25th International Conference on Electricity Distribution Madrid, 3-6 June 2019

Paper n° 1608

LARGE SCALE PQ, TEMPERATURE AND ENERGY

MONITORING IN SECONDARY SUBSTATIONS

José María ROMERO GORDON

ENDESA – Spain
josemaria.romero@enel.com

fully compliant to IEC 61000-4-30 [1], but instead to

ABSTRACT deliver really useful data for the operation and
This paper deals with an ongoing project for monitoring maintenance of the distribution network.
distribution transformers and their associated low voltage Next follows the main features of these devices and their
network. The main goal is not only to deliver a set of installation:
statistics and figures for assessing power quality, but using
it for improving network operation and maintenance.  Similar price to low cost instantaneous voltage and
power monitors.
Instead of relying on complex data-management systems  Inbuilt large memory for storing power quality, event
for gathering and analysing the information throughput, a and waveform data.
simpler and low-cost edge-computing approach has been  In order to improve the expected lifetime no battery is
designed and built into the PQ device firmware. used.
 Current is measured by using split core transformers.
One of the most challenging issues has been the estimation  Indoor temperature monitoring.
and coding of the hot-spot temperatures based on existing  Estimation of the hot-spot temperature of transformer
based on IEC 60076-7 [2] and former IEC 60354 [3].
IEC standards.
These algorithms do take into account internal time
constants so they allow to keep track of the real
Moreover, the system is open for future enhancements transformer loading and aging. Moreover due to the
without incurring in massive and costly changes in central intrinsic power quality capabilities of the device,
data-management systems. harmonics are taken into account for the overloading
and aging algorithms. This is even more necessary
INTRODUCTION now, when a massive deployment of electric cars and
distributed generation is ongoing.
It is a widely extended throughout the world to deploy  Flexible communication ports and protocols for easy
power quality monitoring systems in primary substations, integration into existing and future networks.
mostly on medium voltage bus-bars. The main reason for  The most valuable data is available by simply
this well-known practice is not only the high cost of querying instantaneous magnitudes. Thus, there is no
devices, but also the exponential expenses on associated need to carefully download internal files for later
communication and data management centres. uploading and processing on a central location.
 3-phase power supply, thus avoiding lack of power
Even though there has been a massive deployment of smart when one phase is lost. Moreover, it allows detecting
energy meters, due to the intrinsic limited features of these this type of failure and reporting in advance.
devices, they do just provide some sort of rudimentary  Even without battery, it can record up to a few
monitoring which can only be useful for a first-time hundred milliseconds interruptions or very deep sags.
filtering. Thus low voltage monitoring is usually out of This data is available by querying the device on a
scope and limited to portable and expensive instruments. single query. When a medium voltage feeder trips,
every device along the line will be queried asking for
As result secondary transformers are most of the times the last recorded voltage dip and its depth. By
poorly monitored. Sometimes average power is measured matching this data to the feeder topology, a location
every 15 minutes, leading to some sort of overloading of the fault can be estimated.
monitoring. However this is not a widely extended practice  Direct connection of internal end-customers to an
throughout our distribution network, being applied on a inbuilt web interface with plenty of energy, aging and
miniscule set of transformers. This type of measurement is power quality data. The main goal is to avoid costly
not even a good signal for the real overloading and aging and unnecessary centralized data centres.
of a transformer.
The main benefits of this approach are as follows:
A new challenging project has started in order to improve
the monitoring of these transformers and their associated
 A closer look on the transformers, leading to better
low voltage network. Instead of using very expensive and
maintenance and planning works based on real
complex power quality monitors, we decided to use low-
loading and aging.
cost devices with plenty of features on energy and power
quality monitoring. These devices are not intended to be  Automatic and instantaneous location of faults along
medium feeders.

