View metadata, citation and similar papers at core.ac.

uk brought to you by CORE

provided by Illinois Digital Environment for Access to Learning and Scholarship Repository

c 2019 Christopher G. Sain

 AN OPENDSS IMPLEMENTATION OF A GENERIC MUNICIPAL
MICROGRID FOR CO-SIMULATION

BY

CHRISTOPHER G. SAIN

THESIS

Submitted in partial fulfillment of the requirements

for the degree of Master of Science in Electrical and Computer Engineering
in the Graduate College of the
University of Illinois at Urbana-Champaign, 2019

Urbana, Illinois

Adviser:

Professor Peter W. Sauer

 ABSTRACT

This thesis provides a summary of the development process of a microgrid

simulation model using OpenDSS software, as well as simulations and co-
simulations using said model. Many power system research problems may
be solved via the deployment simulations. However, for real-world problems
the computational efforts for detailed dynamic modeling may be impractical
or excessive. OpenDSS provides a framework within which a model of a
small- or large-scale system may be implemented without the representation
dynamics, with extensive co-simulation capabilities. We discuss the modeling
of a generic 8-bus microgrid that consists of 200 residential loads plus an
additional load for the local control building, three generation resources –
solar, wind, and gas – and a battery storage resource. We use historical
environmental data from Decatur, Illinois, together with realistic consumer
load shapes to simulate and analyze various unbalanced and quasi-balanced
situations. In addition, we present results of co-simulation studies on such a
model in an OpenDSS application to evaluate various potential scenarios.

ii
For Booster, my most loyal friend.

iii
 ACKNOWLEDGMENTS

I would like to thank my advisor, Professor Peter W. Sauer, for his guidance,
support, and wealth of knowledge over the course of my master’s study. I
would also like to thank Professor Alejandro D. Domı́nguez-Garcı́a and the
members of the ARPA-E NODES group for their assistance and dedication
during my time working with them. Finally, I would like to thank all my
family and friends who helped and supported me along the way.

iv
 TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

LIST OF ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . viii

CHAPTER 1 INTRODUCTION . . . . . . . . . . . . . . . . . . . . 1
1.1 Literature Survey . . . . . . . . . . . . . . . . . . . . . . . . . 2

CHAPTER 2 THE MICROGRID . . . . . . . . . . . . . . . . . . . . 4

2.1 One-Line Representation . . . . . . . . . . . . . . . . . . . . . 4
2.2 Generators and Storage . . . . . . . . . . . . . . . . . . . . . . 6
2.3 Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

CHAPTER 3 MODELING IN OPENDSS . . . . . . . . . . . . . . . 8

3.1 Buses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.2 Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
3.3 Initial Simplified Model . . . . . . . . . . . . . . . . . . . . . . 9
3.4 Refined Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
3.5 Auxiliary Implementation of Dynamics . . . . . . . . . . . . . 20

CHAPTER 4 AN APPLICATION IN DISTRIBUTED CONTROL

STUDIES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
4.1 Co-simulation Goals . . . . . . . . . . . . . . . . . . . . . . . 24
4.2 Microgrid Model . . . . . . . . . . . . . . . . . . . . . . . . . 25
4.3 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

CHAPTER 5 CONCLUSIONS AND FUTURE WORK . . . . . . . . 31

REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

v
 LIST OF TABLES

2.1 DER Parameters . . . . . . . . . . . . . . . . . . . . . . . . . 6

2.2 Transformer Parameters . . . . . . . . . . . . . . . . . . . . . 7

3.1 Buses in Network . . . . . . . . . . . . . . . . . . . . . . . . . 8

3.2 Distribution Line Parameters . . . . . . . . . . . . . . . . . . 9

vi
 LIST OF FIGURES

2.1 One-line diagram of the microgrid to be modeled . . . . . . . 5

3.1 Real power generated in preliminary test . . . . . . . . . . . . 10

3.2 Tie line voltage in preliminary test . . . . . . . . . . . . . . . 11
3.3 Inverter efficiency curve . . . . . . . . . . . . . . . . . . . . . 12
3.4 Historical solar data, Decatur, IL, 07/21/2017 . . . . . . . . . 13
3.5 Historical wind data, Decatur, IL, 07/21/2017 . . . . . . . . . 14
3.6 Generated load shapes . . . . . . . . . . . . . . . . . . . . . . 15
3.7 Combined neighborhood load, per phase . . . . . . . . . . . . 16
3.8 Simulated generation, full day . . . . . . . . . . . . . . . . . . 17
3.9 Simulated power exchanged on tie-line, full day . . . . . . . . 17
3.10 Simulated voltage on tie-line, full day . . . . . . . . . . . . . . 18
3.11 Simulated voltage on neighborhood bus, full day . . . . . . . . 18
3.12 Artificially inflated loads . . . . . . . . . . . . . . . . . . . . . 19
3.13 Neighborhood voltage response to artificially inflated loads . . 20
3.14 Generator dynamic test . . . . . . . . . . . . . . . . . . . . . . 22
3.15 Dynamic tie-line exchange . . . . . . . . . . . . . . . . . . . . 22

