Photovoltaic Generator Modeling for Large Scale Distribution System Studies

A Thesis

Submitted to the Faculty

of

Drexel University

By

Andrew S. Golder

in partial fulfillment of the

requirements for the degree

of

Master of Science in Electrical Engineering

October 2006
 ii

© Copyright 2006
Andrew S. Golder. All Rights Reserved.
 iii

ACKNOWLEDGEMENTS

I would like to thank my advisor Dr. Karen Miu, for all of her help and guidance

in the research and development of this thesis. I also thank her for the opportunity to

study and perform research at the Center for Electric Power Engineering (CEPE). My

education in the field of power systems engineering began with Dr. Miu as an

undergraduate and has continued through my graduate education. I appreciate her skill

and enthusiasm as a teacher and mentor; my education here at Drexel University has

encompassed much more than the field of power engineering.

I would like to thank Dr. Fischl and Dr. Nwankpa for serving on my thesis

committee. Their comments and suggestions throughout the development of this work

have been valuable and are greatly appreciated. I appreciate their time and commitment

to myself and this work.

 iv

TABLE OF CONTENTS

LIST OF TABLES………………………………………………………………..…….VII

LIST OF FIGURES…………………………………………………………..………. VIII

ABSTRACT……………………………………………………………………….….... IX

1. INTRODUCTION………………………………………………………………..…… 1

1.1. MOTIVATION…………………………………………………………….... 1

1.2. BACKGROUND…………………………………………...……………..… 2

1.3. PROBLEM STATEMENT………………………………………...……...… 4

1.4. OVERVIEW…………………………………………………………....…… 5

1.5. CONTRIBUTIONS………………………………………………..………... 7

1.6. ORGANIZATION OF THESIS…………………………………….………. 7

2. BACKGROUND……………………………………………………………………… 8

2.1. THE PHOTOVOLTAIC GENERATOR……………………………………. 8

2.1.1. THE PHOTOVOLTAIC CELL…………………….……………... 9

2.1.2. THE PHOTOVOLTAIC MODULE……………………...……… 11

2.1.3. THE PHOTOVOLTAIC ARRAY………………………...……... 14

2.1.4. THE POWER CONDITIONING UNIT…………………………. 14

2.1.4.1. THE MAXIMUM POWER POINT TRACKER……..... 15

2.1.4.2. THE DC TO AC INVERTER………………………..… 16

2.2. UNBALANCED DISTRIBUTION POWER FLOW……………………… 18

2.2.1. CONSTANT PQ SOURCES…………………………………...... 20

2.2.2. CONSTANT P|V| SOURCES………………………...…………. 21

 v

3. PHOTOVOLTAIC ARRAY MODEL…………………………………………….… 24

3.1. CHARACTERISTIC EQUATION OF THE NON-IDEAL

PHOTOVOLTAIC CELL…………………………………………………...….. 26

3.2. CONVERSION OF CELL CHARACTERISTIC EQUATION TO ARRAY

CHARACTERISTIC EQUATION …………………………………………...... 28

3.3. STEP BY STEP MODELING PROCEDURE…………………...……...… 31

3.3.1. DETERMINE THE CONNECTION SCHEME OF THE

ARRAY…………………………………………………………..…….. 31

3.3.2. CONVERT CELL CHARACTERISTIC EQUATION TO ARRAY

CHARACTERISTIC EQUATION…………………………………… 33

3.3.3. SOLVE FOR PARAMETERS OF CHARACTERISTIC ARRAY

EQUATION…………………………………………………………..… 34

3.3.4. CALCULATE THE I-V CURVE OF THE CHARACTERISTIC

EQUATION AND DETERMINE PGDC ………………………..…….… 39

3.4. FLOW DIAGRAM OF PROCEDURE FOR CALCULATING POWER

OUTPUT OF PHV ARRAY…………………………………………………..... 42

3.5. IMPLEMENTATION AND EXAMPLE…………………..…………….... 43

4. MODEL OF THE POWER CONDITIONING UNIT AND INTEGRATION WITH

POWER FLOW………………………………………………………………………… 45

4.1. POWER CONDITIONING UNIT MODELING……………...…………... 46

4.1.1. THE CONSTANT PQ SOURCE MODEL…………………....… 46

4.1.2. THE CONSTANT P|V| SOURCE MODEL……………...…….... 47

4.2. POWER FLOW INTEGRATION………………………...……………..….48

 vi

4.2.1. PQ SOURCE MODEL INTEGRATION…………………..……. 49

4.2.2. P|V| SOURCE MODEL INTEGRATION………………..……… 49

5. SIMULATION METHODS AND RESULTS………………………………………. 51

5.1. TEST SYSTEM AND CASE FILE CONSTRUCTION…………………... 51

5.1.1. ELECTRICAL SYSTEM DATA…………………………...….... 52

5.1.2. ENVIRONMENTAL DATA………………………..………….... 53

5.2. SINGLE FLOW STUDIES……………………………………….………... 54

5.3. TWENTY FOUR HOUR SIMULATION STUDIES……………….…….. 55

5.3.1. EXAMPLE TWENTY FOUR HOUR SIMULATION STUDY.... 56

5.3.2. OBSERVATIONS AND REMARKS……………...……………. 59

6. CONCLUSION AND FUTURE WORK………….…………………………...….… 61

6.1 CONCLUSIONS…………………………………………………...………. 61

6.2. SUMMARY OF RESEARCH CONTRIBUTIONS……….…………….… 62

6.3. FUTURE WORK…………………………………………………………... 63

LIST OF REFERENCES……………………………………………………………..… 65

APPENDIX A: List of Symbols by Section..................................................................... 69