CIRED 2019 1/5

 25th International Conference on Electricity Distribution Madrid, 3-6 June 2019

Paper n° 1608

 Internal end-customers will have access to plenty of temperature might seem easier and cheaper, but hot-spot
event and power quality information without data temperature might be quite higher.
processing delays.
 A better look of the real power quality closer to the IEC 60076-7 and IEC 60354 formulas
end customers will be available.
 By matching energy measurements together with the IEC 60076-7 is entitled “Power transformers – Part 7:
smart meter data, many frauds might be detected and Loading guide for oil-immersed power transformers”. Its
thus non-technical losses fairly reduced.
predecessor IEC 60354 (“Loading guide for oil-immersed
power transformers”) is somehow simpler and easier to be
ON-SITE INSTALLATION implemented in a computer language, although does not
Each installation consists of a PQ device in a plastic change the underlying formulae.
housing together with voltage and current connectors (see
fig. 1) . There is also an external temperature probe (PT100 As stated in IEC 60076-7 table E.1, the hot-spot to top-oil
resistor) for measuring the indoor temperature. Current is (in tank) gradient at rated current for distribution
measured by split-core transformers of 1500/1 ratio. transformers can be approximated by 23ºC. According to
table 5 of this standard, a winding exponent can also be
Each device has an internal display for on-site verification fixed at 1.6. Thus by using its equation no. 5 the overall
of the most valuable data, .e.g. voltages, currents, powers, top-oil to hot-spot temperature rise can be approximated
power factors and temperatures. They do also have RS232, by
RS485 and Ethernet ports for connection to existing and
future communication appliances. Communication is 23𝐾 1.6 (1)
possible via MODBUS through the serial interfaces RS232
and RS485, or by HTTP or FTP through the Ethernet port. being K the load factor for each phase (RMS phase current
as a fraction of nominal current). At rated power this
formula leads to 26ºC, 52ºC for twice, etc. Assuming a
rated hot-spot temperature of 98ºC for non-thermally
upgraded insulation paper (IEC 60076-7 table 1), it is
obvious that the hot-spot temperature rise cannot be
neglected nor assumed constant.

For distribution transformers the top-oil temperature rise

can be taken as uniform within the tank. According to
equation no. 5 and table E.1 of this standard, this
temperature rise can be approximated by
0.8
1 + 5𝐾 2 (2)
55 ( )
1+5

fig. 1 PQ device inside plastic housing. being as high as 55ºC at full load and 132ºC at twice the
rated power. Since the top-oil temperature is assume to be
It is quite common for these meters to have separate uniform (see IEC 60354 page 31), the 3-phase average of
terminals for its power supply, so they can be connected to the oil formula will be taken as overall estimation of the
an UPS in a substation or an industry site. Although it was top-oil temperature rise.
decided to not use any battery due to the increasing costs
and early aging, it was mandatory to be able to keep it on These two aforementioned formulas correspond to the
during one phase lost. A very simple solution based on a steady-state temperature rises. Indeed both top-oil and hot-
half-wave rectifier was designed, thus creating a simple spot do have different time constants. IEC 60076-7
DC supply and improving the ride-through capability emulates such behaviour by two first-order decoupled
during voltage dips and interruptions systems, but the solution is stated as a set of differential
equations which are then transformed into difference
equations.
TRANSFORMER MAGNITUDES
Table 5 within the standard states that time constant for oil
Monitoring transformer temperature is most of the times
can be set at 180 minutes (3 hours) and 4 minutes for the
expensive and complex. Although there are very accurate
winding. Its annex C describes practical calculations of the
methods based on fibre optics probes (see IEC 60076-7),
time-dependant temperature rises with a 3-minute
they could only be applied on distribution transformers
resolution, which is not bad but of the same order of the
during the manufacturing process. Monitoring top-oil