4.1 System architecture for C-HIL co-simulation . . . . . . . . . . 27

4.2 Simulation results for “stepped ramp” input signal . . . . . . . 28
4.3 Simulation results for first 5 minutes of PJM RegD signal . . . 29
4.4 Simulation results for second 5 minutes of PJM RegD signal . 30

vii
 LIST OF ABBREVIATIONS

ADN Active Distribution Network

DER Distributed Energy Resource

EPRI Electric Power Research Institute

HIL Hardware in-the-Loop

ISO Independent System Operator

PQ Active and Reactive Power

PV Power and Voltage

RTO Regional Transmission Organization

TAC Technology Applications Center

TCUL Tap Change under Load

UCSD University of California, San Diego

UIUC University of Illinois at Urbana-Champaign

viii
 CHAPTER 1

INTRODUCTION

When simulating power systems for research applications, often there arises
the need for simulation tools that are not necessarily in the scope of the
problem. For example, when attempting to test a system’s response to a
control algorithm, the researcher requires a system on which to apply the
control scheme. It would be valuable to have a system model already created
that was made available for such purposes. The topic of this thesis is to
create such a model.
In particular, the goal is to create a “plug-and-play” model of a microgrid
which may be used to test the efficacy of control algorithms under develop-
ment. Because of the close ties between the University of Illinois at Urbana-
Champaign (UIUC) and Ameren Illinois, the microgrid to be modeled is
based upon Ameren’s Technology Applications Center (TAC) microgrid [1],
located next to the university. This model will be used by the NODES group
in the Power and Energy Systems area at UIUC, in support of ongoing re-
search into distributed control algorithms for distributed energy resources
(DERs).
The simulation framework chosen for this model is OpenDSS, developed
by the Electric Power Research Institute (EPRI). This software was chosen
for several reasons.

1. It is free and open source, making the model easily transferred, modi-
fied, and updated as needed.

2. It has extensive co-simulation capabilities through its COM server, al-

lowing for a driver program to run the simulation from almost any
programming language/environment [2]. The driver used in this thesis
is Matlab.

3. It handles the simulation of both balanced and unbalanced networks.

1
 4. It does not implement full dynamic modeling natively, allowing for a
lighter load on computation resources and fast solution convergence.

5. It is widely used in industry for simulation of large and small scale

networks, yielding comprehensive support documentation [2].

Rather than implementing a full dynamic model of a power system, which

is often infeasible or unnecessary, the system may be modeled in OpenDSS,
allowing for fast simulation yielding the steady-state response of the system
to inputs. However, if dynamic response is desired, they may be added
piecewise to the model via the driver program. This allows any critical
dynamic performance to be modeled, while not requiring the entire network
to be simulated as such. If desired, complete system dynamics could be
added to the simulation [3].
This thesis is organized in the following manner. Chapter 1 outlines and
motivates the problem to be solved, with a survey of contemporary research
being performed in this area. Chapter 2 provides an in-depth definition of the
microgrid: the nature of its network and topology, as well as power conversion
and delivery elements. Chapter 3 summarizes the process of developing the
model of such a microgrid using OpenDSS and preliminary simulations of its
behavior. Lastly, Chapter 4 presents the results of an application for this
simulation framework, using a similar microgrid model in co-simulation with
a distributed control architecture currently under development at UIUC.

1.1 Literature Survey

Applications of open-source modeling are widespread, from industry super-
visory control and data acquisition (SCADA) systems to university research.
The most abundant literature concerns modeling of photovoltaic systems
for impact studies. The authors of [4] seek to use OpenDSS to supplement
existing models for a hybrid time-and-frequency domain simulation of the
implementation of such photovoltaics on a large scale in an urban setting.
They have chosen OpenDSS because many such hybrid simulations have been
performed on the transmission level, rather than distribution.
Other applications of OpenDSS include interconnection studies [5], real-
time simulation [6], and voltage control [7]. These papers focus on the ef-

2
fects of high-penetration levels of photovoltaics into the grid, and how they
can cause problematic voltage conditions around the interconnections. The
models attempt to take into account variations of residential and commer-
cial loads and how different control schemes may assist in mitigating the
over/undervoltage events at the interconnections. This includes Volt/VAR
and Volt/Watt control, as well as the introduction of battery storage units
into the network.
The authors of [6] describe their use of OpenDSS in real-time hardware-
in-the-loop (HIL) studies similar to those that constitute the subject of this
thesis. They use their HIL system to simulate their controls and FPGA hard-
ware for the IEEE 13-bus test feeder, while supplementing their simulation
with OpenDSS to simulate power flow in their distribution system. However,
their OpenDSS simulation differs from the topic of this thesis in that it is
not a microgrid; it is a standard distribution system with a substation and
load buses, lacking generation and storage capabilities.
The authors of [8–10] use OpenDSS and a similar distribution framework,
GridLAB-D, to model active distribution networks (ADNs) independent of
other simulations. The simulations run in these studies are an attempt to
use the various DERs to dynamically maintain power balance and optimize
generator cost. Particularly, they emphasize control of storage resources and
load shaping.
The great advantage of OpenDSS, evident in the current applications, is its
flexibility to be applied effectively in different programming environments.
Among the mentioned papers are OpenDSS implementations in Matlab,
Python, C/C++, Pascal, and even Microsoft Excel. The literature highlights
the co-simulation capability of the simulation framework, one of the chief
reasons it was chosen for this application.