APPENDIX B: Derivation of VocAE……………….……………………………..…...… 73

APPENDIX C: CALCULATING THE LOAD LEVEL FOR A GIVEN POWER FLOW

…………………………………………………………………………………………....75

APPENDIX D: EXAMPLE SOLAR MODULE SPEC SHEET …………………….....78

 vii

LIST OF TABLES

Table 5.1: Summary of simulation study conditions……………………………………51

Table 5.2: Imbalance study varying PHV penetration…………………………………...55

 viii

LIST OF FIGURES

Figure 2.1: Typical connection scheme of a residential or commercial PHV generator….8

Figure 2.2: Circuit model of an ideal solar cell………………………………………….10

Figure 2.3: String of cells with bypass and blocking diodes…………………………….12

Figure 2.4: Components of the power conditioning unit………………………………...15

Figure 2.5: I-V Characteristic of a solar module under varied solar irradiance…………16

Figure 2.6: I-V Characteristic of a solar module under varied ambient temperature……16

Figure 2.7: The basic building block of a distribution system…………………………...19

Figure 3.1: I-V curve showing the three points given on a typical specification sheet…25

Figure 3.2: Electrical model of a solar cell with shunt resistance……………………….26

Figure 3.3: Model of a solar cell without shunt resistance………………………………28

Figure 3.4: Circuit diagram of a 3x3 solar array…………………………………………29

Figure 3.5: Flow diagram of PHV array model………………………………………….42

Figure 3.6: I-V Characteristic of example PHV Array…………………………………..43

Figure 3.7: P-V Characteristic of example PHV Array………………………………….44

Figure 5.1: One line diagram of 394 bus test system...…………………………………..52

Figure 5.2: Real power demand of the 394 bus test case for peak load….………………57

Figure 5.3: TMY2 Irradiance profile for Binghamton NY, July 16……………………..58

Figure 5.4: TMY2 Ambient temperature profile for Binghamton NY, July 16................58

Figure 5.5: Real power demand at the substation under varying capacity of PHV

generation………………………………………………………………………………...59

Figure 5.7: Real power output of a 1000W rated PHV generator……………………….60

ix
 x

ABSTRACT
Photovoltaic Generator Modeling for Large Scale Distribution System Studies
Andrew S. Golder
Dr. Karen Miu, Ph.D.

Geographic regions with favorable conditions for photovoltaic (PHV) power

generation are seeing increasing numbers of three-phase commercial installations and

single-phase residential sized installations. PHV sources can be described as intermittent

sources because their power production is a function of ever changing environmental

conditions. Increased numbers of these intermittent sources will begin to have a

significant impact on the distribution system. This work presents the development of a

simulation model for utility distribution systems with installed PHV generation. The

mathematical modeling of PHV generators is investigated and a model is developed to

calculate power output based on the environmental conditions which determine generator

behavior. This output is then integrated into an unbalanced distribution power flow

solver using either a constant PQ or constant P|V| generator model. The model has been

implemented and simulations will be presented for a 394 bus test system. In order to

accommodate intermittency, hourly simulations which utilize corresponding

environmental conditions and 24 hour load curves have been automated. Simulations are

performed in order to perform system studies in the areas of demand profiles, phase

balancing, feeder balancing, power factor changes and voltage rise. Analysis is

performed to identify issues that will be most relevant to engineers working in planning

and operations of distribution systems with installed PHV generation.

xi
xii
 1

1. Introduction

1.1 Motivation

Photovoltaic (PHV) power generation has an increasing importance in the

renewable energy marketplace due to its ability to quietly produce emission free power at

the load site. The main advantage of these systems over other forms of renewable energy

is the feasibility of small PHV systems which can be owned and operated by the

consumer. Economic and political trends, along with increased social desire for

alternative energy sources has produced a significant increase in the installation of small

scale (on the order of 1kW) residential and commercial PHV power systems on the utility

distribution system. In the US, as of 2003, 15 states have implemented programs which

encourage or mandate the growth of renewable energy [9]. In addition some utilities

such as Austin Energy of Texas, USA have offered their customers rebates up to $4.00

per watt to help offset the installation cost of PHV systems [2]. In 2006 the California

state government passed the California Solar Initiative (CSI) which is the largest solar

energy policy enacted in the US. The initiative allots $2.9 billion dollars for solar energy

rebates which will aid residents and businesses in reducing the installation costs of PHV

generation systems. The goal of the CSI is to increase installed solar capacity in

California by roughly 3000 MW over an 11 year span from 2006-2017 [4].

While PHV generation is not a new technology, the socio-economic factors

described above are currently causing a significant and rapid increase in the total installed

capacity of grid connected PHV generation in the United States. Because PHV

generation is most often owned and operated by the consumer, system planners and

operators have reduced knowledge of and control over the distribution systems yet it is
 2

their responsibility to manage these networks. This situation is similar to system loads

where the day to day behavior of which is determined by the customer. However in the

case of loads the utility has more control over where new loads are connected to the

existing network, where as the location and sizing of PHV generators is determined by

the customer. As installed capacity increases, planners and operators will be faced with

an entirely new set of problems and challenges which will require new engineering

solutions.

When developing engineering solutions a basic tool used to predict system

behavior, identify problems which are most likely to occur, and test the optimal control

schemes is power flow analysis. Power flow analysis requires mathematical models of

every component connected to the power system being studied. Thus it is important to

develop models which allow for distribution system analysis with PHV generation to be

performed.

1.2 Background

PHV generators are most often connected to the distribution system for several

reasons. A high percentage of grid connected PHV generators are owned by small,

residential and commercial consumers, and are therefore connected at the load site on the

distribution system. When compared to other forms of distributed generation, PHV

generators are typically physically small, low voltage, and low power systems and it is

therefore appropriate to interface these systems with the low voltage distribution system.

