two feeders of similar length is shown in Figure 2.4. Feeder 2 has slightly higher overall peak
load. However, Feeder 1 is characterised by 50% higher losses due to the long first section to
the first load point, while for Feeder 2 the loading is clustered in the middleof the feeder. Load
distribution is a key driver for losses on these two feeders.

21
 Primary 1

Primary 2

Figure 2.4: Illustration of two UK Power Networks HV feeders of similar length

UK Power Networks HV feeders are disaggregated into groups according to losses levels.
Assuming that losses mitigation measures have higher potential in feeders with relatively high
losses, the focus of the investigation is on those feeders. Figure 2.5 shows the UK Power
Networks HV feeders with losses greater than 550 MWh/year. The X-axis shows feeder
reference, the Y-axis shows losses in MWh/year and secondary Y-axis shows percentage of
losses. Feeders were analysed to understand drivers for relatively high losses and whether
there is any similarity between feeders.

1,600 7%
Losses (MWh/year) Losses (%)
1,400 6%

1,200
5%
Losses (MWh/year)

1,000

Losses (%)
4%
800
3%
600
2%
400

200 1%

0 0%

Feeder

Figure 2.5. UK Power Networks HV feeders with high losses (>550 MWh/year)

The EPN HV feeder with the highest losses is feeder E00112d76. It is 55 km long as denoted
and schematically illustrated in Figure 2.6. Total maximum peak demand is 4.7 MW and
annual losses are 1,422 MWh/year or 6%. It is relatively long feeder with two major branches.
On one of the major branches, relatively high load is located towards the end of feeder, which
results in relatively high losses.

22
 Figure 2.6. Schematic illustration of EPN HV feeder with the highest losses supplied from the Thaxted local
primary substation. Size of circles represent level of peak load. Total feeder length is 55 km.

Figure 2.7 shows LPN HV feeder EDNA005W0Y, which is characterised by relatively high
absolute volume of losses. The feeder length is 9.9 km and peak loading is 7.4 MW. It is
shorter than the one in Figure 2.6 but loading is greater. There is a long 3.5 km section of
feeder to the first load point, which is about a third of the total feeder length. 71% of feeder
losses are generated on this section alone. All load points are located towards the end of the
feeder which results in relatively high losses of 598 MWh/year or about 2%.

Figure 2.7. Illustration of LPN HV feeder with high losses supplied from Glaucus street primary substation. Size of
circles represent level of peak load. Red lines represent sections where NOP is located. High proportion of losses
are generated on first sections of feeder before the first load point connection.

Figure 2.8 shows SPN HV feeder, EDSO003ZGH, which has comparatively high losses. It is
relatively long (16.8 km) and characterised by a high load of 7.2 MW at peak. The annual
losses are 804 MWh/year, or about 2.5%. About 65% of total losses are generated in the first
two sections of the feeder.

23
 Figure 2.8. Illustration of SPN HV feeder with relatively high losses supplied from Crayford primary substation.

Each of the high losses feeders investigated have widely differing characteristics. It is
therefore difficult to derive standard templates that could enable simple feeder and network
classifications and easy understanding of effectiveness of different losses mitigation
measures.

2.4 Cluster-based network analysis

In order to draw general conclusions about different types of network, an analysis of the LV
and HV network dataset for UK Power Networks’ three licence areas was carried out to identify
network categories, or “clusters” of networks with similar characteristics, which can then be
used to represent different types of network. For example, it might be that many circuits fall
into the category of being a certain length, having a certain number of customers, where the
customers are mostly of a certain Elexon load type. It was anticipated that this clustering could
be used to identify types of network in different licence areas and to identify types of loss
interventions that may be beneficial for different network types.
The dataset contains 30,470 Pole Mounted Transformer (PMT) networks and 53,198 Ground
Mounted Transformer (GMT) networks. The following parameters are used on a per-DT basis:
total transformer rating and maximum demand, total overhead and underground circuit length,
total number of connected customers and their mix in terms of the Elexon profiles (DU, DR,
NDU, NDR), and area type (rural, semi-rural, semi-urban, and urban). A neural networks-
based self-organising map was used to disaggregate networks into 16 clusters for PMT
networks and 16 clusters for GMT networks.
The feeder parameters considered are: feeder length, number and rating of pole and ground
mounted distribution transformers, load distribution along feeder, feeder losses. All
parameters are split onto main and latter part of feeder.
A Self Organising Map (Neural Networks) approach is used to disaggregate feeders into
clusters. Networks supplied from pole (PMT) and ground mounted transformers (GMT) are
considered separately and they are split into 16 clusters each. For PMT 10 clusters are
selected and for GMT 12 clusters. Table 3 shows overall characteristics of networks
aggregated into different clusters. For example, in cluster RN1 there are 2,063 distribution
transformers and LV networks. Total DT rating is about 206 MVA and maximum demand about
100 MVA. Total LV network OH and UG length is about 582 and 406 km respectively. Total
connected domestic unrestricted and multi-tariff customers are 14,173 and 14,303,

24
respectively. Total connected non-domestic unrestricted and multi-tariff customers are 2,033
and 2,475, respectively. The area type is predominantly rural. Most of UK Power Networks’
licence areas are contained in three clusters: RN17, RN15 and RN8, to which more than 6,000
LV networks are assigned.

Table 3. Characteristics of network clusters; DT: distribution transformer, LV: low voltage, OH: overhead line, UG:
underground cable, DU: domestic unrestricted, DR: domestic multi-tariff, NDU: non-domestic unrestricted, NDR:
non-domestic multi-tariff, area type index 1: rural, 2: semi-rural, 3: semi-urban, 4: urban

Total LV Total LV
Total DT Avg
Circuit Circuit Number of
Number of Total DT Maximum Number of DU Number of DR Number of NDU Area
Cluster OH UG NDR
DTs Rating kVA Demand Customers Customers Customers Type
Length Length Customers
kVA Index
km km

RN1 2,063 205,985 99,960 582 406 14,173 14,303 2,033 2,475 1.2
RN2 604 133,965 59,240 135 198 5,601 5,301 990 1,054 1.3
RN3 2,463 61,630 10,622 340 95 4,607 3,672 955 1,020 1.1
RN4 1,623 162,225 41,165 682 500 34,580 34,327 1,941 2,744 1.5
RN5 3,191 1,597,216 604,496 255 5,657 185,038 484,828 10,744 16,706 2.6
RN6 4,645 2,321,650 970,018 103 9,585 771,312 157,930 46,633 12,086 2.1
RN7 5,821 582,100 117,267 1,363 989 29,042 23,357 5,709 5,587 1.1
RN8 6,004 300,200 59,796 1,096 562 16,990 13,736 3,588 3,679 1.1
RN9 3,433 1,711,240 1,132,356 80 4,249 185,547 132,749 34,428 16,724 2.4
RN10 2,062 1,028,975 816,695 9 5,455 587,467 57,203 44,004 5,491 3.1
RN11 2,383 693,326 314,189 265 3,227 102,634 277,267 4,202 6,424 2.5
RN12 2,007 2,013,400 716,050 16 2,608 113,492 147,576 20,975 20,022 2.5
RN13 744 149,130 38,432 330 385 24,176 20,329 1,242 1,431 1.6
RN14 4,896 771,075 284,192 592 2,767 86,074 87,124 6,099 5,918 1.4
RN15 6,549 2,008,767 528,835 755 5,452 219,110 179,842 13,533 12,511 1.6
RN16 2,083 417,680 81,984 428 481 11,468 9,554 2,845 2,811 1.2
RN17 8,897 4,448,595 1,008,228 452 8,946 388,951 278,113 40,834 29,403 1.8
RN18 3,529 2,773,730 559,405 42 3,249 146,259 138,333 23,082 16,540 2.3
RN19 3,005 2,370,000 1,085,088 13 2,696 107,821 87,508 36,220 14,657 3.0
RN20 2,570 128,513 58,535 523 254 7,366 6,929 1,532 1,824 1.1
RN21 1,485 1,180,400 470,277 6 3,361 374,799 49,859 32,045 5,071 3.1
RN22 2,925 44,854 12,160 264 83 4,209 3,317 775 774 1.1

Table 4 presents characteristics of selected neural network-derived clusters.

Table 4. Characteristics of selected pole (PMT) and ground mounted transformer (GMT) clusters; OH: overhead
line, UG: underground cable, D: domestic, ND: non domestic, DU: domestic unrestricted, DR: domestic multi-
tariff, NDU: non-domestic unrestricted, NDR: non-domestic multi-tariff

Cluster Mounting Description

RN1 PMT Rural, OH:UG=1.5:1, D:ND=6:1
RN2 PMT Rural, OH:UG=1:1.5, D:ND=5.5:1
RN3 PMT Semi-rural, UG, D:ND=16:1, DU predominantly
RN4 PMT Semi-rural, OH:UG=1.5:1, D:ND=15:1
RN5 GMT Semi-urban, UG, D:ND=24:1, DU:DR=1:2.5, NDU:NDR=1:1.5
RN6 GMT Semi-rural, UG, D:ND=16:1, DU:DR=5:1, NDU:NDR=4:1
RN7 PMT Rural, double rating and maximum demand compared to cluster 1,
OH:UG=1.5:1, D:ND=5:1
RN8 PMT Rural, OH:UG=2:1, D:ND=4:1
RN9 GMT Semi-rural, UG, D:ND=6:1, NDU:NDR=2:1
RN10 GMT Urban, UG, D:ND=14:1, NDU:NDR=8:1

25
 Cluster Mounting Description
RN11 GMT Semi-urban, OH:UG=1:12, D:ND=36:1
RN12 GMT Semi-urban, UG, D:ND=6:1
RN13 PMT Semi-rural, OH:UG=1:1, D:ND=25:1
RN14 GMT Rural, OH:UG=1:5, D:ND=14:1
RN15 GMT Semi-rural, OH:UG=1:7, D:ND=15:1
RN16 PMT Rural, OH:UG=1:1, D:ND=3.5:1
RN17 GMT Semi-rural, OH:UG=1:20, D:ND=10:1
RN18 GMT Semi-rural, UG, D:ND=7:1
RN19 GMT Semi-urban, UG, D:ND=4:1, NDU:NDR=2.5:1
RN20 PMT Rural, OH:UG=2:1, D:ND=4:1
RN21 GMT Urban, UG, D:ND=11:1, DU:DR=8:1, NDU:NDR=6:1
RN22 PMT Rural, UG, D:ND=6:1

For example, cluster RN1 is characterised with the ratio of overhead lines to underground
cable lengths is 1.5:1 i.e. overhead lines are 50% longer than underground cables. Domestic
customers are predominantly connected with a domestic to non-domestic customer ratio of
6:1. Average parameters for each cluster are used to create representative networks. Figure
2.9 shows representative network annual losses.

900
800 SC losses LV losses DT LL DT NLL HV losses
700
Losses, MWh

600
500
400
300
200
100
0

Figure 2.9. Breakdown of annual losses for representative networks

The highest losses are expected on average in rural network types, RN1 and RN2, and in
those networks supplying residential areas as well as in semi-urban network types also
supplying residential areas. All three network types are supplied from pole mounted
transformers.
Figure 2.10 shows percentage of losses for representative networks.

26
 7%
6% SC losses LV losses DT LL DT NLL HV losses
5%
Losses, %

4%
3%
2%
1%
0%

Figure 2.10. Breakdown of annual losses in percentages for representative networks

RN7, RN3 and RN4 have relatively high losses compared to the energy transported. All
network types are supplied from PMTs and are predominantly domestic. This is consistent
with the observation that HV losses are relatively high in those networks.
Figure 2.11 shows the relative share of loses in representative networks across the network.

100%
90%
80%
70%
Losses, %

60%
50%
40%
30%
20%
10%
0%

SC losses LV losses DT LL DT NLL HV losses

Figure 2.11. Losses source for clusters

Given the greater customer and load density in urban areas, greater losses per square
kilometre were observed. In urban areas, LV network losses are dominant while in rural areas
HV network losses are dominant. RN22, RN13 and RN16 type networks are characterised
with relatively short networks and hence distribution transformer losses are dominant.
Representative networks describe averages very well, but do not appropriately describe
extreme cases, i.e. feeders with very high losses that may be targeted for loss reduction.
Hence for these cases it is more beneficial to analyse real rather than representative networks.

27
2.5 Summary
Network losses are strongly influenced by demand density and circuit length, because of this
detailed analysis of network loading and correlation with feeder lengths in different areas is
carried out. UK Power Networks operate a wide range of network types, from rural areas, as
in parts of Norfolk and Suffolk, to very densely populated urban areas like London. These
areas have corresponding peak demand density ranging from very low (0.05 MW/km2) in rural
areas to very high density (137 MW/km2) in urban areas. In this context, the average utilisation
of distribution transformers varies between 51% and 38% in LPN and EPN areas respectively.
Furthermore, the analysis carried out demonstrated that 20% of distribution transformers in
LPN are characterised by peak utilisation greater than 70%. On the other hand, only about
4% of distribution transformers in EPN area are highly loaded.
Detailed power flow modelling revealed that feeders in the LPN area deliver on average 50%
more energy than in the EPN area, while the circuits in LPN region are on average about 60%
shorter than in the EPN. This analysis demonstrated that the losses in LPN area are primarily
driven by high network utilisation, while in EPN losses are driven by long feeder lengths.
Overall, losses in LV networks are comparable (for the minimum case) in both areas or slightly
greater in LPN area (for the maximum case) even though the LPN LV network is significantly
shorter but characterised by higher loading. On the other hand, losses in HV networks are
greater in the EPN area.
The analysis also demonstrated that the magnitudes of losses vary significantly across the
networks. A relatively small number of feeders are accountable for the majority of losses.
Detailed modelling demonstrated that 20% of feeders are responsible for 70% of HV network
losses. Similarly, more than 50% of losses in LV networks are in only 20% of feeders. This
clearly demonstrates that the loss reduction schemes in HV networks needs to target only a
relatively small proportion of the feeders characterised by high losses, in order to achieve cost-
effective loss mitigation.
Based on an advanced neural networks methodology, UK Power Networks’ networks are
classified into 22 clusters according to the number of connected customers and their mix, LV
network length and construction, and distribution transformer mounting, rating and loading.
Average parameters for each cluster are quantified and corresponding representative
networks created, ranging from rural to urban networks and the related loss performance
assessed. However, as the significant amount of losses turned out to be associated with a
relatively small number of very specific feeders, generic feeders with average parameters may
not provide appropriate evidence to inform the development of effective losses reduction
strategies.