3
 CHAPTER 2

THE MICROGRID

This chapter focuses on the configuration and parameters of the microgrid

to be simulated by OpenDSS. The network topology will be explored as gen-
erators, transformers, loads, and lines will be enumerated in detail. This
microgrid is inspired by the TAC microgrid located in Champaign, Illinois.
As the name suggests, this microgrid is mainly for research purposes, in-
cluding testing of infrastructure resources, new distribution equipment, and
development of smart grid technologies. Ameren Illinois and the University
of Illinois at Urbana-Champaign can perform research jointly using these re-
sources before wide-scale adoption [1]. In particular, this microgrid model
will be used in research applications centered around distributed control of
DERs to provide ancillary services to the bulk grid [11].

2.1 One-Line Representation

The microgrid to be modeled is illustrated by the one-line diagram shown in
Figure 2.1. This model consists of 219 nodes on 8 buses. The diagram shows
the high-level topology of the microgrid, including power generation, trans-
formers, loads, transmission lines, and the tie-line connecting the microgrid
to the bulk grid.

4
 Figure 2.1: One-line diagram of the microgrid to be modeled

A top-level description of the network is as follows: an infinite 69 kV bus

representing the bulk electrical grid is connected through a transmission line
to a 69 to 12.47 kV transformer, located at the microgrid substation. This
transformer is connected through transmission lines to four more transform-
ers, 12.47 kV to 480 V, each connected to an independent DER: photovoltaic
solar array (through an inverter), wind turbine, battery storage (through an
inverter), and a natural gas driven synchronous generator. There are two
load buses connected to the 12.47 kV substation bus: a single, 3-phase load
representing the local substation controls and monitoring building, and a
bus connecting to 200 single phase loads representing a large neighborhood.
Both loads are connected to transformers stepping down the 12.47 kV to 208
V, with the addition of the neighborhood transformer having an on-load tap
changer. This tap changer utilizes 33 taps (16 both above and below a cen-
ter tap) at 0.625% per-tap for a total of nominal ±10%, acting as a voltage
regulator for the neighborhood.

5
2.2 Generators and Storage
The microgrid contains a total of four energy resources: solar array, wind
turbine, natural gas generator, and lithium-ion battery storage. The two DC
sources (solar and battery) are each connected to inverters, producing 480 V.
The wind turbine and gas generators each produce power at 480 V as well.
The generators and their respective parameters are tabulated in Table 2.1.

Table 2.1: DER Parameters

Element Voltage [V] Capacity [kW] Note

GAS 480 1000

SOLAR 480 125 Inverter
WIND 480 100
BATT 480 250 Inverter, rated 500 kWh

2.3 Transformers
There are a total of 7 transformers in the system. Their parameters are
outlined in Table 2.2 (note that the superscripts P and S refer to “primary”
and “secondary”, respectively). As mentioned previously, the transformer
responsible for the neighborhood connection is a tap-changing-under-load
(TCUL) transformer, acting as a voltage regulator for the neighborhood bus.
For the sake of this model, the tap changer was set to monitor the voltage
on phase C of the load bus (chosen arbitrarily). Because load magnitudes
are reasonably balanced in the testing scenarios presented, this single-phase
monitoring is sufficient to control the voltage.

6
 Table 2.2: Transformer Parameters

Element VoltageP [kV] VoltageS [kV] BusP BusS

T SUB 69 12.47 B GRID B SUB

T SOLAR 12.47 0.480 B SUB B SOLAR
T WIND 12.47 0.480 B SUB B WIND
T GAS 12.47 0.480 B SUB B GAS
T BATT 12.47 0.480 B SUB B BATT
T CONTR 12.47 0.208 B SUB B CONTR
T NEIGHB 12.47 0.208 B SUB B NEIGHB

7
 CHAPTER 3

MODELING IN OPENDSS

3.1 Buses
The OpenDSS framework requires the explicit enumeration of buses in the
system. Each element must be connected to one or more buses, depending
on the element type. A complete list of buses is shown in Table 3.1.

Table 3.1: Buses in Network

Bus Voltage [V]

B GRID 69k
B SUB 12.47k
B SOLAR 480
B WIND 480
B BATT 480
B GAS 480
B CONTR 208
B NEIGHB 208

These buses form the framework of the model. Every element within the
network is connected to one or more buses either directly or through a dis-
tribution line.

3.2 Lines
There are 10 transmission lines in the network. They are modeled based upon
per-unit-length impedance taken from the EPRI Test Circuit 5 [2]. In the
network, there exists a transmission line between any two connected nodes,

8
with two exceptions. The distance between the energy resources and their
corresponding transformers was deemed negligible, and thus line impedances
between these nodes are neglected. Additionally, the 200 single-phase neigh-
borhood loads are not all individually connected via separate lines. The
transmission line traveling from the substation to the neighborhood is mod-
eled, whereafter all the loads are connected directly to its termination, with
no final “line-to-home” impedance. Line connections and parameters includ-
ing per-unit-length impedance at 60 Hz are detailed in Table 3.2.