PHV generation is unique in that the residential and small commercial systems for

example in the U.S. which make up a majority of the installations are often configured

for single phase power generation. It is therefore appropriate to connect PHV generators
 3

to the distribution grid which contains single phase feeders and is designed to handle an

unbalanced flow of power.

The study of renewable generation in utility distribution systems began in the late

1970’s. Procedures for evaluating the impact of intermittent generation on distribution

system planning and operations were presented in [17, 18]. Subsequent studies have

compared power demand curves with PHV production curves in order to predict system

load profiles [12]. Other studies have calculated line loss savings from a distribution

PHV system installation based on simplified distribution line models [13]. To this date

an unbalanced distribution power flow solver incorporating PHV generation has not been

developed and documented publicly. While PHV generation systems have been in use

since the late 1970’s economic considerations have limited their contribution to overall

power production, and the distribution power system impacts of these small number of

installations has not seen extensive study. In addition unbalanced distribution system

modeling and power flow analysis, which increase the accuracy of distribution system

analysis were developed in the late 80’s and 90’s and have only recently evolved as state

of the art methods [35, 36]. Recently in the early 21st century economic and political

trends suggest that certain geographic areas could see high penetrations of PHV

generation on the distribution system, and therefore studies of their impact on large scale

distribution systems are warranted.

The characteristics of utility distribution networks provide engineers with a

unique set of problems. For example in order to counteract the inherently unbalanced

nature of distribution systems load balancing schemes have been investigated and have

been implemented typically in seasonal manners [3, 1] and more recently with higher
 4

frequency [8]. Methods to enhance the voltage stability and reduce power losses of

distribution systems by optimal placement and control switching of capacitors have been

investigated in [23, 32]. Schemes for service restoration of distribution systems have

been investigated in [22, 24, 25] utilizing techniques including network configuration and

capacitor control. Investigations in the field of distribution automation are extensive with

goals such as cost reduction [34], reliability improvement [28], and power quality

monitoring [19].

Now, with a significant number of small scale, single-phase PHV systems coming

online, it may be useful to make updates to these works in order to include the special

case of distribution systems with significant capacity of installed PHV generation. In

addition, the intermittent nature of PHV generation creates new challenges with regard to

distribution capacity planning and maintaining load balance. New distribution system

analysis and automation tools are needed to accommodate the problems associated with

grid connected PHV generation.

1.3 Problem Statement

The objective of this work is to develop a component model for grid connected

PHV generation which can be integrated with an unbalanced distribution power flow

solver. The component model must be practical for use by system planners and operators

in order to serve its function for systems level studies. Then, a combined model, which

describes the PHV generator and system models together, will use state of the art

methods and equations for PHV generator modeling and unbalanced distribution power

flow. The model will be implemented using an existing unbalanced distribution power

flow solver coded in MATLAB in order perform simulation studies.

 5

1.4 Overview

This work presents the development of a system model which simulates

distribution systems with grid connected photovoltaic generation. The model is formed

by the integration of i) a mathematical component model of a photovoltaic (PHV)

generation system, with ii) an existing unbalanced distribution power flow solver. A

simulation tool has been developed using MATLAB which is used to simulate the

behavior of distribution systems with grid connected PHV generation over the course of a

day.

The mathematical model of the photovoltaic generator is composed of two parts

which model the two main components of the PHV generator, the photovoltaic array, and

the power conditioning unit (PCU). First the model of the PHV array calculates the DC

power output under the given environmental conditions. Second the model of the PCU

converts the DC power value to a specific AC power injection to the utility grid. These

two models are now discussed respectively.

The basic building block of the PHV array model is the standard equation which

describes the I-V characteristic of a single solar cell. By making some assumptions it is

possible to perform scaling operations in order to define an equation which describes the

I-V characteristic of a PHV array. The developed model uses the available inputs to

calculate the required parameters of the array characteristic equation. Because the

solution to the characteristic equation is not unique an appropriate initial condition and

iterative technique are used to calculate the solution vector of the array characteristic

equation. The model then determines the solution corresponding to the maximum power

output, and defines this as the DC power output of the PHV array.
 6

The model of the PCU is characterized by the choice of one of two inverter

control schemes, which determines the use of either a constant PQ injection or a constant

P|V| injection model. In both schemes the DC power output of the PHV array is first

scaled down according to the efficiency of the PCU, and second an AC power injection to

the distribution system is calculated.

A previously developed unbalanced distribution power flow method provides the

steady state model of the utility distribution system. A traditional Newton Raphson

method is used to perform three phase power flow analysis using the appropriate

component models for unbalanced distribution power flow. Power injections determined

by the PHV array model are integrated with the power flow using the traditional PQ and

P|V| generator models. The overall model of the power system is used to perform power

flow analysis in order to model the steady state behavior of the distribution system.

Therefore the model is an instantaneous model which describes the steady state behavior

of the power system at a single point in time.

The model of a distribution system with PHV generation is implemented in

MATLAB and simulations are performed using a 394 bus test system. Twenty four

simulations, each of which represents the average behavior of the system for the time

period of one hour, are used to study the system conditions over the course of the day.

Profiles for load demand and environmental conditions are used to simulate the test

system for varying seasonal, environmental, and loading conditions. Analysis is

performed with a primary focus placed on phase balancing and feeder balancing. Several

issues are addressed with the goal of identifying the problems which are most relevant to

distribution system planners and operators.

 7

The following sections describe the characteristics of PHV generators which

make it appropriate for them to be connected to the utility distribution system. It then

details the background and history of distribution systems studies involving PHV

generation. Literature regarding distribution systems which may be impacted by the

installation of grid connected PHV generation is also discussed.