28
3 Identification of potential operational strategies for loss
reduction

3.1 Optimisation of Normally Open Point locations

HV and LV distribution networks are typically operated with a radial topology. However, in
order to minimise customer minutes lost under fault conditions, the networks are designed in
such a way that the load can be supplied from another adjacent feeder via a Normally Open
Point (NOP) switch. In this study, we investigated the benefits of optimising the NOP locations
to minimise network losses.
We then compared the losses in the original network and the losses in the network with
optimised locations of NOP switches. The study is performed on an actual UKPN HV network
with the key parameters shown in Table 5.

Table 5 Key data of the network used in the study

Primaries 114
Feeders 2,286
Nodes 31,063
Branches 29,676
Total lines length (km) 7,826
Loads 17,809
Total Load (MVA) 5,537
NOPs 2,828

The optimisation of NOP locations results in the closer to average loading of each feeder,
leading to lower losses. This is demonstrated in Figure 3.1 where the loading of all 2,286
feeders in the basecase with the orginal NOPs and the case with optimised NOPs is
presented.
In the case with the original NOP positions, the loading of feeders varies in a larger range
compared to the loading of feeders in the case with optimised NOP locations. This explains
the larger network losses in the base case since losses are a quadratic function of network
loading. By optimising the locations of NOP switches, the system loads can be allocated more
evenly to the feeders resulting in lower losses. Some of the spare feeders, which were not
initially loaded, might become loaded after the proposed optimal NOP locations are deployed.
This approach is used to minimise losses by determining the optimal location of NOPs.

29
 20
Base NOP Optimal NOP
18
16
14
Load, MVA

12
10
8
6
4
2
0
0 500 1000 1500 2000
Number of feeders

Figure 3.1 HV feeder loading in the base case and the optimised case.

The optimisation of NOP locations also changes the variation of the feeder lengths. The feeder
length expressed in km for the feeders analysed in the study is shown in Figure 3.2. The range
of variation is lower when compared to the non-optimised case. This also explains the reduced
losses in the optimised case.

25 Base NOP Optimal NOP

20
Length, km

15

10

0
0 500 1000 1500 2000 2500
Feeder

Figure 3.2 HV feeder length in the base case and the optimised case.

Figure 3.3 presents the losses on all feeders under the base case and the optimised case.
The results for the optimised case (corresponding to the brown line) express the number of
feeders having losses above the level indicated in the Y-axis e.g. around 14% (about 300 out

30
of about 2,100 feeders with losses greater than zero) of the feeders exhibit losses higher than
50 MWh/year.

450
Base Optimal NOPs
400
Feeder losses (MWh/year)

350
300
250
200
150
100
50
0
0 500 1000 1500 2000
Number of feeders

Figure 3.3. HV feeder losses in the base case and the optimised case.

The overall losses in the system are presented in Table 6, which demonstrates that the
optimisation of the NOP locations leads to a 17% reduction of losses in the network.
Table 6.Reduction of total losses in the optimised case.

Base Case Optimised Case Reduction

Losses (MWh/year) 67,398 56,253 17%

The above findings imply that the determination of the locations of NOP should consider the
effect of losses. The simple approach proposed in this study yields a 17% reduction of total
losses if the locations of the NOP switches are optimised. The equivalent capitalised value is
between £5.4-8.9m.
With smart HV network switched potential for losses reduction might be even greater and
achieve meshed network losses of 50,839 MWh/year which yields additional losses reduction
of 8% and the equivalent value between £2.6-4.3m.

3.2 Power factor correction

Since both active and reactive power flows contribute to the overall losses in an AC network,
a useful strategy in reducing losses lies in minimising the reactive power load. In this context,
the following study investigates the impact of installing power factor correction to reduce the
reactive power load.
Since the exact value of the present power factor in the actual networks employed in these
studies is not accurately known, in order to comprehensively evaluate the benefits of power
factor correction, a number of studies assuming four different levels of power factor
improvement are carried out. The first scenario assumes that the base (present) power factor
is 0.95 and after the deployment of power factor correction, the power factor is equal to 1. The
second, third and fourth scenario assume that the reference power factor is 0.9, 0.85 and 0.8

31
respectively. For each scenario, the network losses in the base case are compared against
the losses in the case with power factor correction. The study is carried out for real UKPN LV
and HV networks. It should be noted that the observed losses reduction does not include
improvement of losses at all upstream levels. Hence, shown losses reduction represent
conservative value and, hence, the same losses reduction in LV networks would have greater
impact then losses reduction on HV networks.

40%
35%

Losses reduction (%)

30%
25%
20%
15%
10%
5%
0%
0.95->1 0.9->1 0.85->1 0.8->1
Power factor improvement

Figure 3.4. Impact of power factor correction on LV and HV networks

The reduction of losses under each of these scenarios is presented in Figure 3.4. The results
demonstrate that power factor improvement can achieve very significant reduction of losses,
which, depending on the scenario, can reach up to 29% reduction in LV networks and 36%
reduction in HV networks. As expected, the benefit is higher when the power factor of the base
case is lower. However, even if the base power factor is 0.95, the achieved reduction of losses
is still substantial, i.e. 7% in LV networks and 10% in HV networks.
In order to further investigate the benefits of power factor correction on different UKPN
networks, studies have been carried out on selected SPN, EPN and LPN feeders. For each
type of feeders, two power factors are considered, i.e. 0.85 and 0.95. The benefits of improving
the power factor from 0.85 to 0.95 in each type of network are shown in Figure 3.5, Figure 3.6
and Figure 3.7.
The graphs are explained as follows. The x-axis shows HV feeders used in the studies; the
HV feeders are ranked according to the reduction of peak losses (in absolute values) - from
the highest to the lowest- achieved by the deployment of power factor correction. In this way,
the study can help network planners to determine priority areas where power factor correction
strategies would have significant benefits. The 1st y-axis (left) denotes the annual losses in
MWh/year. The brown and orange lines denote the losses for the case with 0.85 and 0.95
power factor respectively. The relative reduction in losses due to power factor improvement
from 0.85 to 0.95 is expressed by the green line (referring to the 2nd y-axis).

32
 2,000 30%
PF 0.95 PF 0.85 Reduction, %
1,800
25%
1,600

Losses reduction (%)

Losses (MWh/year)

1,400
20%
1,200
1,000 15%
800
10%
600
400
5%
200
0 0%

Figure 3.5. List of SPN feeders ranked by reduction of losses achieved by power factor improvement.

The study on the SPN feeders (Figure 3.5) demonstrates that the range of losses reduction
that can be expected is between 20% and 25%. The results demonstrate that prioritising the
deployment of the power factor correction to feeders with high losses (in absolute values)
would be most beneficial. However, even for the feeders with lower absolute losses (right hand
part of the graph), the level of losses reduction is still significant. If the power factor is improved
from 0.85 to 0.95, the corresponding losses reduction for the above 30 feeders is more than
50 MWh/year. It should be noted that actual power factor is unknown and it is recommended
to investigate further the above selected HV feeders.
The results for the EPN feeders (Figure 3.6) offer the same kind of conclusions. The range of
losses reduction is between 20% and 27% and this reduction is consistent across all feeders
irrespectively to the level of their absolute annual losses. The losses reduction for the specified
30 feeders is more than 115 MWh/year.

2,000 30.0%
1,800 PF 0.95 PF 0.85 Reduction, %
25.0%
1,600
Losses reduction (%)
Losses (MWh/year)

1,400
20.0%
1,200
1,000 15.0%
800
10.0%
600
400
5.0%
200
0 0.0%

Figure 3.6. List of EPN feeders ranked by reduction of losses achieved by power factor improvement.

For the LPN feeders, the losses reduction seems to be consistent around 20%.

33
 2,000 30.0%
PF 0.95 PF 0.85 Reduction, %
1,800
25.0%
1,600

Losses reduction (%)

Losses (MWh/year)

1,400
20.0%
1,200
1,000 15.0%
800
10.0%
600
400
5.0%
200
0 0.0%

Figure 3.7. List of LPN feeders ranked by reduction of losses achieved by power factor improvement.

We can conclude that the deployment of power factor correction on the UKPN networks could
potentially achieve significant benefits in terms of reduction in network losses. The study
shows that the expected loss reduction lies between 20% and 27% if the power factor can be
improved from 0.85 to 0.95. Losses are reduced at least about 110 MWh/per for each of the
above 30 feeders.
Single point power factor compensation
In order to improve the power factor, investing in reactive power compensation may be
beneficial. The location and the size of the compensation are critical factors to be considered
in order to maximise the benefit of this investment. In this study, we have employed
mathematical analysis to determine the optimal location and size of the reactive power
compensation under certain assumptions.
Figure 3.8 illustrates a feeder with uniformly distributed load along its length and the change
in current due to the deployment of compensation. This model is used for deriving the location
and size of the compensation achieving the lowest losses.
Flow, p.u.

i-i
c

Length, p.u.
0

0 1
x dx

Figure 3.8. Illustration of radial feeder and change in current due to deployment of compensation.

34
In this study, we assume that the distribution network is operated in radial topology and the
load is distributed uniformly across the feeder. In this case, the feeder losses without any
reactive compensation can be formulated as follow:

(1)

Considering reactive compensation, the formula can be re-written as follows:

(2)

Thus, the losses reduction can be derived from the difference between (1) and (2):

(3)

In order to determine the optimal location and size of the single point reactive compensation,
the optimality condition to the 1st derivative of (3) are derived. Based on this approach, the
optimal location of the reactive compensation is at 2/3 of the feeder length (with the substation
used as the reference point) and the optimal size of the compensation is 2/3 of the reactive
power load.
This strategy is then analysed, i.e. installing a single point reactive compensation in this
optimal location and with this optimal size, on selected HV UKPN feeders in LPN, SPN, and
EPN areas. For each type of feeders, two cases are simulated, i.e. a base case without
compensation (assuming a power factor of 0.85) and a case with a single point reactive
compensation with the above optimal characteristics.
The results corresponding to LPN feeders are presented in Figure 3.9. The x-axis shows the
feeder IDs (sorted from high to low losses feeders) and the y-axis shows the annual feeder
losses for the base case (corresponding to the orange line) and the case with optimised
reactive compensation (corresponding to the brown line). The relative losses reduction is
expressed by the green line. The results clearly indicate the significant benefits of reactive
compensation as the losses at all feeders considered in this study decrease by more than
12%. This is equivalent to capitalised benefits between £28-46k for each of identified 30
feeders.

35
 3,000 13.0%

PF 0.85 Single point compensation Reduction, %

2,500
12.5%

Losses reduction (%)

Losses (MWh/year)

2,000
12.0%
1,500
11.5%
1,000

11.0%
500

0 10.5%

Figure 3.9 List of LPN feeders ranked by reduction of losses achieved by reactive power compensation.

In a similar fashion, the results of the study corresponding to the SPN feeders are presented
in Figure 3.10. The results demonstrate that the deployment of reactive compensation can
reduce losses about 12%. The capitalised value of savings for the first 30 feeders could be
between £25-41k per feeder.

3,000 18%
PF 0.85 Single point compensation Reduction, %
16%
2,500
14% Losses reduction (%)
Losses (MWh/year)

2,000 12%

10%
1,500
8%

1,000 6%

4%
500
2%

0 0%

Figure 3.10 List of SPN feeders ranked by reduction of losses achieved by reactive power compensation.

36
The results of the study correspond to the EPN feeders are presented in Figure 3.11. The
improvement in losses varies between 12% and 14%, with corresponding capitalised value
per feeder being between £30-49k.
3,000 18%
PF 0.85 Single point compensation Reduction, % 16%
2,500
14%

Losses reduction (%)

Losses (MWh/year)

2,000 12%

10%
1,500
8%

1,000 6%

4%
500
2%

0 0%

Figure 3.11 List of EPN feeders ranked by reduction of losses achieved by reactive power compensation.

Based on the presented results, it is clear that the deployment of reactive power compensation
of suitably designed size and location could potentially achieve significant reduction of network
losses, ranging from 12% to 14% in the examined studies.
Based on the above simple analysis, the optimal location is at 2/3 of the feeder length (with
the substation used as the reference point) and the optimal size of the compensation is 2/3 of
the reactive power load, for feeders with uniformly distributed load. Given that the load
distribution may be different in reality, further analysis is required to determine the optimal
location and size of reactive compensation.

3.3 Voltage control driven loss reduction

Another strategy for reducing losses investigated in this project is related to voltage control
aimed at reducing electricity load. The rationale behind this strategy lies in the fact that some
electricity loads are voltage dependent, such as motors, resistive heating, etc. By lowering the
voltage, the power demand of may reduce, leading to reduction in network losses. The voltage
control technology can be implemented at the system level by optimising position of tap-
changing transformers or through the application of voltage optimisation technologies (e.g.
powerPerfector) at the customer connection points.
For illustrating the potential impact of voltage control on network losses, studies have been
carried out assuming that electricity demand is composed of constant power, constant current
and constant impedance type loads. The study is performed on LV networks and HV feeders
in EPN area. Losses in the base case (without voltage control) and the case with optimised
voltage are presented in Figure 3.12 for LV networks and in Figure 3.13 for HV feeders.
The x-axis shows the feeder IDs (sorted from high to low losses feeders) and the y-axis shows
the annual feeder losses for the base case (corresponding to the orange line) and the case
with optimised voltage level (corresponding to the brown line). The results indicate that the

37
benefits achieved by voltage control are relatively modest (around 5%). The corresponding
capitalised value of loss reduction per site is between £1.5-2.4k. It should be noted that this
corresponds to maximum benefit as calculation is conducted assuming constant impedance
based demand model. For example, if voltage dependent load could be represented by one
third constant power, one third constant current and one third constant impedance than
changes in losses would be marginal.