Table 3.2: Distribution Line Parameters

Ω Ω nF
Line Conn 1 Conn 2 L [km] R [ km ] X [ km ] C [ km ]

L1 B GRID T SUB P 1.0 0.0656 0.193 3.165

L2 T SUB S B SUB 0.2 0.0656 0.193 3.165
L3 B SUB T SOLAR S 0.3 0.0656 0.193 3.165
L4 B SUB T WIND S 0.1 0.0656 0.193 3.165
L5 B SUB T BATT S 0.1 0.0656 0.193 3.165
L6 B SUB T GAS S 0.1 0.0656 0.193 3.165
L7 B SUB T CONTR P 0.1 0.0656 0.193 3.165
L8 T CONTR S B CONT 0.01 0.0656 0.193 3.165
L9 B SUB T NEIGHB P 2.0 0.0656 0.193 3.165
L10 T NEIGHB S B NEIGHB 1.0 0.0656 0.193 3.165

3.3 Initial Simplified Model

The model was initially created in a highly reduced fashion. All generator
buses were treated simply as PV buses (real power and voltage given), and
load buses were treated as PQ buses (complex power given). Transformers
were ideal, shifting voltage levels with no losses. From here, the network
was connected using the parameters given in Tables 2.1, 2.2, 3.1, and 3.2.
OpenDSS was used as a power flow calculator for the network, solving the

9
standard power flow equations [12].

N
X
Pk = |Vk ||Vj | (Gkj cos (θk − θj ) + Bkj sin (θk − θj )) (3.1)
j=1
N
X
Qk = |Vk ||Vj | (Gkj sin (θk − θj ) − Bkj cos (θk − θj )) (3.2)
j=1

Here Pk and Qk are the net real and reactive powers injected at bus k, and
Gkj and Bkj are the real and imaginary components of the kj entry in the bus
admittance matrix. OpenDSS calculates the admittance matrix and solves
the equations.
Using this ideal model, a preliminary no-load simulation was run to ensure
the proper setup of the model. In this scenario, the DERs were assigned
arbitrary outputs, in keeping with the PV nature of the model. These outputs
were varied 5 times, with OpenDSS simulating the network behavior after
each variation. Figures 3.1 and 3.2 show the per-phase real generation and
tie-line voltage for this simulation.

Figure 3.1: Real power generated in preliminary test

10
 Figure 3.2: Tie line voltage in preliminary test

This initial no-load simulation shows that the model does seem to be work-
ing as intended, with the generators causing voltage effects on the tie-line (the
expected increase in generation causing an increase in voltage). However, the
changes to the system are instantaneous and not reflective of real-world be-
havior. To remedy this, the model was refined with the goal of successfully
simulating an “normal” day in central Illinois.

3.4 Refined Model

3.4.1 Inverter-Based Source Models

Rather than simply arbitrarily assigning power outputs to the solar and bat-
tery units, a more realistic representation is desired.
The active power output of the solar array is a function of the solar irra-
diance Ee , ambient temperature T , and a power rating. This power rating,
Pmpp , is equal to the real power produced at the maximum power point for
a given temperature (this case 25◦ C) and irradiance of 1.0 kW m2
[13]. The
inverter behavior is abstracted out by applying an efficiency factor (ηinv ) to

11
the array output [14]. Therefore the output power of the solar array given
certain environmental parameters can be calculated.

Parray = Pmpp · Ee · Kt · ηinv (3.3)

The temperature factor Kt is provided by the manufacturer of the panel. For

this simulation, the factor from the Yingli 60-series solar modules in use at
the TAC was used [15].

Kt = 1 − 0.17 · (T − 25) (3.4)

The efficiency factor for the inverters used for both the solar array and the
battery unit was taken from the California Energy Commission (CEC) ex-
perimental efficiency database matching the voltage and rated power of this
solar array [14]. The efficiency curve [14] is shown in Figure 3.3.

Figure 3.3: Inverter efficiency curve

As shown, the AC power output of the solar array in the model will have
a maximum of ≈ 97% of its maximum output before the inverter at a given
temperature, with a minimum of ≈ 86% at lower power levels, such as during
overcast periods, or the sunrise/sunset periods. This inverter efficiency curve
is also used for the battery storage unit for simplicity.

12
3.4.2 Historical Location Data
To perform as accurate a daily simulation as possible, it is crucial to use
realistic data in setting power output levels for the sources in the system. As
such, historical environmental data from Decatur, Illinois, (approximately
50 miles southwest of Champaign), is used to calculate the outputs of the
solar and wind energy sources. This data includes temperature (T ), solar
irradiance (Ee ), and wind speed (vwind ) for July 21, 2017 [16–18]. It was
assumed for this simulation that the wind turbine is always optimally aligned
with the wind direction. The temperature and solar irradiance data for the
chosen date are shown in Figures 3.4 and 3.5.

Figure 3.4: Historical solar data, Decatur, IL, 07/21/2017

13
 Figure 3.5: Historical wind data, Decatur, IL, 07/21/2017

The solar data was fed into the model based upon Equations 3.3 and 3.4.
The wind data was used to calculate the power output of the wind turbine
using equation 3.5.