1.5 Contributions:

The following contributions are presented in this thesis:

• Practical PHV generator model for systems engineers requiring limited input data

• Matlab based simulation tool which integrates the PHV model with an unbalanced

distribution power flow solver for system studies

• Simulation methods, results, and analysis for 24 hour system studies using a 394

bus test case.

This thesis provides an additional component model for the field of steady state

distribution power system analysis.

1.6 Organization of Thesis

Chapter 2 is a review of i) PHV generators, their components and functions, and

ii) unbalanced distribution power flow and PQ and P|V| generator models. Chapter 3

describes the PHV array model, specifying all input and output parameters, assumptions

and procedures. In Chapter 4 the PCU model and integration of the calculated power

injection with unbalanced distribution power flow analysis is described. Chapter 5

presents simulation methods, procedures, results, and analysis. In Chapter 6 some

conclusions are made.

 8

2. Background

This chapter reviews two subjects important to modeling installed PHV

generation on a distribution system:

• First, a grid tied photovoltaic generator, its components and their functions.

• Second, unbalanced distribution power flow methods with a focus on co-

generator modeling.

2.1. The Photovoltaic Generator

A photovoltaic generator is a device which converts energy in the form of solar

radiation to energy in the form of electric current. The basic component of a PHV system

is the photovoltaic cell. Multiple cells are connected in series and parallel to form solar

panels or modules, which are sold commercially for applications such as grid tied solar

generation. Solar modules are connected in series and parallel in order to create a PHV

array. The solar array produces DC power which for grid tie systems must be adjusted by

a maximum power point tracker (MPPT) and then converted to AC by a DC to AC

inverter. These functions are performed by the power conditioning unit (PCU) which is

composed of the MPPT and inverter [29]. An example a PHV generator system without

energy storage devices is shown in Figure 2.1. In the discussion below the PHV array

and its components will be addressed first followed by the PCU and its components.

Figure 2.1: Typical connection scheme of a residential or commercial PHV generator

 9

2.1.1. The Photovoltaic Cell

The photovoltaic cell is a semiconductor device which behaves as a current source

when driven by a flux of solar radiation from the sun. This occurs when radiation is

incident upon absorbing material and separates positive and negative charge carriers in

the presence of an electric field. The electric field exists permanently at junctions or in-

homogeneities in solar cells which can be described as silicon semiconductor junction

devices. A silicon semiconductor junction device contains a p-n junction similar to that

of a common diode, however in a solar cell it exists over a large surface area. When not

illuminated and connected to a forward bias the ideal solar cell mimics the electrical

characteristics of an ideal diode, modeled by the Shockley equation (2.1), where the

current produced is referred to as the dark current ID:

⎛ kTVc q ⎞
I D = I o ⎜ e cK − 1⎟ (2.1)
⎜ ⎟
⎝ ⎠

where:
I D = dark current (A)
I o = saturation current of the diode (A)
Vc = cell voltage (V)
q = 1.6*10−19 (coul )
k = 1.38*10−23 ( j / K )
TcK = Cell Temperature (K)

The current in the cell that results from solar radiation is called the photo current

Il which flows in the direction opposite of the forward dark current. Its value remains the

same regardless of external voltage and therefore it can be measured by the short circuit

current (Isc = Il). This current varies linearly with the intensity of solar radiation as
 10

increased radiation is able to separate increased charge carriers. The overall current is

then described as the difference between the dark current and the photocurrent. If the

sign convention of current flow is reversed to describe the current which is produced by

an illuminated cell (Ic) then the cell equation can be written as shown in (2.2):

⎛ kTVc q ⎞
I c = I l − I o ⎜ e cK − 1⎟ (2.2)
⎜ ⎟
⎝ ⎠

where:
I c = cell current (A)
I l = photo current (A)
G = solar irradiance (W/m 2 )
TA = ambient temperature (°C)
I l = I l (G )
Vc = Vc (G, TA)

This is the mathematical equation which models the behavior of an ideal

photovoltaic cell shown in Figure 2.2 [29]. The production of photocurrent is modeled

with a DC current source and the dark current is modeled with a diode referred to as the

diffusion diode.

Figure 2.2: Circuit model of an ideal solar cell

Modeling of non-ideal solar cells and the behavior of solar cells under varying

environmental conditions will be discussed further in Section 3.2.

 11

2.1.2. The Photovoltaic Module

A photovoltaic module is composed of a series and parallel connection of solar

cells, with the additional components of blocking and bypass diodes. While the

manufacture and size of solar cells vary, in general a single cell has a relatively low

voltage handling capability on the order of 0.6 V [10]. In order to package solar cells as a

more practical device most manufacturers produce solar modules; a group of solar cells

connected in series and parallel in order to increase the voltage and current handling

capability. While a variety of connection schemes exist for a multitude of applications, a

common scheme for PHV modules used in grid tied power generation is a connection of

72 cells in series. As an example, the Shell SQ150-PC module which is so composed

provides rated voltage and current of 34 V, 4.4 A [10].

In most PHV design and modeling applications the following assumption is made,

which is also used in this work:

A1: Connection of solar cells in series will directly multiply the voltage handling

capability of the system and connections in parallel will directly multiply current

production.

For this assumption to be reasonable it is necessary for all solar cells to behave in a

uniform manner (further discussion in Section 3.2.2.), however it is clear that this is not

always the case. Solar modules contain blocking and isolation or bypass diodes as shown

in Figure 2.3 to reduce the effects of cells which do not act uniformly. Without these

devices it is possible for cells or strings which are not operating uniformly to consume

power generated by other cells or strings. These devices therefore reduce the power loss
 12

associated with non-uniformity and allow the behavior of the array to more closely

resemble the uniform case by reducing the negative impact of non uniform cells.