140 9%
Base Voltage control Losses reduction, %
8%
120
7%

Losses reduction, %
Losses, MWh/year

100
6%
80 5%

60 4%
3%
40
2%
20
1%
0 0%

Distribution site reference

Figure 3.12 The annual EPN LV network losses with and without voltage optimisation and maximum losses
reduction in %; sites are ranked by losses reduction in MWh/year; assumed load model is constant impedance

3,000 80
Base Optimised voltage Reduction
70

Losses reduction, MWh/year

2,500
60
Losses (MWh/year)

2,000
50

1,500 40

30
1,000
20
500
10

0 0

Figure 3.13 List of EPN HV feeders ranked by reduction in losses achieved by voltage control.

The analysis carried out demonstrates that the deployment of voltage control could deliver
benefits in terms of losses reduction by exploiting the voltage dependency of certain loads. In
the presented studies, these benefits are moderate (around 5%). Given the assumptions
related to voltage dependency of demand the capitalised value of losses reduction per each
of the identified 30 feeders is between £9.2-15.2k.

38
3.4 Balancing load across phases
In this section, we analyse the impact of balancing the load across different phases on
reduction in network losses. As most of end-of-use appliances constitute single phase loads,
the loading of different phases can be quite unbalanced, especially in LV networks. The
installation of small-scale DG can aggravate the phase imbalance problem. The presence of
imbalanced demand increases losses as the flows are not evenly distributed across phases.
In addition, load imbalances also trigger current flowing through the neutral conductor which
contributes further to losses. The imbalance problem tends to be less pronounced in HV
networks due to load diversity; for this reason, this analysis is focused on the LV networks. In
this context, modelling has been carried out to investigate the opportunities for loss reduction
by improving the load distribution across three phases through optimisation of transformation
points, e.g. modern power electronic-based voltage regulators which can be combined with
the application of power factor compensation and phase balancing.
The study was carried out on actual UKPN LV networks; the length of the LV circuits
considered in this study is given in Table 7 below.
Table 7 Length of installed LV network

Length (‘000 km) EPN LPN SPN Total

LV Cable 133 37 55 225
LV Overhead Line 90 57 146
Total 222 37 111 371

Losses in LV networks have been quantified under three scenarios: (i) balanced load, (ii) 10%
imbalance load and (iii) 30% imbalance load. The load flows and losses for each scenario are
recorded and analysed. The results are presented in Figure 3.14.

90 30
0% 10% 30%
80
LV network losses (MWh/year)

LV network losses reduction

Reduction (10%) Reduction (30%) 25
70
60 20 (MWh/year)
50
15
40
30 10
20
5
10
0 0

Site reference

Figure 3.14. Impact of load balancing on the LPN LV network losses.

The x-axis shows the LV sites IDs (sorted from high to low losses sites). The left y-axis
presents the annual network losses for the three scenarios, while the right y-axis shows the
achieved losses reduction by correcting a 10% and 30% imbalance. The results indicate that
the benefits of such correction are up to 5% (=2.8/55.7) and 45% (=25.0/55.7) respectively,
highlighting the importance of this strategy in reducing losses. The minimum value of losses
reduction per site LV network shown in Figure 3.14 is between £90-£1,360 per year. Further

39
analysis is required to understand in-depth the actual level of imbalance in the UKPN network
through suitable measurements and consequently identify priority areas to deploy this
strategy.

3.5 Harmonics
3.5.1 Impact of harmonics on transformer’s and Joule losses
The increased penetration of load appliances and DG with power electronics in the distribution
network tends to increase generated harmonics. The resulting presence of multiple
frequencies will increase heating in the equipment and conductors as well as power quality
problems. In this context, we analyse the impact of increased voltage harmonics on losses,
particularly on transformer (iron) no-load losses and thermal losses. The impact of total
harmonic distortion (THD) on losses is illustrated in Figure 3.15.

1.1
Joule losses
1.08
Fe losses
Losses factor

1.06

1.04

1.02

1
0% 2% 4% 6% 8% 10%
THD, %

Figure 3.15. Impact of THD on Fe and Joule losses

The graph shows that increased THD increased transformer no-load losses linearly. For
example, 10% THD will cause losses to increase by 10%. THD also increases the Joule losses
to a lower extent and in a non-linear fashion. Therefore, it would be beneficial to keep the THD
in the system as low as possible.
According to the current standards, the maximum limit of THD at distribution transformer is
5% for distribution transformers, 4% for primary substations and 3% for EHV networks11. This
means that the increase in joule losses is already capped to be modest (<0.2%), but there is
a significant opportunity for reducing the Fe losses. A possible strategy is to use transformers
with low non-load losses. Another strategy lies in installing filters that can reduce harmonics
in the system, e.g. technologies such as powerPerfector, which would optimise voltage and
simultaneously reduce harmonics.

3.5.2 LV Network Harmonics

The insights from section 3.5.1 motivate an estimation of the impact of filtering harmonics on
the losses of selected UKPN LV sites which are shown on the x-axis of Figure 3.16 and are
ranked based on the annual losses (from high to low). Two cases are illustrated in the graph:
(i) network losses based on the assumption that there are no harmonics (expressed by the
orange line) and (ii) losses assuming 10% THD (expressed by the blue line). The green line
expressed losses reduction achieved by filtering harmonics. It should be noted that this study
considers only the thermal losses in LV network circuits.

11ENA ER G5, Energy Network Association Engineering Recommendation (ENA ER): Planning
Levels For Harmonic Voltage Distortion

40
 60 600
0% 10% Reduction
LV network losses (MWh/year)

LV network losses reduction

55 500

50 400

(kWh/year)
45 300

40 200

35 100

30 0

Site reference

Figure 3.16. Impact of reducing total harmonic current distortion on losses at LV networks.

The results of the study indicate that improving the THD can offer a modest reduction of losses
of around 1%. The equivalent capitalised value of improving THD per each site LV network is
between £200-300.

3.5.3 Long HV Feeders, Harmonic Resonance

Harmonic resonance appears in the electricity network when the natural frequency coincides
with the frequency of harmonic current. As the network contains ferrous components that are
magnetically energised, the network itself can be the source of harmonics, which increase
losses. In this study, we estimate the resonant frequency for different HV cable types (1-core
and 3-core), lengths (5 to 30 km), and cross-sectional areas (70 - 630 mm2).
For this study, we use data for Prysmian 11 kV 1-core and 3-core cables and assume a
lumped-capacitance cable model. The estimated resonant frequencies for different lengths
and diameters of these cables are presented in Table 8 and Table 9 for 1-core and 3-core
cables respectively. It shows that the resonant frequency is lower for longer cables and greater
cross-sectional areas.
Table 8 Approximate resonant frequency for different lengths and cross-sections (Prysmian 11 kV 1-core).

Reactance at Maximum Approximate resonant frequency, Hz, for different

CSA 50 Hz, capacitance, cable lengths, km
milliohms/km nanofarad/km 5 10 15 20 25 30
70 130 288 2,916 1,458 972 729 583 486
95 123 323 2,831 1,415 944 708 566 472
120 118 353 2,764 1,382 921 691 553 461
150 117 380 2,676 1,338 892 669 535 446
185 112 416 2,614 1,307 871 653 523 436
240 109 460 2,520 1,260 840 630 504 420
300 105 506 2,448 1,224 816 612 490 408
400 101 561 2,370 1,185 790 593 474 395
500 99.8 619 2,270 1,135 757 567 454 378
630 96.1 697 2,180 1,090 727 545 436 363

41
 Table 9 Approximate resonant frequency for different lengths and cross-sections (Prysmian 11 kV 3-core)

Reactance at Maximum Approximate resonant frequency, Hz, for different

CSA 50 Hz, capacitance, cable lengths, km
milliohms/km nanofarad/km 5 10 15 20 25 30
70 108 298 3,145 1,572 1,048 786 629 524
95 102 334 3,057 1,528 1,019 764 611 509
120 98.8 365 2,971 1,485 990 743 594 495
150 96.2 392 2,905 1,453 968 726 581 484
185 93.1 430 2,820 1,410 940 705 564 470
240 90 476 2,726 1,363 909 681 545 454
300 87.4 524 2,636 1,318 879 659 527 439
400 84.9 580 2,542 1,271 847 636 508 424

For 15-20 km HV cables, the harmonic resonant frequency might be lower than 700 Hz at
which point it is recommended to carry out a detailed harmonic study. This parallel resonance
is characterised by low impedance to the flow of harmonic currents at the resonant frequency
which contributes to higher losses12. Therefore, for a long HV feeder, e.g. network connection
to a wind farm, there may be a case for analysing the harmonic resonance.

3.6 Upgrading single phase spur to three phase

Another possible strategy to reduce losses lies in upgrading the single-phase spur to a three-
phase one. In this section, we demonstrate theoretically that the implementation of this
strategy can reduce losses significantly. For the purpose of this study, 95 mm 2 Al 11 kV
conductor is analysed assuming that the neutral path is characterised by the same resistance
as the phase conductor. The losses for the single-phase and three-phase spur are presented
in Figure 3.17. It should be noted that the upstream losses reduction due to imbalance and
load reduction are not included in the analysis.

90 Single-phase
80
Three-phase
Losses (MWh/year)

70
60
50
40
30
20
10
0
0 0.5 1 1.5
Peak load (MW)

Figure 3.17. Impact of spur upgrading on the LV network losses

Based on the presented results, upgrading the single-phase spur to a three-phase one
achieves a very substantial losses reduction of 83%. This is attributed to the fact that after the
upgrade more conductors are used to transfer the same energy. However, the cost of such a
strategy may be high and therefore its implementation on the spur with high peak demand
should be recommended.

12 UK Power Networks, Business plan (2015 to 2023) Annex 7: Losses Strategy, March 2014

42
3.7 Impact on transmission losses
Implementation of losses reduction strategies at distribution networks will also have impact on
the transmission network losses. This impact of course is location-specific, depending on the
electrical location of the distribution area in question within the national transmission system.
In this context, we have investigated the impact on the losses reduction in the UKPN LPN
network on the GB transmission network losses. Due to the geographical location of the UKPN
network, i.e. the south of England, and considering that power flows in the transmission
network are from north to south, the reduction of losses in LPN distribution network should
result in lower losses at the transmission level.
In order to analyse the impacts on transmission losses, the following approach is applied:
1. Marginal losses of the GB transmission network are calculated based on peak demand
conditions. The marginal losses are location-specific and indicate the increase in
transmission system losses if the demand at the location in question is increased by 1
MW.
2. Based on the results from step 1, the average marginal losses for the GSPs in the LPN
area are calculated.
3. Considering the Elexon class 8 profile (with 76% load factor), the potential annual
losses reduction in the transmission system is estimated, driven by reduction of losses
in LPN’s GSP. This analysis suggests that on average, 5.5% reduction in transmission
losses can be attributed to the distribution losses reduction in LPN. In other words, 1
MWh reduction in losses in distribution network contributes to 0.055 MWh loss
reduction in transmission network.
As losses are a function of system loading conditions, the impact of reduced loading in the
LPN network on transmission losses is expected to vary according to the system loading as
well. This is illustrated in Figure 3.18.

0.12
Marginal losses (MW/MW)

0.1
0.08
0.06
0.04
0.02
0
0 20 40 60 80 100
Load (%)

Figure 3.18. Marginal transmission losses corresponding to losses of LPN network as a function of system
loading.

For example, during peak load condition (100%), 1 MWh losses reduction in the LPN area
would reduce transmission losses by 0.11 MWh. However, when system loading is 60%, 1
MWh losses reduction in the LPN area will reduce transmission losses by 0.024 MWh. This
highlights the importance of losses reduction during peak demand conditions.

43
3.8 Matching Demand with PV Output
Demand side management strategies can also yield significant benefits in terms of losses
reduction, given that demand flexibility can be used to match DG output and therefore reduce
the loading of the network. In this study, we investigate the benefits achieved on UKPN
networks by optimising the operation of smart domestic load appliances to match PV output.
Elexon Class 1 demand profiles (Table 10), as well as PV output profiles (Figure 3.19) for 15
characteristic days, are used in the analysis.
Table 10. Demand profiles corresponding to 15 characteristic days.

Autumn High Summer Summer Spring Winter

Week Saturday Sunday
WD Sa Su WD Sa Su WD Sa Su WD Sa Su
Day (WD) (Sa) (Su)
Index 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

100%
1 2 3
90%
4 5 6
80%
7 8 9
70%
10 11 12
Power, pu

60%
13 14 15
50%
40%
30%
20%
10%
0%
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of day, hour

Figure 3.19. PV output profiles corresponding to 15 characteristic days.

Figure 3.20 presents the net demand (demand minus PV output) profiles of the 15
characteristic days considered.

900
1 2 3
800
4 5 6
700
7 8 9
600
10 11 12
Power, kW

500
13 14 15
400
300
200
100
0
-100 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of day, hour

Figure 3.20. Net demand profiles corresponding to 15 characteristic days.

The net demand profiles are modulated by shifting demand of smart appliances from peak to
off-peak periods, as shown in Figure 3.21. It can be observed that this demand-side
management strategy leads to a reduction of peak demand from about 860 kW to about 760

44
kW. At the same time, the off-peak load is increased in order to keep the total consumed
energy over the day unchanged, satisfying consumers’ requirements.

800
700 1

600 2
Power, kW

500 3
400 4
300 5
200 6
100 7
0
8
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Time of day, hour 9

Figure 3.21. Net demand profiles corresponding to 15 characteristic days after demand side management
actions.

This change of the net demand profiles results in an overall reduction of network losses of
15% and 40% for a PV penetration of 10% and 20% respectively. This implies that demand-
side management actions will have a higher impact on network losses as the PV penetration
increases.