CIRED 2019 2/5

 25th International Conference on Electricity Distribution Madrid, 3-6 June 2019

Paper n° 1608

winding constant. On the other hand the PQ device does Effect of harmonics
have quite powerful processing capabilities, so that time
step can be fairly reduced. Several tests were made down As stated above, IEC 60076-7 estimates temperature rises
to 1-second intervals, although a 5-second window was as a function of current (i.e. transformer load) and thus
selected because otherwise PQ processing was affected. losses. However these losses do not take into account the
However a 5-second time constant is two orders of harmonics effect, which can be indeed non-negligible. EN
magnitude shorter than the winding time constant, so it is 50464-3 [4] weighs up that loading factor as a function of
by far low enough. the eddy and joule losses according to the following
equation:
First order system approximations
𝑚
𝑒 𝐼1 2 𝐼ℎ 2
As stated above, IEC 60076-7 estimates that both top-oil 𝐾𝑓𝑎𝑐𝑡𝑜𝑟 = √1 + ( ) ∑ [ℎ𝑞 ( ) ] (6)
and winding are decoupled first-order systems. Then it 1+𝑒 𝐼 𝐼1
ℎ=2
tries to be as accurate as possible in the definition of the
differential equations and their associated difference
being e the ratio between eddy current losses at
equations. However in order to deploy a more stable
fundamental frequency and DC (typically 0.1), h the
algorithm with a similar degree of accuracy, a simpler and
harmonic order, q a factor that is dependent on the type of
probably better method is built. Instead of relying on
winding and frequency (typically 1.7 for distribution
incremental steps of temperature (which could lead to
transformers), Ii the i-harmonic current, m the highest
divergence within the algorithm), it emulates the first-
measured harmonic (e.g. 40) and I the full RMS current.
order system by smoothly approaching its steady-state
value.
Therefore each temperature raise must be weighted up
according to the above formula. The selected PQ device is
Being n the time step (in seconds) within the algorithm,
able to calculate every 200-ms up to 40th order harmonic,
any temperature rise is updated according to a -second
leading to a very accurate calculation.
moving window:
For instance, for an ideal 6-pulse rectifier (THDi=33%)
𝑛 ∙ 𝑇𝑟𝑖𝑠𝑒 |∞ + (𝜏 − 𝑛) ∙ 𝑇𝑟𝑖𝑠𝑒 |𝑛𝑜𝑤 (3)
𝑇𝑟𝑖𝑠𝑒 |𝑛𝑒𝑥𝑡 ∶= that value reaches 1.27, which means an overheating above
𝜏 25%. Nowadays there are plenty of loads based on than
scheme (e.g. air conditioners), so taking into account
The overall top-oil and hot-spot temperatures are
harmonics might be considered.
calculated as follows:
Next follows real data taken from a secondary substation:
𝑇𝑡𝑜𝑝−𝑜𝑖𝑙 = 𝑇𝑎𝑚𝑏𝑖𝑒𝑛𝑡 + 𝑇𝑟𝑖𝑠𝑒 𝑜𝑖𝑙 (4)

𝑇ℎ𝑜𝑡−𝑠𝑝𝑜𝑡 = 𝑇𝑡𝑜𝑝−𝑜𝑖𝑙 + 𝑇𝑟𝑖𝑠𝑒 ℎ𝑜𝑡−𝑠𝑝𝑜𝑡 (5)

The method stated in (3) gives a very good approximation

to the time-decay formula of a first-order system, being its
error always below 0.16%. In fact it is not possible to see
any difference when plotted together with the exponential
formula:

fig. 3 Current THD of an existing secondary substation.

fig. 2 Time-decay function of a first order system ( scale)

CIRED 2019 3/5

 25th International Conference on Electricity Distribution Madrid, 3-6 June 2019

Paper n° 1608

 Ambient and top-oil temperatures (average, lowest

and highest).
 Hot-spot temperatures (overall and per phase –
average, lowest and highest-).
 Average aging across the period (average and per
phase).

This data can be stored for weeks in the device and

downloaded either by MODBUS or FTP.

Energy records can be compared with downstream

measurements from smart meters and allow a better
detection of frauds.
fig. 4 Harmonics 5, 7 and 11 of an existing secondary
substation. FAULT LOCATION AND ANALYSIS
Another important tool to be exploited is the analysis of
It can be noticed that its THDi is sometimes close to that both MV and LV faults. In fact these PQ devices are able
upper burden (above 30%). Moreover harmonics up to 11 th to record voltage dips, sags and interruptions. It is true that
order are well below that value, thus measuring up to e.g. the lack of battery prevents a detailed study of any event,
n=15 might not be enough. but that drawback is partially amended by the DC supply
from the 3-phase half-wave rectifier.
Transformer aging
Analysis of LV faults
As stated in IEC 60076-7 (section 6.2), aging is basically
a factor of hot-spot temperature and type of insulation. For In case the event source is in the downstream LV network
distribution transformers it is rather uncommon to use (e.g. fuse is blown), the device does not shut down so a full
thermally upgraded paper. Therefore the instantaneous recording of the event is yet possible. A 10-ms RMS
rate of aging can be estimated as recording for any voltage and current phase is made
𝑇ℎ𝑜𝑡−𝑠𝑝𝑜𝑡 −98
available few seconds prior and after the event, as well as
(7) a 50-µs transient recording.
𝑉=2 6

leading exactly 1.0 for 98ºC. By using equations (1), (2) Analysis of MV faults
and (7), it can be inferred that a typical distribution
transformer is designed for reaching exactly 98ºC at full
power (K=1), ambient temperature of 20ºC and top-oil The main goal should be to better estimate the fault
temperature of 20+55=75ºC. location along the feeder. Since there is no battery, there is
a serious risk of shutting down the device. However there
That rate of aging can be calculated per phase. is always some energy available and indeed the DC-supply
improves the ride-through performance.