Figure 2.3: String of cells with bypass and blocking diodes

Thus for this work, the following additional assumptions are made:

A2: Each cell behaves in a uniform manner. Under a given environmental condition the

electrical characteristics of every cell in a solar array is identical.

A3: For the purposes of modeling a solar array in this work the bypass diodes and

isolation/blocking diodes are considered as ideal diodes.

With these assumptions in place, bypass diodes are reverse biased and act as open circuits

while isolation or blocking diodes are forward biased and act as short circuits. Therefore

it is not necessary to include these diodes in the mathematical model to be described in

Chapter 3.

Bypass diode:

Bypass diodes are placed across each cell in a solar module in order to ensure that

each series string of cells produces the maximum possible voltage. This protects against

the case where a given solar cell does not receive the same illumination as other cells in

the string. For example, given a string of cells, if a single cell is covered in dirt and thus

absorbs less radiation it is possible for the voltage gain of this cell to reverse its direction
 13

(ID > Il, as shown by the center cell in Figure 2.3) which would significantly reduce the

voltage gain of the entire string. Therefore, a bypass diode is connected in parallel to

each solar cell which becomes forward biased if the voltage gain of a cell is reversed, and

effectively creates a short across this cell. Then, when a cell voltage is reversed the

voltage capability of the string is increased by the action of the bypass diode [5].

Isolation/Blocking Diodes:

Isolation or blocking diodes are often placed between each string of modules and

the PCU in order to prevent module strings from absorbing power from the system when

not illuminated. Under zero illumination and in the presence of forward voltage

(provided by a battery, the grid, or other module strings) a string of solar modules will

draw current from the circuit [5]. Therefore an isolation or blocking diode is placed

between the module string and the voltage source as shown in Figure 2.3. The term

blocking diode often refers to a single diode used in PHV systems with battery storage,

while isolation diodes refer to multiple diodes placed on the positive terminal of each

series string of solar modules. These diodes perform the same function of blocking a

reverse current flow however multiple isolation diodes provide additional protection in

the case of shading on a single string of modules. These diodes become reversed biased

when the string is not illuminated thus preventing current flow to the string. In some

cases, the function of a blocking diode is performed by switching in the PCU, however

this only prevents the array from absorbing power from the grid, it does not prevent a

single module string from absorbing power from the rest of the array.
 14

2.1.3. The Photovoltaic Array

A photovoltaic array is composed of series and parallel connections of solar

modules. The number of modules connected in series and parallel is based on the voltage

requirements of the system and the desired power output of the array. Grid connected

systems require an inverter which in order to successfully interface with the grid requires

a specific DC input voltage range generally on the order of 200-600 Volts DC [33]. PHV

arrays are designed to produce a voltage close to the top of this range at rated power

production. This allows the inverter to operate for the largest possible range of DC

voltage input and therefore the largest range of environmental conditions.

Once the voltage requirement is met, power handling capability can be increased

by connecting additional module strings in parallel. This is done by first connecting the

correct number of modules in series to fulfill the voltage requirement creating a string of

modules. Next, additional strings of the same number of modules may be added to

increase the current production and thus the power capability of the array. It is important

that each string is composed of an identical number of modules of the same brand and

power rating. If strings are not matched the voltage production of each string may not be

identical. In this case the voltage output of the array will be lower than that of the highest

producing string, therefore the highest producing string does not operate at maximum

efficiency.

2.1.4. The Power Conditioning Unit

The power conditioning unit (PCU) is a device which interfaces the solar array

with the utility grid. In a grid tie system without storage the PCU consists of a maximum

power point tracker (MPPT) and a DC to AC inverter, as shown in Figure 2.4. In most
 15

cases the PCU also contains protection devices which disconnect the PHV system from

the grid in the case of grid failure. Many PCU’s use the existing controllable switches of

the DC to AC inverter in order to perform these protective functions. These components

are described in detail in the following sections.

Figure 2.4: Components of the power conditioning unit

2.1.4.1. The Maximum Power Point Tracker

The MPPT is a DC to DC voltage converter which ensures that the PHV array

provides the maximum available power to the inverter. Typical I-V characteristics are

shown in Figure 2.5 and Figure 2.6 where environmental conditions are varied. As can

be seen in the figures, a solar array is able to provide constant current for a range of

voltages below a given threshold, where the maximum power point occurs just beyond

this threshold. The MPPT device assures that the input to the inverter remains constant at

this optimal operating point. A variety of DC to DC converters such as buck, boost,

buck-boost, or Cúk can be used for this function [7].

 16

Voltage vs. Current

6

G = 1200 W/m2
5

G = 1000 W/m2
4

DC Amperes
G = 800 W/m2
3

0
0 5 10 15 20 25 30 35 40
DC Volts

Figure 2.5: I-V Characteristic of a solar module under varied solar irradiance

Voltage vs. Current

5

4.5

3.5

3 Ta=35°C
DC Amperes

2.5

2 Ta=25°C
1.5

1
Ta=15°C
0.5

0
0 5 10 15 20 25 30 35 40
DC Volts

Figure 2.6: I-V Characteristic of a solar module under varied ambient temperature

2.1.4.2. The DC to AC Inverter

The DC to AC voltage converter ties the DC output of the MPPT with the AC bus

of the utility grid. Most grid tie PCU’s use self-commutated voltage source inverters
 17

which are built with switching devices such as IGBT’s or MOSFET’s. These types of

switches have the advantage of allowing control of both the on and off switching events.