3.9 Switching off transformers

Switching off primary and grid transformers reduces no-load losses and increases load losses.
Hence, if this action is carried out during low loading conditions, there is potential for overall
losses reduction, if the reduction in no-load losses outweighs the increase in load losses.
The impact of such a strategy is illustrated on a primary substation containing four EHV/11
kV, 15 MVA transformers. For the purpose of this example, considered load losses are 45 kW,
and no-load are losses 30 kW. Furthermore, load profiles for Hyde Park Estate A 11kV (HYPA)
substation for the period of 1/2/2013-31/1/2014 are used. The assumed power factor is 0.95.
Breakeven loading for three and four switched on transformers is set at 40.3 MW loading. For
two and three it is at 28.5 MW, and for two and one is 16.5 MW loading. This is dependent on
the ratio of the load and no-load losses. Hence, if the load is higher than 40.3 MW, transformer
losses are the lowest if all four transformers are switched on. If loading is between 28.5 and
40.3 MW transformer losses are lowest if one of the transformers is switched off, and so on.
Figure 3.22 shows the potential for losses reduction depending on the minimum duration of
switching off a transformer. If transformers could be switched off and back on again in the next
period, potentially 10% of substation losses could be saved. For the one-month minimum
duration of the transformer being switched off, about 6% of substation losses could be saved.
This is equivalent to £4.9-8.1k cost saving.
However, the potential for savings is highly dependent on the ratio of load to no-load losses.
For a typical high ratio (10:1) there is no benefit. Hence, the potential for losses reduction by
switching off primary transformers during low demand conditions is very low.

45
 1640 12%
1620
1600 10%

Losses reduction (%)

Losses (MWh/a)

1580 8%
1560
1540 6%
1520
1500 4%
Losses (MWh/a)
1480 2%
Base losses (MWh/a)
1460
Losses reduction (%)
1440 0%
0 168 336 504 672
Minimum switching period (h)

Figure 3.22. Operational diagram assuming minimum switching off period of transformers

3.10 Summary
A number of key drivers of network losses have been identified and analysed in order to inform
the development of future strategies for reduction of losses. These include changes in network
operational topology, improvement of power factor, changes in load profile, management of
phase imbalance and harmonics.
Key results of the conducted studies are as follows:
 The analysis aimed at optimising network topology through Normally Open Points (NOP)
in the UKPN HV networks, demonstrated that this could result in significant loss reduction
in the order of 15%. This concept hence may provide an important opportunity for reducing
losses, as the economic case for this method is likely to be strong.
 For the three UK Power Networks licence areas, feeders are ranked by the possible
reduction in losses driven by power factor improvement. The potential for loss saving is
assessed assuming power factor improvement from 0.85 to 0.95. This led to reduction in
losses in each feeder between 20 and 25%. For a single point, PF compensation losses
reduction that could be achieved is between 11 and 14%. It is interesting that the analysis
demonstrated that improving power factor in only one third of HV feeders could achieve
90% of potential losses reduction. Hence, the list of 30 highest ranked HV feeders is each
licence area is created and measurements of the actual power factor in future trials are
proposed to be carried out.
 Modelling carried out demonstrated that the impact of imbalance level on losses is
nonlinear. Imbalance of 10% and 30% would potentially lead to increase in losses of 5%
and 45%, respectively, and hence feeders with high phase imbalance should be identified.
List of 30 highest ranked LV networks are identified for further measurements and
consideration for reduction of possible load imbalance.
 Given that current harmonics are kept low by design, the impact on network losses and
transformer load losses is unlikely to be significant. On the other hand, impact of voltage
harmonics on transformer no-load losses is linearly dependant on the total harmonic
distortion (THD), and hence impact on losses may be more significant. Given the standard,
THD is below 5% at distribution transformer, below 4% at primary substation and below
3% at EHV. On the other hand, given future deployment of Eco design transformers with
lower no-load losses, the impact of THD is likely to reduce.

46
4 Application of smart technologies for reduction of
losses

Context: Maximising the Value of Smart Grid for Network Congestion and Losses
The need to accommodate growing penetration of low carbon technologies (renewable
generation, electrification of heat and transport sectors) has challenged the traditional
operation and design paradigm of distribution networks. Therehave been very significant
innovation activities aimed at enhancing utilisation of existing infrastructure in order to reduce
network reinforcement needs. A range of smart-grid technologies and systems (e.g. active
network control, energy storage, demand side response, dynamic line rating etc.) has been
successfully applied in enhancing network asset utilisation, whichgenerally leads to an
increase in network losses. Analysis demonstrates that the increase in network losses driven
by the application of smart-grid technologies is economically justified, as savings in network
reinforcement costs outweigh the additional losses of the smart grid paradigm.
On the other hand, some of these smart grid technologies and systems could also potentially
be applied to reduce losses, in addition to deferring network reinforcement and facilitating
larger penetration of low carbon technologies. In this context, analyses have been carried out
to assess the potential opportunities offered by smart grid technologies to reduce network
losses, with particular focus on existing solutions that have been trialled by UK Power
Networks to facilitate connection of low carbon technologies.
4.1 Case Study No 1 - Flexible Plug and Play (FPP) Project
In this work, we investigate the opportunity of using the Quadrature Booster (QB) to minimise
network losses. The QB was deployed in the Flexible Plug and Play (FPP) project to manage
load flows on congested circuits. The FPP was a Second Tier Low Carbon Network Fund
(LCNF) project aimed at trialling new technologies and novel commercial solutions to achieve
cost-effective integration of distributed generation (DG), such as wind power or solar, into the
electricity distribution network. The QB has been deployed to relieve the network constraint
in the Wissington area. The system is presented in Figure 4.1.
Due to the thermal limit constraint on the 2nd circuit, the output of CHP at Wissington would
need to be constrained (especially during low demand conditions). Instead of reinforcing the
circuit, a QB was installed to enhance utilisation of the spare capacity on the adjacent circuit
(shown in the diagram) which otherwise could not be used without the QB (due to its high
impedance characteristic). By managing the flows on these two circuits, it is demonstrated
that the QB can defer network reinforcements and allow higher utilisation of the CHP plant at
Wissington.
In this context, the QB is controlled such that the flows are divided evenly between the 1st and
2nd circuit. While this approach solves the network congestion issue; this may lead to higher
losses since the impedance of the 1st circuit is high, and therefore, it is not recommended to
always increase the flows at the 1st circuit.

47
 2nd 1st circuit
circuit

Figure 4.1 Wissington test system

In the next section, we will analyse a different control approach for the QB that leads to smaller
losses.
4.1.1 Loss performance of different control approaches for QB
There are two control approaches that have been analysed:
1. Balancing the flows: in this approach, QB is used to balance the flows in the 1st and
2nd circuit at all times. The control algorithm is relatively simple as it only needs to
monitor the flows at both circuits and adjust the setting of the QB to balance the loading
of the two circuits.
2. Minimising the losses: in this approach, the QB is used to manage the thermal loading
but also network losses. In order to optimise the setting, a Security Constrained
Optimal Power Flow (SCOPF) model based on the real-time system operating
conditions is applied. The impact of the flows in the broader system can be analysed
and computed to achieve the optimal results from the overall system perspective.
While this requires a more complex control algorithm, this approach would deliver more
optimal losses performance.

The results of these approaches are shown in Table 11. The results demonstrate that the
losses in the scenario where the QB is controlled only to balance the line loading, the losses
are higher by 11% - 25% compared to the losses if the QB is optimised concurrently to
minimise the losses.

48
 Table 11 Performance of different QB control approaches on losses

Losses per hour (MWh)

Loss
System conditions Balancing line Flow management with
reduction
loading minimisation of losses
High demand –
1.7 1.5 11%
high DG
Low demand –
2 1.5 25%
low DG

This finding suggests that there is a significant opportunity to reduce network losses by
optimising the settings of QB for losses management, in addition to managing constraints and
voltage violations. This analysis suggests that the QB could be used to increase the flows in
the low-losses circuit (2nd circuit) – up to its thermal limit, rather than balancing the load
between both circuits, as the impedance of the 1st circuit is three times the impedance of the
2nd circuit. This is shown in Figure 4.2.

Figure 4.2 QB is used to manage the flow and losses

Optimising the setting of QB for management of losses requires detailed understanding of the
network impedance characteristics and may not be straightforward; hence, advanced network
optimisation tools, such as SCOPF, would be needed to support the system operator in
determining the optimal QB setting while respecting the network constraints (voltages and
thermal limits).
In summary, more value can be extracted from smart grid technologies like the QB if it could
be used to simultaneously manage network congestion and also to reduce losses. In this
study, the losses could be reduced by 11% to 25%, which are attractive opportunities to be
explored. However, it requires significantly more complex control algorithms and the cost
implications of this strategy need to be investigated. For example, costs resulting from more
frequent changes to the QB settings.

49
4.1.2 Distributed Generation (DG) reactive power capability for losses reduction
Traditionally, DG operates at unity power factor. This is achieved by internal compensation of
the DG’s reactive load. This control approach is practical for the passive distribution network
as the distribution system operator/planner would not need to consider the impact of DG’s
reactive load on the system. However, when the system voltages, flows, and losses are being
managed more actively, there may be significant opportunity to apply this “hidden” reactive
capability to manage the voltages and losses in the system.
In this context, a set of studies have been carried out analysing the benefits of controlling DG’s
reactive power capability to reduce network losses. For this purpose the 33 kV EPN
distribution system between Peterborough and March is analysed, see Figure 4.3. This is the
FPP trial zone which serves an area of approx. 30 km diameter (700 km2) and is particularly
well suited for renewable generation. Over recent years UK Power Networks have actively
facilitated significant growth in connection applications and corresponding deployment of
renewable generation in this area. More than 230 MW of DG have been already connected
and it is expected to increase further in future. The system used for the study is shown below.

slack
sl a ck

BHWS DOWN_1 NOWNT NOWN

B11 B14 BOAR
WHIT_2 Wind2
PETC CHAT_2 LITT
PV1
MARCH_TEE BOAR_1 SOPR
uPV4 UPWL_1
NOWN_2 SWAFT_1
FARC_1 BANK_FM uPV3 uPV5
FARC_2 FUNT_T2 FUNT_T1 AF_1
MARCH_FM
WHIT_T1
WHIT_T2

S1 PV3 S4
MARCH132 S3
AF SWAF
WISS
W6

FARC FUNT RANWT 2 Tx

MARCH
WHIT C1
uPV6 CHAT_T1
CHAT_T2 STAG
W4

S2
W5

GLAS_1 GLAS GLAS2 MARCH_T1

PV2 MARCH_T2
uPV7 W7
W3
slack

MARCH_P
slack

BURY WOODT CHAT

W1

RAMSEY REDT_2
REDT CHP1 Wind4 uPV2
REDTT_1
Wind5

Wind6
Wind3
uPV1

PV4

Wind1 W2

Figure 4.3 Flexible Plug and Play network

For this analysis, we evaluate the impact of DG’s reactive power on the system losses across
a set of operating conditions considering the variation in demand and the output of variable
generators (wind, PV) across one-year period. The study varies the DG’s reactive power
capability between 0.75 (lag/lead) and unity power factor. DG’s reactive power output is
optimised using the Optimal Power Flow (OPF) algorithm to minimise the total operation cost
of the system including the cost of curtailing DG and losses. The impact of having different
reactive power capability on losses are shown in Figure 4.4.

50
 Figure 4.4 Benefits of controlling DG’s reactive power capability on losses

The results demonstrate that by controlling DG’s reactive power capability via active network
management, the volume of DG curtailment can be reduced from 2.4% to 0.2% and the losses
also decrease from 2.9% to 2.6% (a reduction of about 10%). This is expected since the
reactive power from DG can be used to control system voltages and relieve the voltage-driven
network constraints, which may cause DG curtailment. Modelling demonstrates that in this
particular case there is a trade-off between cost of losses and cost of DG curtailment.
This analysis also suggests that the value of the first available reactive capacity is higher than
subsequent capacity. This is demonstrated in Figure 4.4 where the losses are reduced sharply
when DG is able to operate beyond the unity power factor. However, the reduction in losses
starts to saturate at 0.95 pf. Increasing the reactive capability further does not reduce the
losses further.
These results provide some new insight related to the efficient level of DG reactive capability
needed by the system, and this can inform the design or sizing of the DG’s reactive power
support. It is important to note that the impact on losses in the study is limited within the
perimeter of the test system. Furthermore, there may be additional losses reduction obtained
in the upper voltage network but this is expected to be marginal.
As the impact of reactive power tends to be local, the value of reactive power services tends
to be very locational and system specific as well. In this context, to illustrate the point, we have
analysed the utilisation of DG reactive power across all operating conditions. The results are
shown in Figure 4.5.

51
 Figure 4.5 Utilisation of reactive sources

The results show the limit for injecting (QMAX) and absorbing (QMIN) reactive power, the
maximum reactive power injection (Max Qinj) and absorption (Max Qabs), and the average
reactive power usage (Average Q usage). It is shown in Figure 4.5 that not all of the capacity
available (injection or absorption) is used by the system. For example: DG with ID=10 should
only provide capability to inject reactive power, while for DG with ID=9, should be able to
operate with leading or lagging power factor. This type of analysis is important for identifying
the requirement for local reactive power services; in the future, this service may have
commercial value which would incentivise the provision of such services to reduce network
losses.
In conclusion, the ability to utilise DG’s reactive power capability by controlling its reactive
output according to the system needs can contribute to a reduction in network losses in
addition to relieving voltage-driven network constraints that may trigger DG curtailment and
increase the need for network reinforcement. While the potential losses reduction is system
specific, this study shows around 10% improvement in losses could be achieved. The study
also shows that enabling DG to operate with 0.95 pf may be desirable; the benefit of increasing
further the reactive capability in terms of losses reduction is small. This analysis also
demonstrates that the value of reactive power services is location specific.