1 3
Pout = ρAvwind Cp (3.5)
2

Here ρ is the density of the air at the altitude and mean temperature of
Decatur (ρ ≈ 1.17 mkg3 ), A is the swept area of the blades (A ≈ π · 102 =
314.2 m2 given 10 m blade length). The power coefficient Cp is set to the
Betz limit of 0.59, assuming maximum efficiency of the wind turbine [19].
These power outputs were calculated for each data point and then fed to the
model.

3.4.3 Load Models

To accurately simulate a day’s operation, it is necessary to develop loadshapes
to describe the power draw of the neighborhood loads. In this scenario, the
creation of 200 independent load shapes is infeasible. To solve this problem,
3 independent loadshapes were developed, with each household load taking

14
on a linear combination of each shape. Using the method described in [20],
these loadshapes were determined based on the assumption that most homes
contain certain common appliances (refrigerator/freezer, electric oven/stove,
microwave oven, heating/cooling, television, lighting), while a smaller num-
ber of homes contain more “luxury” appliances (clothes washer/dryer, dish-
washer, personal computer). Note that due to the fact that the day chosen
to be simulated is July 21st , air-conditioning was assumed to be in heavy use
for most of the day. The generated load shapes are plotted in Figure 3.6.

Figure 3.6: Generated load shapes

These three shapes form the basis of the neighborhood loads. In the Mat-
lab driver, each of the 200 residential loads is formed by a linear combination
of these shapes.

loadn = kn1 S1 + kn2 S2 + kn3 S3 (3.6)

Here S1 , S2 , and S3 are the load shapes, and kn1 , kn2 , and kn3 are randomly
chosen constants between 100 and 700, such that the power draw per home
lies between a maximum of 2.1 kW and minimum of 300 W [20]. These
loads are then evenly distributed across the three phases of the load bus (66
phase A, 66 phase B, and 67 phase C), in an attempt to make the system

15
quasi-balanced. The total three-phase load of the neighborhood is shown in
Figure 3.7.

Figure 3.7: Combined neighborhood load, per phase

Because the function of the control building is constant monitoring and

control, the load is assumed for the purposes of this simulation to be constant.
A value of 65 kW was chosen. It must be lit and climate controlled for the
entire duration of the day, along with sensory and control equipment being
run constantly, making this a fair assumption.

3.4.4 Full-Day Simulation

The simulation was run for the full day model, using the real-world data from
July 21, 2017. For simplicity, it was assumed that the gas generator was
running at 90% capacity for the duration of the day, and that the microgrid
was operating in grid-connected mode; that is, power was able to freely pass
either direction along the tie line. The results of the simulation are shown in
Figures 3.8 through 3.11.

16
 Figure 3.8: Simulated generation, full day

Figure 3.9: Simulated power exchanged on tie-line, full day

17
 Figure 3.10: Simulated voltage on tie-line, full day

Figure 3.11: Simulated voltage on neighborhood bus, full day

Figure 3.8 shows the generation of all modeled sources. The renewable
sources generally follow the previously discussed curves, with the addition

18
of the sharp “turn-on” step of the solar array, taking the minimum power
capability of the attached inverter [21]. The battery storage in this case is
used in an attempt to ease the burden on the bulk grid during the heavy
power-consuming portion of the day. It charges during the low-load period
overnight, and discharges during peak hours. Figure 3.9 shows the power
exchanged with the bulk grid through the tie-line. Negative values indicate
power flowing into the microgrid, and positive values indicate power flowing
out of the microgrid. The simulation indicates that for this particular day
and with the gas generator running at 90% capacity, the microgrid both
draws power from and supplies power to the grid depending mostly upon
the demand of the neighborhood. The voltage plots in Figures 3.10 and
3.11 show that across the day, the tie-line voltage varied by approximately
0.15% and the neighborhood voltage varied by approximately 0.6%. The
voltage swings on the neighborhood bus were not enough to trigger the tap
changer (±0.625%), so it remained at the center tap for the duration of the
day. A subsequent simulation was run with artificially inflated load values
to specifically verify that the tap changer behaves as desired. The results of
that simulation are shown in Figures 3.12 and 3.13.

Figure 3.12: Artificially inflated loads

19
 Figure 3.13: Neighborhood voltage response to artificially inflated loads

As the neighborhood load increases, the voltage starts to drop. When it

has dropped by the requisite 0.625% = 0.75 V, the tap changer (monitor-
ing Phase C) switches to tap 1 to boost voltage back to desired levels, as
expected.

3.5 Auxiliary Implementation of Dynamics

Though OpenDSS simulates the network discretely, dynamics can be added
in an auxiliary fashion by the driver program [2, 3]. Rather than issuing the
simulator step commands as in previous simulations, the operator can define
the desired dynamic behavior within Matlab. For example, dynamics were
added to the gas generator by calculating its dynamic response to a step
command in output power. This response was calculated, and then fed to
the simulator, allowing the network to be solved for this dynamic behavior.

20
3.5.1 Generator Model
The synchronous generator was modeled as described in [12] in dq0 rotor-flux
reference frame. This model is outlined in the following equations.