PQ-device implementation The device is able to work when fed by a single AC-supply
and voltage drops below 40%. When fed by the three
As previously stated, temperature calculations are updated phases and previously half-rectified, it means a real hard
every 5 seconds. This gives a very good trade-off between three-phase dip or an interruption. Lab testing has
performance, accuracy and resolution. demonstrated that it takes around 300 ms until the device
shuts down when subjected to a full interruption, being this
As any PQ analyser, this device is able to periodically time gap below the time that a feeder protection takes for
record measured and calculated variables. Together with a very hard three-phase short-circuit. Therefore the real
the typical 10-minute averaging window for PQ shutting down problem will be indeed noticeable for trips
magnitudes (e.g. voltages, currents, voltage and harmonic of its own MV feeder.
currents, unbalance, etc.) a larger 15-minute window is
selected for energies, temperatures and aging. In short, When this happens, the device does not intend to be
these are the full set of variables: accurate on recording the voltage dip, but fast and useful.
In this sense it starts recording in flash memory the three
 Active, reactive and reactive energies (overall and per ongoing RMS voltages. If it suddenly shuts down, when
phase). Reactive energy is split into four quadrants. power is recovered it can be queried via MODBUS and get

CIRED 2019 4/5

 25th International Conference on Electricity Distribution Madrid, 3-6 June 2019

Paper n° 1608

those three measurements and when they happened. The tests. By far this is the most accurate solution, although
idea is to query any PQ device along the feeder as soon as consisting of very few samples. Therefore it should be
the feeder is back, without bulky and slow PQ files combined with the massive on-site verification. Effect of
processing. Thus a residual-voltage image can be obtained harmonics should also be an important topic to be analysed
along the MV feeder and matched with the existing together with these manufacturers.
topology, leading into an instantaneous fault locator.
FUTURE ENHANCEMENTS
The most typical distribution transformers in ENDESA by
far are Dy. Except for very specific regions with MV There are certain fields nowadays out of the scope of this
isolated neutral, the most common neutral grounding in paper, so they could be investigated on future stages:
substations is by resistor or impedance. Thus one-phase
MV faults are not usually noticeable on LV networks  Dry-type transformers.
unless there is a subsequent interruption.  Outdoor oil-immersed transformers, since the effect
of direct solar radiation, wind and rain makes very
However the main goal is mainly to help fault location hard to estimate the exchanged heat across
when directional fault passage and voltage loss indicators transformer’s surface.
are not installed.

MODEL VALIDATION CONCLUSIONS

The most challenging and newest issue is the estimation of ENDESA has initiated an ambitious project for monitoring
temperature and aging of transformer. Although formulas distribution transformer within secondary substations.
have been taken from the well-stablished standard IEC Well stablished formulas from IEC standards have been
60076-7 (dated back to 2005, not even changed for used for temperature estimation. Moreover these formulas
distribution transformers from the legacy standard IEC have been improved for taking into account both
60354 from 1991), an on-site verification of temperatures harmonics and current unbalance.
is mandatory.
A low-cost but flexible and powerful PQ device has been
The existing distribution transformers within the ENDESA employed, thus allowing recording of PQ magnitudes,
network lack of any in-tank temperature probe for remote energies and temperatures. Other tools have been
measurement. However many of them do have simple implemented, such as aided assistance for fault location.
temperature gauges which can be easily inspected. The PQ Inbuilt standard features such as event and transient
device has been programmed for displaying on its LCD recording will allow a better assessment of customers’
ambient, top-oil and overall hot-spot temperatures. Thus it power quality.
is quite straightforward for trained personnel to take visual
measurements periodically and asses the top-oil REFERENCES
temperature. Next follows a snapshot of the LCD display:
[1] IEC, 2015, IEC 61000-4-30, Electromagnetic compatibility
(EMC) - Part 4-30: Testing and measurement techniques - Power
quality measurement methods.

[2] IEC, 2005, IEC 60076-7, Power transformers - Part 7: loading

guide for oil-immersed power transformers.

[3] IEC, 1991, IEC 60354, Loading guide for oil-immersed power
transformers.

[4] CENELEC, 2010, EN 50464-3, Three-phase oil-immersed

fig. 5 LCD snapshot with ambient, top-oil and average hot-spot distribution transformers 50 Hz, from 50 kVA to 2 500 kVA with
temperatures. highest voltage for equipment not exceeding 36 kV - Part 3:
Determination of the power rating of a transformer loaded with
non-sinusoidal currents
By using these on-site visual inspections, the model can be
readjusted and both constant values as well as the
algorithm itself can be upgraded. In fact the PQ device can
be reprogrammed without reloading the full firmware
image, just the required algorithms, which makes it a
robust and flexible solution.

It will be also explored future collaborations with

transformer suppliers in order to make intensive in-factory

CIRED 2019 5/5