52
4.2 Case study No 2: Soft open points for loss minimisation
4.2.1 Introduction
The overarching aim of the Flexible Urban Networks at Low Voltage (FUN-LV) project was to
explore the use of power electronics devices (PEDs) to enable deferral of reinforcement and
facilitate the connection of low carbon technologies and distributed generation in urban areas.
This was to be achieved by meshing existing networks which are not currently meshed, and
by removing boundaries within existing meshed networks. In this section, we evaluate the
potential for soft open points (SOPs) in reducing losses through balancing loading among
distribution substations..
A non-linear optimisation model was developed to simulate system operation. The model’s
objective function was the minimization of losses while a DC power flow formulation was
adopted throughout this analysis.
Using three case studies, based on data provided by UK Power Networks, this analysis
demonstrates that the potential for loss minimisation can be substantial, yet it depends highly
on network loading and the efficiency of the SOP devices. In particular, whereas fully efficient
SOPs are shown to lead to energy loss reductions in the range of 10%, the benefit of less
efficient SOPs may be severely limited.
4.2.2 Case Study 1: Eastbourne Terrace
We evaluate the potential for a 240kW dual-terminal soft open point to minimize transformer
losses in the area of Eastbourne Terrace; The SOP is assumed to be fully efficient. The
network diagram and corresponding demand profiles are shown in Figure 4.6 below. We have
isolated two substations for analysis; West Terrace and East Terrace. West Terrace has a
750kW transformer and a peak demand of 440 kW. East Terrace has a 500kW transformer
and a peak demand of 290 kW. In terms of transformer load losses, these were estimated at
10kW at full loading for West Terrace and 6.68kW at full loading for East Terrace.

Figure 4.6: Eastbourne Terrace network (left panel) and typical demand profiles (right panel).

Network operation with and without SOP for one typical day is shown in Figure 4.7 below. As
can be seen on the left, in the base case without SOP, East Terrace is more loaded than West
Terrace. This provides potential for re-balancing between the two transformers. Average
losses over a typical day were found to be 83.37kWh. With the installation of an SOP the
losses were reduced to 71.08kWh i.e. a 14.7% reduction. As shown in the right panel of Figure
4.7, this is achieved by increasing the loading of the West Terrace and then transferring this
excess energy back to East Terrace via the SOP.

53
 Figure 4.7: Network operation for a typical day without SOP (left panel) and with SOP (right panel).

4.2.3 Case Study 2: Boyce’s Street

We evaluate the potential for a 240kW dual-terminal soft open point to minimize transformer
losses in the area of Boyce’s Street. The network diagram and corresponding demand profiles
are shown in Figure 4.8 below. Two substations are considered in this analysis; West Terrace
and East Terrace. Churchill Square has a 1000kW transformer and a peak demand of 800kW.
Duke Street has a 1000kW transformer and a peak demand of 350 kW. In terms of transformer
losses, these were estimated at 11.8kW for full loading for both Churchill Square and Duke
Street transformers.

Figure 4.8: Boyce’s Street network (left panel) and typical demand (right panel).

Network operation with and without SOP for one typical day is shown in Figure 4.9 below. As
can be seen on the left, in the base case without SOP, Duke Street is more loaded than West
Terrace providing potential for demand re-balancing between the two substations. Average
losses over a typical day were found to be 80.17kWh. With the installation of an SOP the
losses were reduced to 70.89kWh. The SOP was able to achieve 11.5% reduction by
increasing the loading of Churchill Square, particularly during midday and afternoon hours.
Note that as shown in the second plot of the right panel, the SOP reaches its maximum rating
of 240kW at around 1pm and thus this particular network would benefit from a larger SOP.

54
 Figure 4.9: Network operation for a typical day without SOP (left panel) and with SOP (right panel).

4.2.4 Case Study 3: Prudential North

We carry out a similar analysis on the Prudential North network to evaluate the potential for a
400kW multi-terminal soft open pint to minimize transformer losses. The network diagram and
corresponding demand profiles are shown in Figure 4.10 below. We have focused on three
substations for analysis; Vokins, Prudential North and New Road. Prudential North and New
Road have 1000kW transformers and a peak demand of 390 and 600 kW respectively. Vokins
has an 800kW transformer and a peak demand of 160 kW. In terms of transformer losses,
these were estimated at 11.8kW and 10.0kW at full loading for the large and smaller
transformers respectively.

Figure 4.10: Prudential North network (left panel) and typical demand profiles (right panel).

Network operation with and without SOP for one typical day is shown in Figure 4.11 below. As
can be seen on the left, in the base case without SOP, New Road is more loaded than Vokins
and Prudential North, providing potential for re-balancing between the three substations.
Average losses over a typical day were found to be 74.97kWh. With the installation of a three-
port SOP the losses were reduced to 65.50kWh i.e. a 12.6% reduction. As shown in the right
panel of Figure 4.11, this is achieved by transferring energy via the SOP from Vokins to
Prudential North and from Prudential North to New Road, resulting in increasing the loading
of Vokins and decreasing losses in the other two substations.

55
 Figure 4.11: Network operation for a typical day without SOP (left panel) and with SOP (right panel).

4.2.5 Impact of SOP efficiency

The preceding analysis has assumed 100% efficiency for the SOPs. When performing the
three case studies under the assumption of a 90% efficient SOP, the benefits were found to
be considerably reduced. In particular, in the case of Eastbourne Terrace and Boyce’s Street,
the SOP could not assist in loss reduction while in the case of Prudential North, average daily
losses were reduced by 3.3%. Results of the three case studies are summarised in Table 12.
Table 12: Summary of results – percentage off loss reduction per case study.

Case Study 100% efficiency 90% efficiency

Eastbourne Terrace 14.7% 0.0%

Boyce’s Street 11.5% 0.0%

Prudential North 12.6% 3.3%

4.3 Case study No 3: Role and value of energy storage systems in

minimizing distribution network losses
4.3.1 Introduction
Energy storage systems can offer a large number of valuable services to current and future
low-carbon power systems, including balancing of demand and supply across multiple
timescales, and the reduction of peak demand. Energy storage can also enhance the cost
efficiency of distribution networks by mitigating thermal and voltage constraints and therefore
avoiding the capital intensive reinforcement of distribution assets. This is primarily achieved
by reducing peak demand levels and shifting demand to adjacent off-peak periods. However,
this also creates a value stream of storage which has not been fully appreciated and involves
the reduction of network losses and the associated societal costs. This can be achieved by
suitable coordination of storage active power; but also through the coordination of their

56
reactive power contributions, given the advanced reactive control capabilities of modern
inverters.
This section analyses the impacts of different storage operating policies on network losses. In
order to achieve that, the interdependencies between the provision of local distribution network
services and the participation in nation-wide energy and balancing markets are explored.
Finally, the impacts of different operational parameters of storage (including round-trip
efficiency, power and energy capacity and reactive power control capabilities) on the overall
performance of these services are investigated.

4.3.2 Modelling considerations

4.3.2.1 Input Data
The analysis discussed above is carried out on a model of the Leighton Buzzard primary
substation, illustrated in Figure 4.12. Detailed data regarding the network parameters in this
test network were considered in the executed power flow analysis.

Figure 4.12. Topology of distribution network at Leighton Buzzard substation

The two feeders connected at the Leighton Buzzard substation (indicated in red and green
colour in Figure 4.12) are comprised by a number of sections with underground cables and
overhead lines. The different sections along with their detailed parameters are presented in
Figure 4.13.

57
 Section Length Resistance Reactance Capacity
ID (km) (Ohm/km) (Ohm/km) (MVA)
1 0.15 0.165 0.101 21.4
2 0.13 0.093 0.101 28.0
3 0.13 0.140 0.109 21.7
4 0.13 0.165 0.101 21.4
5 9.84 0.137 0.326 30.5
6 9.82 0.137 0.326 30.5
7 1.63 0.165 0.101 21.4
8 1.64 0.165 0.101 21.4
9 1.02 0.165 0.101 21.4
10 1.06 0.165 0.101 21.4
11 0.64 0.148 0.347 30.5
12 0.64 0.148 0.347 30.5
13 0.64 0.148 0.347 30.5
14 0.64 0.148 0.347 30.5

Figure 4.13. Simplified network diagram of Leighton Buzzard primary substation and respective feeder section
parameters

The two transformers corresponding to sections 15 and 16 in Figure 4.13 exhibit significant
no-load (iron) losses (11.8kW each) generated by magnetizing current at their core and
depending on the magnetic properties of the materials in the transformer’s core.
Hourly data corresponding to the total demand at the Leighton Buzzard substation and the
energy market prices for a typical summer day and a typical winter day are presented in Figure
4.14. The power factor of the demand is assumed equal to 0.9, while the energy market prices
have been determined based on 2015 average conditions. The model assumes that the cost
of losses for the distribution network operator is determined according to the same energy
prices.

Figure 4.14. Typical profiles of (a) local demand at Leighton Buzzard substation and (b) energy market price

The storage considered in this analysis is assumed to participate both in the energy market to
seize arbitrage opportunities (buying energy during periods of low prices and selling energy
during periods of high prices) but also in the balancing market through the provision of
frequency response services. In this context, typical availability prices for the provision of Firm
Frequency Response (FFR) in the UK have been used in the analysis.
The storage operational parameters considered are as follows:

 Active power capacity: 6 MW

 Apparent power capacity: 7.5 MVA
 Energy capacity: 10 MWh
 Round-trip efficiency: 90%

58
4.3.2.2 Key modelling aspects
The employed modelling approach goes beyond considering provision of individual services
by energy storage and investigates a multi-service business framework. In other words, it
assumes that storage can simultaneously participate in multiple markets and provide multiple
valuable services. In this framework, storage needs to optimally allocate its power and energy
capabilities to various services while accounting for the interdependencies and conflicts
between these services. More specifically, the model accounts for:
• Participation in energy market: storage participates in the day-ahead energy market to
seize arbitrage opportunities i.e. buy energy during periods of low prices and sell
energy during period of high prices.
• Provision of frequency response services: storage participates in the balancing market
through the provision of firm frequency response. This requires rapid response times
and would be particularly well suited for Li-ion batteries.
• Provision of network services: storage is capable of mitigating distribution network
constraints during peak demand periods as well as reducing network losses.
In the context of analysing the impact of energy storage operation on network losses, various
operating policies have been established focusing on the potential synergic and conflicting
actions with other services. Specifically, the following operating policies have been
considered:
• Base case: This case serves as a benchmark and assumes that energy storage is not
available in the network. Therefore, the model determines the network losses
associated with meeting the local demand without storage.
• Min Losses: This operating policy prioritises the reduction of network losses and thus
optimises the operation of storage to minimise network losses. Therefore, storage
effectively disregards the market signals from energy and balancing markets when
optimising its actions.
• Max Profit: This operating policy prioritises the participation of storage in energy and
balancing markets and thus optimises the operation of storage to maximise its revenue
from these markets. Therefore, storage effectively disregards the impact of its actions
on network losses.
• Optimized: This operating policy reconciles the conflicts between the previous two
policies by maximising the difference between the revenue of storage in energy and
balancing markets minus the energy cost of network losses.

4.3.3 Results
Figure 4.15 presents the net demand profiles (combination of local demand and energy
storage actions) as well as the energy storage output corresponding to the Base Case and
Min Losses operating policies in a typical winter day. Application of the Min Losses operating
policy leads to a reduction of peak demand (between hours 17 and 22) and shift of demand
towards off-peak periods (between hours 1 and 7).

59
 Figure 4.15. a) Net demand and (b) storage output corresponding to the Base Case and Min Losses operating
policies

Figure 4.16 illustrates the reduction in network losses achieved with the Min Losses policy in
comparison to the Base Case.

15%
Reduction
(620 kWh)

Figure 4.16. Network losses corresponding to the Base Case and Min Losses operating policies

Furthermore, by analysing the load duration curve over 1 year of operation for both Base Case
and Min Losses operating policies (Figure 4.17), it becomes clear that the latter policy reduces
local peak demand. The above results demonstrate that reducing network losses is synergic
with the reduction of peak demand levels and the avoidance of network reinforcements.

Figure 4.17. Load duration curve corresponding to Base Case and Min Losses operating policies

Going further, Figure 4.18 presents the net demand profiles and the network losses
corresponding to the Base Case, Min Losses and Max Profit operating policies. Although the
Max Profit policy disregards the impact of storage actions on network losses, it leads to lower
losses with respect to the Base Case. This is driven by the fact that energy market prices and
local demand are often correlated, i.e. they exhibit coincident peaks and valleys. However, the
reduction of losses is not as significant as in the case of the Min Losses policy.

60
Figure 4.18. (a) Net demand and (b) network losses corresponding to the Base Case, Min Losses and Max Profit
operating policies

Figure 4.19 illustrates the energy storage power and energy levels corresponding to these
three operating policies. Application of the Max Profit policy results in additional charging /
discharging cycles with respect to the Min Losses policy, since storage operation is driven by
the energy market price differentials.

Figure 4.19. Storage (a) output and (b) energy level corresponding to the Base Case, Min Losses and Max Profit
operating policies

Charging and discharging actions by energy storage are subject to energy losses given its
round-trip efficiency. Therefore, a comprehensive analysis needs to account for these losses
apart from network losses. Figure 4.20 presents both network and storage losses
corresponding to different operating policies. Although energy storage reduces network losses
(with respect to the Base Case) under all the examined operating policies, the total losses are
higher than the Base Case due to the significant round-trip losses of storage (its round-trip
efficiency is assumed 90% in this case), even if a Min Losses policy is adopted. Furthermore,
it should be noted that Max Profit and Optimized operating policies do not only lead to higher
network losses with respect to the Min Losses policy, but they also yield higher storage losses
due to the additional charging / discharging cycles they entail (Figure 4.19).

61
 Figure 4.20. Network and storage losses corresponding to different operating policies

Beyond its round-trip efficiency, the performance of storage depends on other factors such as
its power and energy capacity. Figure 4.21 presents the reduction in network losses achieved
by the adoption of the Min Losses operating policy with respect to the Base Case, under
different scenarios concerning the round-trip efficiency, the power capacity and the energy
capacity of storage.
As the round-trip efficiency increases, the performance of storage in reducing network losses
is enhanced, as storage can efficiently shift more demand from peak to off-peak periods
without increasing excessively its own energy losses. The same effect emerges when the
energy capacity of storage is increased, since storage has the capability to shift more demand
from peak to off-peak periods. In contrast, an increase in active power capacity has no effect
on the performance of storage.