λd = Ld id + Laf if (3.7)
λq = Lq iq (3.8)
3
λf = Laf id + Lf f if (3.9)
2
λ0 = L0 i0 (3.10)
dλd
vd = Ra Id + − ωm λq (3.11)
dt
dλq
vq = Ra iq + + ωm λd (3.12)
dt
dλf
vf = Rf if + (3.13)
dt
dλ0
v0 = Ra i0 + (3.14)
dt
3 p
Tmech = (λd iq − λq id ) (3.15)
2 2

Using this model, machine dynamics can be calculated. Of particular interest

is the machine response to step commands in power output. Using Matlab,
the dynamic response to such a command is able to be quickly calculated.
This response, rather than the step command itself, is then passed to the
OpenDSS network for simulation.
A no-load simulation was run to verify the operation of the simulator in
response to such dynamical inputs. All additional energy sources were also
set to 0 to capture only the system response to the generator. The test was
run for a duration of 2.5 seconds, with a step command issued at t = 1.1 s to
increase output power of the generator from 460 to 700 kW. The simulated
generator output and tie-lie exchange are plotted in Figures 3.14 and 3.15.

21
 Figure 3.14: Generator dynamic test

Figure 3.15: Dynamic tie-line exchange

The simulation is able to take this dynamic input and show that its effects
are translated through the network. By implementing a dynamic model

22
in Matlab, some dynamics are able to be introduced into the OpenDSS
simulation.

23
 CHAPTER 4

AN APPLICATION IN DISTRIBUTED
CONTROL STUDIES

As noted before, the goal of this model is to be able to insert it into other
simulations on an as-needed basis, rather than creating a bespoke model
for every single research application. This chapter presents the use of an
OpenDSS microgrid model in co-simulation with the C-HIL testbed devel-
oped at UIUC [11]. While the model used in this application is not the same
as the one used in previous chapters due to the milestone requirements of
the project, the developed simulation interface and procedure are exactly the
same; the previously described microgrid could be used to seamlessly replace
the one used here. The microgrid model used in this application was de-
veloped in cooperation with researchers at the University of California, San
Diego [11].

4.1 Co-simulation Goals

The goal of the project was to test the real-world viability of distributed
control techniques developed by the ARPA-E NODES group at UIUC. The
purpose of this control is to allow an ISO to directly control the power output
of a microgrid to provide ancillary services to the grid at large. The chief
requirements for this milestone are that the system meet a maximum initial
response time of 5 seconds and that the reserve magnitude variability of
the output be less than ±5%. In this particular case, the goal is frequency
regulation. As mentioned previously, this is accomplished in a distributed
manner, with controllers located at each controllable node communicating
together to calculate the desired power output of each node necessary to
reach the request of the Independent System Operator (ISO) [11].

24
4.2 Microgrid Model
The milestone requires the use of the microgrid in use on-campus at UCSD.
This model is a radial, balanced network with 3867 nodes and 1289 buses.
Generation is handled by a combination of photovoltaic, gas, and steam
turbines. Loads are composed of buildings and electric vehicle charging sta-
tions. Because the present capability of the C-HIL testbed is 100 controllable
nodes, this model had to be reduced using the distribution feeder reduction
algorithm described in [22].
The reduced model used in this simulation contains 13 generators (10 pho-
tovoltaic) and 107 loads for a total of 120 nodes.

4.3 Simulation
Because the UIUC C-HIL testbed is only capable of simulating 100 control-
lable nodes at the present time, 20 of the microgrid nodes were held constant.
The nodes chosen were 20 load nodes. For the duration of the simulation,
these nodes consumed a constant amount of complex power.
An additional consideration had to be taken into account, as loads are
typically not controllable and there are not 100 generators in the system.
After taking the 13 generators into account, the remaining 87 controllable
nodes had to be loads. To accomplish this in simulation, these nodes were
treated as generators creating “negative” power.

4.3.1 C-HIL Testbed Interface

The chosen environment to interface the testbed and the OpenDSS simula-
tion was Matlab. This was chosen due to its robust mathematical capabil-
ities, ease of communication, and extensive OpenDSS driver documentation.
The simulation procedure is outlined as follows:

1. ISO sends signal to distributed controllers containing desired power

output of microgrid

2. Controllers communicate and arrive at necessary node setpoints via

ratio consensus algorithm [11]

25
 3. Controllers update generator outputs with new setpoints

4. Generators adjust power output in response to updated setpoints

5. Microgrid power flow is simulated (OpenDSS), tie-line power exchange

is calculated and compared to ISO’s desired value

6. Controllers hold setpoints until new ISO command is received

During simulation, the ISO sends its signal through Matlab via the Modbus
protocol to the controllers. After reaching consensus on the desired setpoints,
the controllers then independently update their generators. In a real-world
scenario, these controllers would be attached to their generators. In simu-
lation, they each independently (to preserve the integrity of the distributed
framework) communicate values to a separate Matlab instance via Modbus
again. This Matlab instance is the driver for the OpenDSS simulation.
After updating each node with its new output value, the OpenDSS simula-
tion is then run, recording the tie-line power exchange. This architecture is
illustrated in Figure 4.1, developed by the NODES group [11].