This characteristic is optimal for normal DC to AC inverter operations and in addition

allows for the existing switches of the inverter to be used for protection functions. Thus

the self-commutated inverter can provide control of the power factor and suppression of

harmonic current, as well as protection against grid disturbance. The self-commutated

inverter is capable of executing two separate control schemes, current control and voltage

control. The control scheme is selected based on the size and function of the PHV

generator.

The current control scheme is the most widely used scheme for grid tied PHV

generators. This is because of its ability to provide a very high power factor with simple

control circuitry [14]. This scheme senses the voltage waveform of the grid for a

reference value and matches the phase angle of the current output to the reference within

a certain tolerance. A high power factor results because the current phase is controlled to

match the phase of the grid. The current control scheme is ideal for residential and small

commercial PHV generators where control of the bus voltage magnitude is not necessary;

however, a high power factor and reduced cost of the inverter are priorities.

The voltage control scheme is used for most off grid applications and some grid

tied applications where it is desired to control the voltage magnitude of the AC bus. This

scheme taps the grid for a reference and matches the voltage output to the reference

within a certain tolerance. A PWM control scheme is utilized resulting in a waveform

with less lower-order harmonic components [14]. Control of the output voltage is

necessary for off grid applications where the PHV generator is the only source of power
 18

in the circuit. In this case, it is necessary for a certain voltage magnitude to be

maintained by the generator for proper operation of the loads and storage devices which

rely on the PHV generator for power. Large commercial or industrial grid tied PHV

generators may also use this scheme in order to control the voltage magnitude of the

output bus. The power output of a PHV system varies over a wide range in a single day,

and the power injected to the grid will affect the voltage at the output bus. Therefore,

large PHV generators could create undesirable swings in the voltage level at the output

bus including significant over voltage levels. In order to prevent a large PHV generator

from causing such disturbance problems in the grid a voltage control scheme can be

utilized in the inverter. With this scheme the voltage magnitude of the output bus is

maintained regardless of the power output which prevents the PHV generator from

causing voltage swings in the grid.

The type of control scheme selected for a given PHV generator will determine the

generator modeling scheme to be used when integrating with an unbalanced power flow

analysis. Utilizing the current control scheme will cause a PHV generator to be modeled

as a constant PQ power source while use of the voltage control scheme will result in a

constant P|V| power source model. This is discussed further in Chapter 4.

2.2. Unbalanced Distribution Power Flow

The goal of this work is to study the affect PHV generators will have on the utility

distribution grid. PHV generators are often single phase devices which produce power at

low voltage levels. These characteristics match well with distribution systems which

allow for single and two phase feeders and distribute power at mid to low voltage levels,

therefore a majority of PHV generation systems are connected to the distribution system.
 19

Thus unbalanced distribution power flow analysis is most appropriate for performing the

desired studies.

Distribution systems are unbalanced networks which can contain single phase or

two phase feeders and are often characterized by a radial or tree structure. Figure 2.7

shows the basic building block of a distribution power system which interconnects

several components. The unique characteristics of distribution systems require detailed,

unbalanced component models for elements such as lines, switches, and transformers to

be used in distribution power flow methods [35].

Figure 2.7: The basic building block of a distribution system [36]

Each component can have up to three phases therefore the variables in Figure 2.7 are

vectors with up to three elements, and the following notation is defined:

Vk −1 = steady state voltage at bus k-1

Vk = steady state voltage at bus k
I k = steady state current flow from bus k-1 to bus k
I k = steady state current flow from bus k to bus k-1
I (G ,C , L ) k = steady state current injection at bus k
S(G ,C , L ) k = steady state power injection at bus k
 20

Devices connected to a bus such as multi-phase loads, co-generators and

capacitors can be modeled as current or power injections to the network [36] as shown in

Figure 2.7. Two common methods for solving power flow include backward/forward

sweep techniques for radial structure systems and Newton Raphson for general structured

systems. Each implementation utilizes the multi-phase component modeling in [35] and

each method solves for the steady state voltages at the existing phases concurrently.

A brief discussion of select co-generator models for both the backward/forward

sweep and Newton Raphson methods is provided. The two main types of generator

models are constant PQ sources and constant P|V| sources, which will now be discussed

in detail respectively.

2.2.1. Constant PQ sources

Co-generators are commonly modeled as constant power devices. In general,

synchronous generators are controlled to maintain constant power output and constant

power factor. Some utilities require that a power factor controller must be able to

maintain the power factor within plus or minus one percent [6]. Induction co-generators

are not able to control their power factor as the reactive power will vary with the terminal

voltage. For induction generators the use of a constant PQ model is considered a

simplification based on the assumption of a constant bus voltage [6]. Therefore in this

model, power output of the device is determined completely by its own characteristics

and operations and is independent of the behavior of the utility grid to which it is

connected.
 21

Backward/Forward Sweep:

While the power output remains constant the current output of the device is

dependent on the bus voltage magnitude to which it is connected [36]. The current

injection due to a co-generator connected to bus k is calculated in (2.3):

*
⎛S ⎞
I Gk = ⎜ Gk ⎟ (2.3)
⎝ Vk ⎠

where:
SGk = power supplied by co-generator at bus k
I Gk = current injection due to co-generator at bus k

Newton Raphson:

In the traditional Newton Raphson method load busses are modeled as constant

power PQ busses. It is assumed that the power injection at a load bus is known, which is

key because the mismatch equation for Newton Raphson is defined as the difference

between the known values for power injection and the values for power injection

calculated from the variables at each iteration. Therefore when a PHV generator is

connected to a load bus the power generated is added to the known power consumption

before the power flow is performed [11].

2.2.2. Constant P|V| Sources

Some generators do incorporate control schemes which allow the generator to

regulate the terminal voltage to a specified value. In this case, the reactive power output

of the generator is determined by the behavior of the grid to which it is connected.