Figure 4.21. Reduction in network losses achieved by the adoption of the Min Losses operating policy with
respect to the Base Case, under different scenarios concerning the round-trip efficiency, the power capacity and
the energy capacity of storage

In addition to its capability to reduce network losses by means of active power charge /
discharge actions, energy storage can further reduce network losses through reactive power
actions, given the advanced reactive control capabilities of modern inverters. Although the
level of reactive power demand is significantly lower than the active power demand, this
capability has a major impact on network losses. Figure 4.22 presents the active and reactive
net demand profiles corresponding to the Base Case, the Optimized operating policy without
considering the reactive control capability and the Optimized policy when considering the
reactive control capability.

62
 Figure 4.22. Net demand profiles of (a) active power and (b) reactive power corresponding to Base Case,
Optimized operating policy without reactive power capability and Optimized operating policy when considering
reactive power capability

Without reactive power control capabilities by the storage, the reactive power profile is given
by the local demand and associated power factor. When storage exhibits such capabilities, it
can modulate its reactive power output so as to reduce the reactive power flows and thus
reduce overall network losses. It should be noted that in contrast to active power
(charge/discharge) outputs which are limited by power and energy capacities (i.e. storage is
required to recover its discharged energy) reactive power output is only limited by the storage
rating. As a result, the overall network losses are significantly reduced when reactive control
capability is available (Figure 4.23).

Figure 4.23. Network losses corresponding to Base Case, Optimized operating policy without reactive power
capability and Optimized operating policy when considering reactive power capability

4.4 Summary
 Modelling demonstrated that the use of Quadrature Booster, beyond network
constraint management as tested in the Flexible Plug and Play project, could
deliver savings in local network losses from about 11% (in the case of high demand
and high distributed generation (DG) growth) up to 25% for low demand and low
DG growth.
 Smarter Network Storage installed in Leighton Buzzard could potentially reduce
losses in supplying circuits by about 15%.
 Modelling demonstrated that the Soft open point (SOPs), installed for the
management of constraints in LV feeders, could potentially reduce losses in the
corresponding LV network and distribution transformers by about 10%-15%.

63
 Installing smart switchgear in HV networks could potentially reduce losses further
by up to 10%, in addition to optimised NOP positions as described in Section 3.1.
 As indicated by Ofgem, the rollout of smart meters would lead to a reduction of
energy demand of 2.8%. Modelling carried out demonstrated that this would
potentially reduce distribution network losses for about 5.5%.
 Comprehensive analysis is carried out regarding the impact of demand
redistribution on network losses, showing that demand side response, which could
potentially shift 2.5% load from peak to off-peak period, would lead to reduction of
losses by about 3%.

64
5 Identification of efficient loss reduction investment
strategies

5.1 Eco-design: low-loss transformers

The contribution of transformer losses to the overall system losses is relatively significant
(between 15% and 20%). Recent development in transformer technologies has yielded
considerable reduction of transformer losses, which is important given the 30-40 years lifetime
of the assets. Figure 5.1 shows load and no-load losses performance of 500 kVA (left) and
1000 kVA(right) transformers between pre-1955 and 2021 (future transformers).

9 16
Load losses No load losses Load losses No load losses
8
Tranformer losses (kW)

Tranformer losses (kW)

14
7 12
6 10
5
8
4
3 6
2 4
1 2
0 0

Figure 5.1. Breakdown of transformer losses for different designs for 500 kVA (left) and 1000 kVA (right)

Both load and no-load power losses can be reduced substantially. For example, the pre-1955
transformer no-load losses are about 2.2 kW. This is expected to reduce to 0.2 kW by using
the Eco-design low-loss transformers. Similarly, the load losses can be reduced by 33% from
12 kW to 8 kW. The use of copper instead of aluminium contributes to the significant reduction
in losses with the associated drawback of heavier transformers.
In order to estimate the potential losses reduction if the use of Eco-design 2015 low-loss
transformers is rolled-out widely across the UK Power Networks system, we use the 2016 data
on the losses reduction attributed to the use of low-loss transformers for replacing the old
distribution transformers in UK Power Networks (Table 13).
Table 13. Replacement of distribution transformers in 2016 (number)

Total reduction of
Asset General
Area Mounting annual losses
Replacement Reinforcement
(MWh/year)
PMT 0 12 12
EPN
GMT 122 16 138
PMT 0 0 0
LPN
GMT 59 3 62
PMT 0 3 3
SPN
GMT 67 10 77
Total 248 44 292

65
By analysing the UK Power Networks data, we have identified the possible number of
transformer replacement for PMT and GMT in EPN, LPN, and SPN areas. Based on these
figures, we estimate the total benefit of losses reduction. The results are presented in Table
14.
Table 14. Potential losses reduction by rolling out widely the use of Eco design low-loss transformers in UKPN

Asset General
Area Mounting Total
Replacement Reinforcement
PMT 0 51 51
EPN
GMT 1,390 182 1,572
PMT 0 0 0
LPN
GMT 743 38 781
PMT 0 10 10
SPN
GMT 705 105 811
Total 2,838 387 3,225

The total losses reduction is significant, i.e. 3,225 MWh per year. The capitalised value of this
losses reduction is between £1.6-2.6m.

5.2 Amorphous Steel Transformers

Another different type of energy efficient transformer is an amorphous metal transformer
(AMT). The magnetic core of this transformer is made with a ferromagnetic amorphous metal;
the materials have high magnetic susceptibility, very low coercivity and high electrical
resistance which lead to low eddy current losses. The no-load losses of AMT can be 70 – 80%
lower than with traditional crystalline materials. This is demonstrated in Table 15 below where
the no-load (NLL) and load losses (LL) performance of AMT for different types are compared
against cold rolled grain oriented steel (CRGO) transformers. As shown in the table, while the
load losses performance of both types of transformers is the same, the NLL of AMT are 63-
79% lower than NLL of CRGO transformers.
Table 15. Comparison between losses performance of CRGO and Amorphous transformers

CRGO
Amorphous Transformer
Rating, Phases transformer
Voltage, kV
kVA count NLL
NLL LL NLL LL
reduction
25 1 11/0.25 70 900 15 900 -79%
50 1 11/0.25 90 1100 22 1100 -76%
50 1 11/0.25-0-0.25 90 1100 22 1100 -76%
100 1 11/0.25-0-0.25 145 1750 38 1750 -74%
100 3 11/0.43 145 1750 53 1750 -63%
200 3 11/0.43 300 2750 90 2750 -70%

Based on the data above, we revise the figures in Table 14 by assuming the replacement of
PMT using the AMT technology while other transformers are replaced with Eco-design low-
loss transformers. The results are presented in Table 16. The losses reduction is slightly
improved, from 3,226 MWh/year to 3,251 MWh/year due to the savings from AMT (26
MWh/year). This study did not take new PMT installations into consideration. The additional
capitalised value of losses reduction is between £13-21k.

66
Table 16. Potential losses reduction by rolling out widely the use of Eco design and AMT low-loss transformers in
UKPN

Asset General
Area Mounting Total
Replacement Reinforcement
PMT 0 72 72
EPN
GMT 1,390 182 1,572
PMT 0 0 0
LPN
GMT 743 38 781
PMT 0 15 15
SPN
GMT 705 105 811
Total 2,838 413 3,251

The study concludes that the deployment of low-loss transformers is essential to reduce
losses; the benefit from the improvement of transformer technologies which is leading to the
more efficient operation and reduction in losses should be capitalised. While the cost of low-
losses transformers is relatively higher than the cost of traditional transformers, with mass
deployment, the cost could be reduced. This will then make the proposition to use low-loss
transformers more attractive.

5.3 Conductor Sizes Rationalisation

Another study investigated and quantified the level of loss reduction driven by increase in the
minimum size of conductors. This strategy could be considered as a simplified variant of loss-
inclusive design which optimises the size of conductors taking into account the long-term cost
of losses altogether with the cost of conductors. In this study, load-flow analysis in carried out
on the representative LV networks with different minimum conductor size policies: (i) base
case (without minimum), (ii) 95 mm2 as the minimum size for main conductors, then (iii) 185
mm2 and (iv) 300 mm2. The losses in each case are shown in Figure 5.2.

300

250
Losses, GWh/year

200

150

100

50

0
Base LV Min 95 sqmm LV Min 185 sqmm LV Min 300 sqmm

Figure 5.2. Losses on the representative LPN LV networks with different minimum conductor size policies

Increasing the minimum size of conductors will reduce losses; for example, if the minimum
size is 95 mm2, losses will decrease by 30%. Implementing higher minimum conductor size,
e.g. 185 mm2 and 300 mm2 will yield higher losses reduction, i.e. 52% and 68% respectively
(compared to the losses in the base case).

67
We simulated this policy on the LPN LV networks by creating the relevant representative
networks for selected LV distribution sites (as shown on x-axis of Figure 5.3). The losses for
different cases are presented in Figure 5.3.

140

120

100
Losses, MWh/year

80

60

40

20

Base 95sqmm 185sqmm 300sqmm

Figure 5.3. Impact of increasing the minimum size of conductors on the LPN LV networks

Depending on the number of feeders affected by this strategy, the level of losses reduction
varies. If more feeders are affected, the losses reduction is higher. For the affected feeders,
the reduction in losses due to increasing the conductor size also varies depending on the
loading of the feeder in question.
In a similar fashion, we investigate and analyse the implementation of this strategy on the LPN
HV network by carrying out simulations for the different minimum size of conductors: (i)
95 mm2, 185 mm2, and 300 mm2. The total network length considered in this study is
10,335 km. The length of circuits affected by this policy is given in Table 17.

Table 17. The length of circuits affected by the minimum conductor size policy

Min95 Min185 Min300

Replaced length, km 190 4,308 8,574

Increasing the minimum size of conductors affects more circuits, e.g. for 95 mm 2, the length
of circuits that needs to be upgraded to meet the policy is 190 km. If 300 mm2 is used as the
minimum limit, then the length of the affected circuits is 8,574 km.
The results of the study on the LPN HV network are presented in Figure 5.4.The x-axis shows
the distribution sites selected for this study. The annual losses are given on the y-axis.

68
 8,000
7,000
Losses, MWh/year

6,000
5,000
4,000
3,000
2,000
1,000
-

Base 95sqmm 185sqmm 300sqmm

Figure 5.4. The impact of implementing minimum conductor size on LPN HV network losses

In LPN, the impact of the implementation of this policy varies across distribution sites. For
example, for the network associated with Durnsfold Road, there is no HV feeder below 95
mm2, so constraining the minimum size of the conductor to 95 mm2 does not have any effect.
However, when the minimum size is higher (e.g. 185 and 300 mm2), the losses are lower by
8.5% and 31% respectively. The study also provides insight into the area of LPN where this
policy would have the largest impact. For example: implementing this policy in the Durnsford
Road area has larger impact than implementing this policy in the Wandsworth Central area.
The distribution sites (on the x-axis) are ranked based on the level of losses (from high to low)
in the base case. This will provide insight on which areas this policy should be implemented
first.
The capitalised value of potential losses reduction in LPN LV and HV networks if the minimum
conductor size is 185 mm2 are between £63-104m and £1.1-1.8m, respectively.
In a similar fashion, the study was carried out in the LV and HV networks in the EPN and SPN
areas. The reduction in losses in EPN LV network is shown in Figure 5.5. The findings are
similar to those in LPN. For example, if the minimum size is 95 mm2, losses will decrease by
35%. Implementing a higher minimum conductor size, e.g. 185 mm2 and 300 mm2 will yield
higher losses reduction, i.e. 60% and 75% respectively (compared to the losses in the base
case). The level of losses reduction in EPN LV networks is slightly higher than in the LPN
networks, as the latter has a smaller number of circuits affected by the policy.

69
 450
400
350

Losses, GWh/year
300
250
200
150
100
50
0
Base LV Min 95 sqmm LV Min 185 sqmm LV Min 300 sqmm

Figure 5.5. Losses on the representative EPN LV networks with different minimum conductor size policies

Figure 5.6 shows the losses reduction for implementing the policy on the EPN LV networks.
The results indicate that many feeders in the EPN area will be affected by this policy, and this
can substantially reduce the losses in EPN.

120

100
Losses, MWh/year

80

60

40

20

Base LV Min 95 sqmm LV Min 185 sqmm LV Min 300 sqmm

Figure 5.6. Impact of increasing the minimum size of conductors on the EPN LV networks

In a similar fashion, we investigate and analyse the implementation of this strategy on the EPN
HV network by carrying out simulations for the different minimum size of conductors: (i)
95 mm2 and 185 mm2. The total network length considered in this study is 33,559 km. The
length of circuits affected by this policy is given in Table 18.

Table 18. The length of circuits affected by the minimum conductor size policy in EPN

Min95 Min185

Replaced length, km 18,344 26,968

The results of the study on the EPN HV network is presented in Figure 5.7.

70
 2,500
Losses, MWh/year

2,000

1,500

1,000

500

Base Min 95 sqmm Min 185 sqmm

Figure 5.7. The impact of implementing minimum conductor size on EPN HV network losses

Similar to the findings for LPN, the impact of the implementation of this policy in EPN varies
across distribution sites. The primary substation IDs (on the x-axis) are sorted based on the
level of losses (from high to low) in the base case. In this case, the primary substation at
Northwold primary is the first potential candidate for implementing this strategy.
The capitalised value of potential losses reduction in EPN LV and HV networks if minimum
conductor size is 185 mm2 are between £114-188m and £24-39m, respectively.
We also performed the study for the LV and HV networks in the SPN area. The losses
reduction in SPN LV network is shown in Figure 5.8. The findings are the similar to LPN and
EPN. For example, if the minimum size is 95 mm2, losses will decrease by 20%. Implementing
higher minimum conductor size, e.g. 185 mm2 and 300 mm2 will yield higher losses reduction,
i.e. 47% and 64% respectively (compared to the losses in the base case). The level of losses
reduction in SPN LV networks is comparable to LPN networks and slightly lower than EPN
networks.

450
400
350
Losses, GWh/year

300
250
200
150
100
50
0
Base LV Min 95 sqmm LV Min 185 sqmm LV Min 300 sqmm

Figure 5.8. Losses on the representative SPN LV networks with different minimum conductor size policies

71
Figure 5.9 shows the losses reduction for implementing the policy on the SPN LV networks.
The results indicate that many feeders in SPN area will be affected by this policy and this can
substantially reduce the losses in SPN.