4.3.2 First Experimental Scenario

Three simulation scenarios were performed to test the efficacy and accuracy
of the simulation. They were all variations in the input ISO signal.
The first input signal tested was a “stepped ramp” signal, simply to test
the microgrid response. This consisted of stepping the regulation signal from
0.2 to 0.7 over a period of approximately 2 minutes. This regulation signal
corresponds to a command requesting an increase in output 0.2 to 0.7 times
the difference between the initial power output and the total capacity. The
results of this first experiment are shown in Figure 4.2.

26
Figure 4.1: System architecture for C-HIL co-simulation

27
 Figure 4.2: Simulation results for “stepped ramp” input signal

The results show the microgrid simulation output generally following the
input signal. There is a delay between the signal send time and the response,
which corresponds to the communication delay between the components of
the system, as well as the time required by the distributed controllers to
come to consensus (between 4.2 to 4.7 seconds, depending on the connection
graph of the controllers). This, however, meets the requirement of a 5-second
initial response time. The simulation shows that the power generated by the
microgrid is not exactly equal to the requested value. This error is introduced
by two sources in the model. The first is the transmission loss combined with
the constant loads present in the model. These loads are consuming some
set amount of power for the duration of the simulation, and are not taken
into account by the control scheme. The second is the transmission loss. The
power measured at the tie line is equal to the generated power (commanded
by the ISO), less the line losses from all nodes to the interconnect. These
errors could both be corrected in future simulations by factoring them into
how the updated power setpoints are calculated.

28
4.3.3 Second Experimental Scenario
The second input signal tested was the first 5 minutes of the PJM RegD wave.
The PJM interconnect is a regional transmission organization (RTO) that
operates in 13 states in the eastern United States. As part of their service,
they provide a 40-minute regulation self-test signal to aid in development
of technologies to provide ancillary support to the grid. The results of this
simulation of the first 5 minutes of the signal are shown in Figure 4.3.

Figure 4.3: Simulation results for first 5 minutes of PJM RegD signal

The goal of this scenario was to test the model’s response to a steeply
decreasing signal. Again, the results show that the controllers are able to
accurately control the tie-line power injection from the microgrid, with the
same error sources appearing as in the first test, and with some calcula-
tion/communication delay.

4.3.4 Third Experimental Scenario

The third scenario tested was the second 5 minutes of the same PJM RegD
wave (Figure 4.4). This simulation was conducted to test the system’s re-
sponse to a steeper increasing signal. Again, the same errors and delays are
present in the results, with the possibility of being compensated for in future

29
trials. The system is able to respond adequately to the input commands,
providing the requested power within the required ±5% tolerance for reserve
magnitude variability, even when not compensating for system losses and
unmodeled loads.

Figure 4.4: Simulation results for second 5 minutes of PJM RegD signal

To ensure that the requirements of the test are being met, the signal is
held constant at the end of the test, showing that the microgrid is able to
hold the ±5% magnitude variability.
These simulations confirm that the distributed control is being performed
to meet the time and accuracy requirements, and able to accurately commu-
nicate with the OpenDSS model of the real-world network. The simulation
results of the model show that the system is responding as requested, and
that, even without losses taken into account, the required metrics are still
being met.

30
 CHAPTER 5

CONCLUSIONS AND FUTURE WORK

This thesis describes a model in the OpenDSS framework of a generic mi-

crogrid based upon Ameren’s TAC microgrid located in Champaign, Illinois.
The creation of this model is motivated by the need to evaluate the perfor-
mance of distributed control algorithms, currently under development by the
NODES group in the Power and Energy Systems area at UIUC. Because the
full dynamic simulation of such a microgrid would be computationally inten-
sive, a method for simulating microgrid response to control inputs without
the full dynamic performance is required. This OpenDSS model allows for
the simulation of the microgrid with only the dynamics that the designer
deems appropriate incorporated, reducing computation time and speeding
up the simulation process.
The microgrid consists of four energy resources (solar array, wind turbine,
natural gas generator, and battery storage), 201 loads, including 200 realisti-
cally modeled household loads distributed equally among the three electrical
phases, and a single load representing the control center for the microgrid.
These sources and loads are each connected through a series of transformers
to the microgrid substation, where another transformer links the microgrid
to the bulk grid through a single tie-line. The key conclusions and products
of this thesis are as follows:

• A dynamics-free model of a microgrid based upon Ameren’s TAC mi-

crogrid is available for use, with the option of adding dynamics on a
node-by-node basis in the driver program.

• While specifically designed with the Typhoon C-HIL laboratory in

mind, the model may be co-simulated with any necessary application,
through interfaces written in C, Python, and Matlab, among others.

• Co-simulations between OpenDSS and the C-HIL distributed control

architecture were successful, demonstrating the efficacy of both the

31
 model and the controller architecture.

This model may be useful in many research applications, and future work
may include the following:

• Further implementation of dynamics into the driver code. Presently,

only generator dynamics are included, and only on an as-needed basis.
However, due to the nature of the model, further dynamic models may
be incorporated, particularly in transformer behavior. Simulation could
be further improved by adding control schemes around the applied
dynamical models.