Backward/Forward Sweep:
 22

For this scheme to be modeled it is necessary to calculate the reactive power

output required to maintain the specified voltage at the given real power output. One

method [26] creates a fictitious node and branch which feeds reactive power to the

specified P|V| node thus maintaining the desired voltage value. Using this technique the

P|V| node can actually be treated as a PQ node in calculations where the Q injection is

calculated from [26]:

⎛ V fict , k − | Vkt |⎞
Qk =| Vkt | ⎜ ⎟⎟ (2.4)
⎜ X
⎝ fic , k ⎠

where:
Qk = specified value of reactive power injection at bus k
Vkt = calculated voltage value in bus k at iteration t
V fict , k = calculated voltage of fictitious node of bus k at iteration t
X fic ,k = fictitious branch impedance of bus k

In the case of generators which are connected to the grid through an inverter the

available injection current is limited. If the Q value determined in (2.4) violates the

current constraint then the inverter will not be able to maintain the specified voltage and

control is switched to current control which is modeled by the constant PQ source model.

The Q limit is determined by [26]:

Qk ,lim = ± SGk
2
− Pk 2 (2.5)

where:
SGk = power produced by the generator at bus k
Pk = specified real power output of the generator at bus k
 23

Newton Raphson:

In the traditional Newton Raphson method, busses connected to generators are

modeled as constant P|V| busses. The known value for the real power injection is

included in the mismatch equation and the unknown reactive power value is not. The bus

voltage magnitude is specified and thus is not included as a variable. A limit is placed on

the reactive power that may be produced by the generator and after each iteration the

reactive power output is calculated and compared to the limit value. If the constraint is

violated the power flow is considered invalid because no solution exists which satisfies

the both the reactive power limit and the voltage magnitude requirement [11].

Given the concepts reviewed in this chapter, development of a PHV array model

is presented in the following chapter. Integration of the PHV array model with an

unbalanced distribution power flow analysis is described in subsequent chapters.

 24

3. Photovoltaic Array Model

This chapter develops a model which calculates the DC power output PGDC , of a

photovoltaic array as a function of the array ratings and a set of environmental

conditions:

PGDC = PGDC (array ratings, environmental conditions) (3.1)

The model is intended for use by system planners and operators who will have a limited

amount of data available to them regarding the photovoltaic systems themselves, and the

environmental conditions which determine their output. Inputs corresponding to the solar

modules are limited to the information available on a typical specification sheet, such as

those shown in Appendix D. Inputs corresponding to the size and connection scheme of

an array are limited to a power rating. Inputs corresponding to the environmental

conditions are limited to solar irradiance which determines current production and

ambient temperature which most heavily impacts the voltage handling capability. A list

of the inputs to the model is shown below:

(1) Array power rating:

Pra = rated DC power output of the array (W)

(2) Solar module specifications:

VocM = module open circuit voltage (V)
IscM = module short circuit current (A)
(ImM, VmM) = module maximum power point (A, V)
PrM = module rated power output (W)
NOCT = nominal operating cell temperature (°C)
⎛ ∂I scM ⎞ ⎛ ∂VocM ⎞
⎜ ⎟, ⎜ ⎟ = temperature coefficients
⎝ ∂Tc ⎠ ⎝ ∂Tc ⎠
Ncs = number of cells in series in the module
 25

(3) Environmental conditions:

G = solar irradiance (W/m2)
TA = ambient temperature (°C)

Figure 3.1 shows an I-V curve of a solar module with the three points commonly

provided on specification sheets. In this chapter, the characteristic equation of the non-

ideal photovoltaic cell will be presented first, followed by the method used to scale the

equation to represent an entire array. A step by step summary of the modeling procedure

is then provided, which is followed by subsections containing detailed explanations. A

flow chart of the modeling procedure is then shown which specifies the inputs and output

of each step is shown. Finally the implementation of the PHV array model is described

and the results of an example calculation are presented.

Voltage vs. Current

5

4.5

4 (0, Isc)
3.5
(VM, IM)
3
DC Amperes

2.5

1.5

0.5
(Voc, 0)
0
0 5 10 15 20 25 30 35 40
DC Volts

Figure 3.1: I-V curve showing the three points given on a typical specification sheet.
 26

3.1. Characteristic Equation of the Non-Ideal Photovoltaic Cell

The circuit model of a photovoltaic cell often used for research purposes is

shown in Figure 3.2. This model was selected due to its use of specification sheet inputs,

and its appearance in several texts regarding photovoltaic devices [5, 20, 29].

Figure 3.2: Electrical model of a solar cell with shunt resistance

In addition to the components of the ideal model, the model shown incorporates a

series and shunt resistance to capture the behavior of a physical solar cell more

accurately. Several elements of the solar cell contain resistive properties such as the

semiconductor material itself, the metal grid which collects current from the

semiconductor material, the collector bus and the internal wiring. It is assumed that these

series losses can be modeled using a lumped resistor Rs. Several shunt resistive losses

occur such as localized shorts at the emitter layer of the semiconductor material and

perimeter shunts at the cell borders. The combined effect of these shunts is modeled by a

lumped parallel resistor Rsh. Equation 3.1 describes the I-V characteristic of this cell

model:

⎛ (VcnkT
+ I c Rs ) q
⎞ V +I R
I c = Il − Io ⎜ e cK
− 1⎟ − c c s (3.1)
⎜ ⎟ Rsh
⎝ ⎠

where:
 27

I c = cell current (A)

I sc = cell short circuit current (A)
I l = cell photo-current (A)
I o = saturation current of the diode (A)
Vc = cell voltage (V)
Rs = series resistance
Rsh = shunt resistance
q = 1.6*10−19 (coul )
k = 1.38*10−23 ( j / K )
TcK = Cell Temperature (K)
n = nonideality constant of the diode
Note: I l =I sc

It is shown in [5] that only very small values of Rsh will significantly affect the

power output of the model. Solar cells are designed to reduce shunt losses and therefore

most commercially available solar modules have relatively large shunt resistances. Thus

this work will make the following assumption:

A4: the shunt resistance can be neglected:

Rsh = ∞ (3.2)

This results in the equivalent circuit model shown in Figure 3.2.