160

140

120
Losses, MWh/year

100

80

60

40

20

Base LV Min 95 sqmm LV Min 185 sqmm LV Min 300 sqmm

Figure 5.9. Impact of increasing the minimum size of conductors on the EPN LV networks

In a similar fashion, we investigate and analyse the implementation of this strategy on the SPN
HV network by carrying out simulations for different minimum conductor sizes: (i) 95 mm2 and
185 mm2. The total network length considered in this study is 17,701 km. The potential for
losses reduction on the HV network if minimum cable/conductor is Al 95 mm 2 or 185 mm2 is
about 15% or 32% respectively as shown in Figure 5.10. The length of HV circuits in SPN
affected by this policy is given in Table 19.

120000
Annual losses, MWh/year

100000

80000

60000

40000

20000

0
Base Min 95 sqmm Min 185 sqmm

Figure 5.10. Losses on the representative SPN HV networks with different minimum conductor size policies

72
 Table 19. The length of circuits affected by the minimum conductor size policy in SPN

Min95 Min185

Replaced length, km 8,244 13,961

Similar to the findings for LPN and EPN, the impact of the implementation of this policy in SPN
also varies across distribution sites, as presented in Figure 5.11. The primary substation IDs
(on the x-axis) are sorted based on the level of losses (from high to low) in the base case. In
this case, the primary substation at Shepway primary has the largest losses and becomes the
first potential candidate for implementing this strategy in SPN.

4000

3500

3000
Losses, MWh/year

2500

2000

1500

1000

500

Base Min 95 sqmm Min 185 sqmm

Figure 5.11. The impact of implementing minimum conductor size on SPN HV network losses

The capitalised value of potential losses reduction in EPN LV and HV networks if minimum
conductor size is 185 mm2 are between £87-144m and £15-25m, respectively.

5.4 Voltage rationalisation

5.4.1 HV networks
UK Power Networks inherits the legacy of operating 2.2, 3.3, and 6.6 kV systems in some of
its areas; these systems were designed decades ago and may no longer fit for future. As the
systems are aging and may reach its lifetime, there is an opportunity to upgrade these systems
to 11 kV or 20 kV to reduce losses. In order to analyse the level of losses reduction that can
be achieved, we carried out a set of studies quantifying the reduction in losses attributed to
the replacement of those systems with 11 or 20 kV systems. We assume that the network
topology will not change and the replacement is done on a like-for-like basis. The study was
performed on the SPN networks. The results are presented in Table 20.

73
 Table 20. The benefit (losses reduction) of moving to a higher voltage level

Potential for losses reduction Potential for losses reduction

Voltage Length, km
at 11 kV, MWh/year at 20 kV, MWh/year
2.2 99 183 188
3.3 74 1,597 1,707
6.6 2,180 13,402 18,660
11 15,347 0 100,260
Total 17,701 15,181 120,815

Table 20 shows the length of circuits operate at 2.2, 3.3, 6.6, and 11 kV in SPN. The table
also shows the reduction in losses if all the corresponding circuits are operated at 11 kV or 20
kV. If the voltage level can be standardised to 11 kV, the losses could be reduced by about
15 GWh/year (worth about £0.73-1.2 million per year). This will involve upgrading about 2,300
km of network. If, rather than 11 kV, the HV voltage level is standardised to 20 kV, then the
losses could decrease by around 121 GWh/year (worth of £59-97 million). This would involve
an upgrade of about 17,700 km of network. In this study, the changes in transformer losses
are not analysed.

5.4.2 EHV networks

Another study we carried out investigated the losses impact of replacing the 33 kV system
with a 132 kV system. In this study, we simulate the situation where all 33 kV network in the
LPN region (about 6,100 km circuits) is upgraded to 132 kV. The study indicates that the
expected reduction of losses reaches 24 GWh/year, which is worth between £12-19 million
(capitalised value). In this study, the potential changes in transformer losses are excluded
from the analysis.

5.5 Smart distribution transformer

As previously discussed in section 6.2-6.5, improvement of power factor, optimisation of
operating voltage, and balancing the load across AC phases could be an effective strategy to
reduce losses. Smart distribution transformers can provide all the above functionality to
improve the overall efficiency of distribution networks. In this context, we analyse the losses
reduction potential of such technology. We assume that the smart distribution transformers
can reduce the imbalance by 10%, improve the power factor, and also support voltage
optimisation. This results in more than 13% losses reduction: 5% reduction from reducing the
load imbalance and to the remainder from improved power factor and voltage optimisation.
The study was carried out on selected LV systems in the EPN region. The level of losses (left
y-axis of Figure 5.12) and the associated reduction (y-axis on the right) attributed to different
functionalities are shown in the figure below for each selected distribution site (shown on the
x-axis).

74
 140 12

120 10

Losses reduction, MWh/year

100
Losses, MWh/year

8
80
6
60
4
40

20 2

0 0

Distribution site reference

Base Voltage control Power factor correction Imbalance

Figure 5.12. Benefit of deploying smart distribution transformers in EPN region.

In total, the reduction in losses due to the use of smart distribution transformers is around
58,328 MWh/year, which correspond to the range of capitalised savings between £28-47m.

5.6 Scott Connected Transformers

A Scott-T (also called a Scott connected) transformer is a type of circuit used to obtain two-
phase power from a three-phase source or vice versa. The Scott-T transformer is built with
two single-phase transformers of equal power rating. The MAIN and Teaser sections can be
enclosed in a floor mount enclosure with MAIN on the bottom and Teaser on top with a
connecting jumper cable. They can also be placed side by side in separate enclosures. Due
to this configuration, the losses in this two-system two-phase networks are about twice the
losses in a conventional three phase system, conservatively assuming losses are 50% greater
due to imbalance.
Because of this legacy system, there are still a number of Scott-T transformers operated by
UK Power Networks. Upgrading these transformers to three-phase transformers is a potential
strategy to reduce losses. In this context, we analyse and estimate the level of savings in
losses due to the upgrade of this type of transformers. The 30 most prominent distribution
sites and the level of losses reduction that can be expected if the Scott-T transformers are
replaced by the equivalent 3-phase transformers are listed in Table 21.
In the table, the 30 most prominent distribution sites are listed according to potential losses
reduction (highest on top). The potential losses reduction varies between 40-79 MWh/year per
site. However, there are about 307 Scott-T transformers in the SPN licence area. The total
potential losses reduction if all Scott-T transformers are replaced by three-phase transformers
is 6,635 MWh/year. The corresponding capitalised value is between £3.2-5.3m.

75
Table 21. List of distribution sites with Scott-T transformers and the potential losses reduction if they are replaced
by 3-phase transformers.

Losses annual savings

Site Reference List name
(MWh/year)
212139 ENMORE ROAD 78.6
212197 ASHBURTON ESTATE 73.3
212015 ELY ROAD SB 66.5
211522 ST AUGUSTINES AVE SB 59.5
212160 MORELAND AVENUE 58.9
211990 PENRITH ROAD 55.0
211840 WHITEHALL ROAD 54.6
211568 EDGEHILL ROAD 51.2
211825 ST JAMES ROAD 50.3
212145 CRAVEN ROAD 50.1
212007 SELHURST WELL 48.4
212095 DALMALLY PASS. SB 48.2
212141 LONGHEATH SB 47.7
212068 SYLVAN HILL 47.1
211950 ST.JOSEPHS COLLEGE 47.0
212001 RYEFIELD ROAD 46.0
212165 LESLIE PARK ROAD 45.8
212004 ALL SAINTS 45.5
211922 CRAIGNISH AVENUE 45.1
212167 BINGHAM ROAD SB 44.9
211892 PRIORY ROAD 44.6
212072 STONEY LANE SB 44.4
211912 BRIAR ROAD SB 42.2
211948 VIRGINIA ROAD 42.1
211902 FOREST GARDENS 41.6
212143 SPRING LANE SB 41.0
211718 CROHAM ROAD SB 40.7
211669 EPSOM ROAD SB 40.6
211930 HILLCOTE AVE SB 40.1
211690 LOWER COOMBE STREET 39.8

5.7 Impact of Distribution Transformer Density

The density of distribution transformers is another important factor with respect to the level of
losses in the network. Results on analysis aiming at capturing this dependency are presented
in Figure 5.13, assuming a typical 1979 distribution transformer losses specification. When
the density of distribution transformers is lower (7 DTs/km2), LV feeders supplying the load
need to be longer and more loaded, which results in greater losses on LV networks (indicated
by LL) and load losses of distribution transformers (indicated by DT LL). As a result, losses in
the HV network (indicated by HV) are also higher, given that this network supplies the LV
network. However, when density of distribution transformers is higher (14 DTs/km2), the
distribution transformer no-load losses (indicated by DT NLL) are greater.

76
 6%

5% PS NLL
PS LL
4%
Losses (%) HV
3%
DT NLL
2% DT LL

1% LV
SC
0%
S-urban, 7 DTs/km² S-urban, 14 DTs/km²

Figure 5.13. Breakdown of total losses in semi-urban networks for different densities of distribution transformers.

Losses in service cables (indicated by SC) and LV networks account for almost 50% of the
total losses in networks with a lower density of distribution transformers while this percentage
drops to 37% for a higher density. Distribution transformers’ load losses also drop from 12 to
9% while distribution transformers no-load losses increase from 13% to 27%, as illustrated in
Figure 5.14.

Figure 5.14. % breakdown of total losses in semi-urban networks for different densities of distribution
transformers.

Potential losses reduction by doubling the number of distribution transformes in semi-urban

networks are on average about 26%.
Figure 5.15 and Figure 5.16 present the respective breakdown of total losses for semi-rural
networks. The trends are similar, with the most important difference being the higher
proportion of HV network losses.

77
 8%
7%
PS NLL
6%
PS LL
Losses (%)
5%
HV
4%
DT NLL
3%
DT LL
2%
LV
1%
SC
0%
Semi-rural, 3 DTs/km² Semi-rural, 8 DTs/km²

Figure 5.15 Breakdown of total losses in semi-rural networks for different densities of distribution transformers.

Figure 5.16 % breakdown of total losses in semi-rural networks for different densities of distribution transformers.

Potential losses reduction by increasing the number of distribution transformers in semi-urban

networks are on average about 17%.

5.8 Impact of tapering

The effect of tapering was investigated using a model of tapered cable as shown in Figure
5.17.

Figure 5.17: Tapered cable model.

The specified section model had the following characteristics:

 The first section of 120 meters is 300 mm2 cable

 The second section of 114 meters is 185 mm2 cable
 The final section of 166 meters is 95 mm2 cable
 There are 130 connected customers uniformly distributed along the cable

78
  Diversified peak demand of 1.2 kW and coincidence factor13 of 0.1 are used to
estimate load and flow along cable
 Power factor of 0.96
 Peak utilisation of 100%
A coincidence factor is used to calculate the length of each tapered section such that none of
sections is overloaded at peak condition.
In this configuration, tapered cables increase losses by up to 25%. If cable peak utilisation
were lower the losses increase would be lower.

5.9 Summary
 A detailed analysis of non-load related replacement of distribution transformers is carried
out. If all distribution transformers classed by Health Index 4 and 5 are replaced by the
Eco design transformers the potential for losses reduction is 17 GWh per annum in the
UKPN area. Given the rate of annual replacement of distribution transformers, the savings
in losses could be potentially about 3.2 GWh per year.
 Once conditionally driven cable reinforcement is needed, investment in high-capacity
cables would be economically efficient for reduction in losses. However, it should be noted
that cables replacement is not driven by losses. Analysis was carried out to determine the
benefits in loss reduction by adopting minimum feeder cross section area of 185 mm2. This
would lead to reduction in losses in HV network in LPN area of 10% and EPN 40% and
SPN 32%. For LV networks, the benefits could be up to 52-63% depending on the area.
 Given that service cables typically supply a single customer, quantification of losses based
half-hourly energy consumption may underestimate losses. To inform this process,
analysis of losses is carried out using 5,000 five-second daily profiles and compared with
the amount losses obtained when half-hourly profiles are used. The actual losses are on
average 1.9 times greater compared with the losses calculated using half-hourly profiles
(the range is wide from 1.2 to 5.8), which will clearly impact the choice of cross-sectional
area of service cables.
 If single-phase HV spurs are converted to three phase, losses could be potentially reduced
by up to 80% in the corresponding network. Assuming the neutral path has the same
resistance as the phase conductor, this conclusion is independent from circuit loading and
conductor cross sectional area.
 Increasing the number of distribution transformers could potentially reduce losses between
17-26%.
 Removing tapering could potentially decrease losses by up to 25%.

13Coincidence and diversity factor are assumed to be directly opposite and proportional i.e.
multiplying each other gives 1. The coincidence factor is lower than or equal to 1 and the diversity
factor is greater than or equal to 1.