• More realistic battery behavior, including optimal charge/discharge pe-

riods and different deployment strategies.

• Simulations of this microgrid in islanded mode. Though this model

is primarily built for co-simulation, with grid service-providing ends,
this model could be used independently, with particular interest in its
capability of simulating the microgrid in islanded mode, something not
explored in this thesis.

• Development of a driver program using Python in keeping with the

open-source nature of the simulation.

32
 REFERENCES

[1] M. Hagge and O. Doyle, “Ameren Illinois TAC microgrid.” Presented

at PECI 2018 Conference, Champaign, Illinois, February 23, 2018.
[2] D. Montenegro and R. C. Dugan, Reference Guide, The Open Distribu-
tion System Simulator (OpenDSS). EPRI, March 2018.
[3] A. Arguello, W. L. Cunha, T. R. Ricciardi, R. Torquato, and W. Fre-
itas, “Dynamic modeling in OpenDSS: An implementation sequence for
object Pascal,” 2018 IEEE Power & Energy Society General Meeting
(PESGM), 2018.
[4] A. Hariri, A. Newaz, and M. O. Faruque, “Open-source Python-
OpenDSS interface for hybrid simulation of PV impact studies,” IET
Generation, Transmission & Distribution, vol. 11, no. 12, p. 3125–3133,
2017.
[5] J. A. Ahamioje and H. Krishnaswami, “Interconnection study of dis-
tributed PV systems by interfacing Matlab with OpenDSS and GIS,”
2017 IEEE 44th Photovoltaic Specialist Conference (PVSC), 2017.
[6] D. Montenegro, M. Hernandez, and G. A. Ramos, “Real time OpenDSS
framework for distribution systems simulation and analysis,” 2012 Sixth
IEEE/PES Transmission and Distribution: Latin America Conference
and Exposition (T&D-LA), 2012.
[7] L. C. Ribeiro, J. P. O. S. Minami, B. D. Bonatto, P. F. Ribeiro, and A. C.
Z. D. Souza, “Voltage control simulations in distribution systems with
high penetration of PVs using the OpenDSS,” 2018 Simposio Brasileiro
de Sistemas Eletricos (SBSE), 2018.
[8] W. Guo, J. Lin, Y. Song, and X. Chen, “A simulation platform of active
distribution network based on co-simulation of GridLAB-D and Mat-
lab,” 2014 International Conference on Power System Technology, 2014.
[9] D. P. Chassin, K. Schneider, and C. Gerkensmeyer, “GridLAB-D: An
open-source power systems modeling and simulation environment,” 2008
IEEE/PES Transmission and Distribution Conference and Exposition,
2008.

33
[10] J. Ma, F. Yang, Z. Li, and S. J. Qin, “A renewable energy integration
application in a microgrid based on model predictive control,” 2012
IEEE Power and Energy Society General Meeting, 2012.

[11] C. Sain, S. Nigam, M. Muralidharan, A. D. Domı́nguez-Garcı́a, P. W.

Sauer, and J. Kleissl, “Simulation of microgrid response with distributed
controls and HIL tests for USCD microgrid,” Tech. Rep., 2018.

[12] A. E. Fitzgerald, C. Kingsley, and S. D. Umans, Electric Machinery,

7th ed. McGraw-Hill, 2014.

[13] D. Montenegro and R. C. Dugan, OpenDSS PVSystem Element Model.

EPRI, March 2018, p. 1–10.

[14] “Inverter performance test summaries.” [Online]. Available: https:

//www.gosolarcalifornia.ca.gov/equipment/inverter tests/summaries/

[15] Yingli Solar, “YGE 60 cell HSF smart datasheet.” [Online]. Avail-
able: http://www.yinglisolar.com/static/assets/uploads/products/
downloads/DS YGE60CELLSERIES2HSFSMART-29b 35mm EU EN
20180615 V04.pdf

[16] “Solar resource data and tools.” [Online]. Available: https://www.nrel.

gov/grid/solar-resource/renewable-resource-data.html

[17] “Wind generation and total load in the BPA balancing authority.” [On-
line]. Available: https://transmission.bpa.gov/Business/Operations/
Wind/

[18] National Centers for Environmental Information, “Climate data online:

Dataset discovery.” [Online]. Available: https://www.ncdc.noaa.gov/
cdo-web/datasets

[19] The Royal Academy of Engineering, “Wind turbine power calculations.”

[Online]. Available: https://www.raeng.org.uk/publications/other/
23-wind-turbine

[20] J. V. Paatero and P. D. Lund, “A model for generating household elec-

tricity load profiles,” International Journal of Energy Research, vol. 30,
no. 5, p. 273–290, 2006.

[21] S. Kalogirou, McEvoy’s Handbook of Photovoltaics: Fundamentals and

Applications, 3rd ed. Academic Press, an imprint of Elsevier, 2018.

[22] Z. K. Pecenak, V. R. Disfani, M. J. Reno, and J. Kleissl, “Multiphase

distribution feeder reduction,” IEEE Transactions on Power Systems,
vol. 33, no. 2, p. 1320–1328, 2018.

34