79
6 Conclusions

Comprehensive studies have been carried out to investigate losses drivers and to identify
opportunities and strategies for reducing network losses through improving system operation,
system design, and deploying loss-reduction technologies. The analysis quantified the
effectiveness of alternative strategies and identified the priority areas in UK Power Networks.
In order to carry out this analysis, a new modelling tool, called Loss Operation & Investment
Model (LOIM) was developed for detailed quantification of losses in real distribution networks,
from low voltage networks to grid supply points. This is in contrast to previous analyses of
network losses based on the application of representative distribution networks14. The LOIM
has also been applied to generate Losses Heat Maps for UK Power Network areas in order to
identify regions in which the volume of network losses are most significant. The effectiveness
of various network loss-reduction techniques in different UKPN areas was analysed in detail.
Core insights were produced regarding the business case for alternative loss mitigation
strategies and loss-driven network infrastructure investment.
The key findings of this work can be summarised as follows:
Quantification of network losses
The analysis carried out highlighted that more than 75% of network losses are associated with
LV networks, HV networks and distribution transformers. Overall:
 36-47% of the total losses are in LV networks
 9-13% of losses are associated with distribution transformer load related losses
 7-10% of losses are associated with distribution transformer no-load losses
 17-27% are in HV networks
 17-24% of total losses are in primary and grid transformers, and EHV and 132 kV
networks.
Understanding the contribution of different network sections to the total losses will be important
when identifying loss management strategies, assessing corresponding cost effectiveness
and determining the potential impact of those strategies.
Distribution of losses across network segments
Asset utilisation and circuit lengths are major losses drivers and hence their impacts have
been investigated and analysed across each region. UK Power Networks operate a wide
range of network types. These range from rural areas, such as parts of Norfolk and Suffolk, to
very densely populated urban areas like London. The corresponding peak demand density
varies from a very low 0.05 MW/km2, to a relatively high density of 137 MW/km2. In this context,
average utilisations of distribution transformers of 51% and 38% are observed in LPN and
EPN areas respectively.
Furthermore, the proportion of transformers which have a utilisation factor in excess of 70%
in LPN is 20%, while in EPN this figure is only 4%.
Detailed power flow modelling revealed that HV feeders in LPN deliver an average of 50%
more energy than feeders in EPN, while circuits in LPN are typically about 60% shorter than
in the EPN region. In this context, the analysis demonstrated that losses in LPN are primarily

14Imperial College London and Sohn Associates, Management of electricity distribution network
losses, supported by UKPN and WPD, 2014

80
driven by high network utilisation, while in EPN, losses are driven by long feeder lengths.
Overall, the LV network losses are comparable in both areas despite LPN LV networks having
significantly shorter lengths but higher loading. Conversely, losses in the HV networks are
greater in the EPN region.
The analysis demonstrated that the magnitudes of losses vary significantly across each
network type. Modelling quantified losses for more than 4,000 HV feeders, demonstrating a
relatively small number of HV feeders are characterised with high losses. About 70% of the
total losses are in 20% of the feeders. This clearly demonstrates that loss reduction initiatives
in HV networks should target a relatively small proportion of the feeders characterised by these
high losses. Undertaking a targeted approach will maximise the cost efficiency of this activity.
An unequal distribution of losses was noted in the LV network with more than 50% of losses
noted to occur in only 20% of LV feeders.
Based on advanced neural networks methodology, UK Power Networks’ HV feeders and LV
networks were classified into 22 clusters. These clusters were determined according to the
number of customers and their load characteristics, network length, rating, type and
construction. Average parameters for each cluster were quantified and corresponding
representative networks created. These included a range of rural and urban networks, and the
related loss performance for each was assessed.
As a significant amount of losses are associated with a small number of very specific feeders,
it should be noted that use of generic feeders with average parameters may not provide
appropriate evidence to inform the development of effective losses reduction strategies.
Identification of potential operational strategies for loss reduction
A number of key losses drivers were identified and analysed. Learning from this analysis can
be used to inform the development of future losses reduction strategies. These include
changes in network operational topology, improvement of power factor, changes in load
profile, controlling phase imbalance and harmonic distortion.
Key results of conducted case studies are as follows:
 Analysis demonstrated that Normally Open Point (NOP) reconfiguration could reduce HV
feeder losses by up to 15% in specific areas. The economic case for this operational
strategy, as a result, appears to be strong.
 For the three UK Power Networks licence areas feeders are ranked by the possible
reduction in losses driven by power factor improvement. The potential for loss reduction is
assessed assuming power factor improvement from 0.85 to 0.95. This would lead to
reduction in losses on each feeder between 11% and 14%. It is interesting that the
modelling demonstrated that improving power factor in only one third of HV feeders could
achieve 90% of potential losses reduction. Hence, the list of 30 highest ranked HV feeders
in each licence area is created and measurements of the actual power factor in future trials
are proposed to be carried out.
 It was noted that phase imbalance increases losses non-linearly. For example, phase
imbalance ranging from 10% to 30% would increase losses by 5% to 45% respectively.
As a consequence, we identified a list of 30 LV networks that would deliver the highest
benefits for imbalance improvement, based on the networks’ electrical characteristics.
 Implementing voltage management across UK Power Networks’ three licence areas could
potentially reduce losses by around 5%. Further investigation is required to understand
the voltage dependency of customer loads. Measurements are recommended to enhance
the understanding of voltage dependency in real time. This information will aid the
formation of future loss mitigation strategies. Performing actual measurements of voltage

81
 dependency of demand in different segments of the network should provide key
information related to the potential development of corresponding loss mitigation
strategies.
 Harmonic distortion is limited though network design standards, which ensure that the
impact of harmonic currents on networks are limted. The impact of voltage harmonics on
transformer no-load losses is linearly dependant on the total harmonic distortion (THD),
and hence, the impact on losses in this domain is more significant. Eco design
transformers’ iron losses are lower than previous transofmer specifications. The net effect
of this should mean that the impact of harmonic distortion on no-load losses will decrease
over time.
Application of smart-grid technologies for reduction of network losses

 Modelling demonstrated that the use of UK Power Networks’ Quadrature Booster, beyond
the network constraint management utilised by their Flexible Plug and Play (FPP) project15,
could deliver savings in the local network losses from about 11% in the case of high
demand and high distributed generation (DG) growth, up to 25% for low demand and low
DG growth.
 Furthermore, modelling demonstrated that optimally controlling the power factor of
distributed generators in the FPP project area could potentially reduce 33kV network
losses by 13%.
 Smarter Network Storage (SNS)16 installed in Leighton Buzzard to manage peak demand
and postpone network reinforcement (in addition to delivering system balancing services),
could potentially reduce losses in supplying circuits by about 15%.
 Modelling demonstrated that Soft Open Points (SOPs)17, installed for the management of
constraints in LV feeders, could potentially reduce losses in the corresponding LV network
and distribution transformers by about 10%-15%.
 Potentially further reduction in losses could be achieved by optimizing NOP positions in
real time to take into account changes in demand and generation.
 The former Department of Energy and Climate Change (DECC) indicated that smart
meters, combined with home display units, could reduce energy consumption by 2.8%18.
Analysis showed that correspondingly, distribution network losses would reduce by 5.5%
due to the decrease in consumption.
 Furthemore, analysis demonstrated that demand side response, which could potentially
shift 2.5% load from peak to off-peak period, would lead to a reduction of losses by about
3%.
Identification of efficient loss reduction investment strategies

 UK Power Networks could save 17GWh per annum by replacing all Health Index 4 and 5
distribution transformers with Ecodesign units. Given the current rate of replacement,
savings could reach up to 3.2 GWh per year.
 Loss reduction benefits alone are not sufficient to justify the upgrade of existing
underground cables. Howerver, when thermal constaints drive network reinforcement ,
installing cables of higher capacity would significantly reduce losses. In this context,

15 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Flexible-Plug-and-
Play-(FPP)/
16 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Smarter-Network-

Storage-(SNS)/
17 http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Flexible-Urban-

Networks-Low-Voltage/
18 https://publications.parliament.uk/pa/cm201617/cmselect/cmsctech/161/161.pdf

82
 analysis carried out to determine the benefits in loss reduction by adopting a minimum
feeder cross-section area of 185 mm². This would reduce LPN HV feeder losses by 10%.
The corresponding values for EPN and SPN are 40% and 32% respectively. Removing
tapering could potentially decrease losses by up to 25%. For LV networks, the benefits of
applying larger cables would be very significant, ranging from 52% to 63%, depending on
the area.
 Using 30-minute samples tends to understate network losses, particularly in service cables
that supply one customer only. To inform this process, 5,000 five-second samples from
the Low Carbon London (LCL)19 project were used comparatively. This modelling
demonstrated that applying higher sampling rates increases calculated losses by a factor
of 1.9 compared with the losses estimated using half-hourly profiles (the range is from 1.2
to 5.8). This further reinforces the case for significantly increasing the standard capacity
of service cables.
 If single-phase HV spurs are converted to three phase, losses could potentially be reduced
by up to 80% in the corresponding network.
Benefits of loss reduction strategies
Based on the analysis carried out, the capitalised value of the benefits associated with
alternative loss reduction strategies are summarised in Table 22. The annual capitalised
benefit is calculated by applying a discount rate of 3.5%.
Table 22 - Capitalised value of the benefits associated with alternative loss reduction strategies

Strategy Capitalised value Comment

NOP optimisation £5.4-8.9m LPN area
HV smart switches £2.6-4.3m LPN area
Multiple power factor SPN £48-80k Minimum for 30 ‘best’ HV feeders
correction per HV EPN £56-92k per each licence area if power
feeder LPN £53-88k factor is reduced from 0.85 to
0.95; the power factor is not yet
measured and hence potential
value might be lower of higher
Single point power SPN £25-41k Minimum for 30 ‘best’ feeders per
factor compensation EPN £30-49k each licence area if single point
per HV feeder LPN £28-46k compensation is installed.
Potential value depends on actual
power factor
Voltage control LV £1.5-2.4k per site Maximum expected value for
HV £9.2-15.2k per voltage dependent loads (constant
feeder power and constant impedance);
for mixes different types of loads,
i.e. constant power, constant
current and constant impedance
based loads, the benefits of
voltage control are marginal.

LV load balancing £0.9-13.6k per site LPN LV network

LV harmonics £200-300 per site

19http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-projects/Low-Carbon-London-
(LCL)/

83
 Strategy Capitalised value Comment
Primary transformer Negligible For typical transformer load and
de-energisation no-load losses, the benefit is
during low load negligible; in the event of high no-
conditions load losses relative to load
losses20, the potential benefit
could be £49-81k per substation
Eco-design £4-7.4k per Average savings per transformer
transformers transformer (392 transformers considered)
Amorphous £0.9-1.4k per PMT Average savings per PMT
transformers transformer (15 pole mounted
transformers (PMTs) considered)
Conductors LPN LV £63-104m All conductors lower than Al 185
rationalisation LPN HV £1.1-1.8m mm² are replaced with Al 185 mm²
EPN LV £114-188m conductors.
EPN HV £24-39m LPN HV network already uses
SPN LV £87-144m relatively higher conductor sizes
SPN HV £15-25m and hence benefit is relatively
lower than in EPN and SPN.
SPN HV voltage Min 11 kV £7.3-12m - SPN HV voltages 2.2, 3.3 and
rationalisation Min 20 kV £59-97m 6.6 kV are upgraded to 11 kV,
2,300 km of conductors
- All HV voltages are upgraded
to 20 kV, 17,700 km of
conductors
- Impact of transformers is not
taken into account
LPN EHV voltage £12-19m LPN 33 kV network is upgraded to
rationalisation 132 kV, 6,100 km of conductors;
Impact of transformers is not
taken into account
Smart distribution £7.4-12.3k per Minimum benefit per site
transformer21 secondary site [considering EPN 30 ‘best’ sites
for voltage control on LV network
(4% loss reduction), HV network
(5% loss reduction), power factor
improvement (8% loss reduction)
and phase imbalance reduction
(5% loss reduction)]
Scott connected £10.4-17.3k per site SPN LV networks supplied from
transformers 307 Scott connected transformers
Impact on Average savings on Savings are due to reduced active
transmission system National Grid’s power on UK Power Networks
networks of up to regions. Control of reactive power

20 High no-load losses imply older transformers, which based on life expectancy, could reduce the
indicative value of the capitalised benefits as these might be replaced, based on condition, before the
full benefits are achieved.
21 Typically distribution transformers are equpted by off-load tap changers to adjust for a seasonal

variation in expected voltage range. Smart distribution transformers could control voltage during
operatioin in order to, for example, reduce losses.

84
Strategy Capitalised value Comment
5.5% could be could potentially generate
achieved additional savings.

85
7 References

Elexon. (2017, 06 07). What is a load profile? Retrieved from Elexon:

https://www.elexon.co.uk/reference/technical-operations/profiling/
HVDC.ca. (2017, 05 23). HVDC. Retrieved from HVDC:
https://hvdc.ca/webhelp/EMTDC/Transmission_Lines/PI-
Section_Equivalent_Circuits.htm
Prysmian 11kV. (2017). BS6622/BS7835 Three Core Armoured 11kV.
Prysmian 33kV. (2017). BS6622/BS7835 Three core Armoured 33kV.
Prysmian LV. (2017). Waveform Cable.
Wikipedia. (2017, 06 06). Diversity factor. Retrieved from Wikipedia:
https://en.wikipedia.org/wiki/Diversity_factor#Coincidence_factor
UK Power Networks. (2014a). Flexible Plug and Play. London: UK Power Networks.
Retrieved from http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-
projects/Flexible-Plug-and-Play-(FPP)/
UK Power Networks. (2014b). Low Carbon London. London: UK Power Networks. Retrieved
from http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-
projects/Low-Carbon-London-(LCL)/
UK Power Networks. (2016a). Smarter Network Storage. London: UK Power Networks.
Retrieved from http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-
projects/Smarter-Network-Storage-(SNS)/
UK Power Networks. (2016b). Flexible Urban Networks Low Voltage. London: UK Power
Networks. Retrieved from
http://innovation.ukpowernetworks.co.uk/innovation/en/Projects/tier-2-
projects/Flexible-Urban-Networks-Low-Voltage/
House of Commons, Science and Technology Committee (2016). Evidence Check: Smart
metering of electricity and gas, Sixth Report of Session 2016-17. London: Science
and Techology Committee. Retrieved from
https://publications.parliament.uk/pa/cm201617/cmselect/cmsctech/161/161.pdf

86
8 Acronyms

AMT - Amorphous Metal Transformer

CRGO - Cold Rolled Grain Oriented Steel
CSA - Cross Sectional Area
DT - Distribution Transformer
EPN - Eastern Power Network
GMT - Ground Mounted Transformer
GSP - Grid Supply Point
LDR - Losses Discretionary Reward
LF - Peak Load Factor
LL - Load Losses
LLF - Load Loss Factor
LOIM - Loss Operation & Investment Model
LPN - London Power Network
OPF - Optimal Power Flow
NLL - No Load Losses
NOP - Normally Open Point
PF - Power Factor
PMT - Pole Mounted Transformer
SPN - South East Power Network
THD - Total Harmonic Distortion
UKPN - UK Power Networks

87