Also, it is not possible to consistently calculate a value for Rsh given the available

parameters listed in Section 3.3. Rsh is determined by the slope of the I-V curve at the

short circuit point [29]:

dI 1
=− (3.3)
dV v = 0 Rsh
 28

For this calculation to be made accurately, empirically determined I-V characteristic

curves must be available. Therefore the model will be modified according to (3.2) which

may result in a bias that overestimates the power production of the solar array.

Figure 3.3: Model of a solar cell without shunt resistance

⎛ (VcnkT
+ I c Rs ) q
⎞
I c = Il − Io ⎜ e cK
− 1⎟ (3.4)
⎜ ⎟
⎝ ⎠

Figure 3.3 depicts the model of a solar cell neglecting the effects of shunt

resistance. (3.4) describes the electrical behavior of this circuit. This is the circuit model

that will be used to simulate the behavior of a solar cell in this thesis. All of the

parameters of (3.4) can be determined from the input parameters listed. This equation is

the basic building block of the array model just as the solar cell is the building block of

an actual array. The equation must now be modified in order to model the behavior of a

photovoltaic array.

3.2. Conversion of Cell Characteristic Equation to Array Characteristic Equation

In order to model a solar array, a circuit composed of series and parallel

connections of solar cells as shown in Figure 3.4 is utilized. For the purpose of modeling

an array an assumption is made that the current and voltage characteristics of each cell in

the array is uniform (A3). This is important for two reasons: First, it ensures the bypass
 29

diodes will be reverse biased and are treated as open circuits and the blocking diode is

forward biased and is treated as a short. Second, it ensures that the exponential

expression (3.4) may be scaled in a linear manner based on the series and parallel

connection of cells because the parameters in the exponent (Vc, Ic, Tck) will be identical

for each cell.

Figure 3.4: Circuit diagram of a 3x3 solar array

Where with ideal blocking and bypass diodes:

I A = array current (A)

VA = array voltage (A)

The following simplifying assumption is made:

A5: A solar module is composed only of series connections of solar cells.

Many solar modules used for grid tie systems are composed of 72 cells connected in

series [10]. This assumption greatly simplifies the modeling procedure and the

characteristic equation that is used as the final model of the array. With this assumption

in place, an array can be simply described as an Ns x Np grid of cells as shown in Figure

 30

3.4 where Ns describes the number of cells in a string and Np describes the total number

of strings in the array. While it is true that modules exist which contain parallel cell

connections, these are generally used for low voltage applications such as stand alone DC

systems with battery storage and are not utilized for grid connected photovoltaic

generators.

Given that in a series connection voltages sum and in a parallel connection

currents sum the following relations apply:

I A = N p Ic
I scA = N p I sc
(3.5)
VA = N sVc
VocA = N sVoc

where:
N s = number of cells connected in series
N p = number of cells connected in parallel
I scA = short circuit current of the array (A)
VocA = open circuit voltage of the array (V)

Remembering that Il = Isc and substituting these into equation (3.4) results in:

⎛ (VAnN+ I kT
A RsA ) q
⎞
I A = I scA − N p I 0 ⎜ e s cK
− 1⎟ (3.6)
⎜ ⎟
⎝ ⎠

The scheme presented above is a commonly accepted method for scaling the cell

characteristic equation in order to form the array characteristic equation [5]. Inputs to the

model developed here include parameters related to solar modules instead of single cells
 31

and thus this scaling method must be modified to account for module inputs. This is

described in section 3.3.2.

3.3. Step by Step Modeling Procedure

In order to develop the model parameters basic steps are outlined below. Detailed

discussions of each step are contained the subsections.

Step 1) Determine the connection scheme of the array (Section 3.3.1).

Step 2) Convert the cell characteristic equation to an array characteristic equation

(Section 3.3.2.).

Step 3) Solve for parameters of the characteristic array equation (Section 3.3.3.).

Step 4) Calculate the I-V curve of the characteristic equation and determine PGDC
(Section 3.3.4.).

3.3.1. Determine the Connection Scheme of the Array:

The array connection scheme refers to the number of solar modules connected in

series to form a string and the number of parallel strings used to form the array. The

array characteristic equation requires the module connection scheme in order to specify

(Ns, Np), however this information is not available therefore the model determines an

appropriate connection scheme from the desired power rating and equipment

specifications. In many cases the only information a system planner/operator will have

concerning the connection scheme will be the rated power output PrA of the array.

However since the mathematical model is built by scaling the equation of a single solar

cell to describe an array, a specific connection scheme is needed. Therefore, the

following procedure of (3.7) has been developed to determine the probable connection

scheme:
 32

N sM = ciel (VDC / VmM );

PCol = N sM i PrM ;
(3.7)
N pM = ciel ( PrA / PCol );
ActualPrA = N sM i N pM i PrM ;

where:
ciel () = function that rounds to the next highest integer
VDC = The desired nominal DC voltage (usually 500V)
N sM = Number of modules in series required to achieve VDC
PCol = Power rating of a single string of N sM modules
N pM = number of series strings connected in parallel required to achieve PrA

First the minimum number of series modules, NsM, required to achieve the desired

voltage output at rated operation is determined. Second, the power rating of this string of

modules, PCol, is calculated, and third, the minimum number of strings required to

achieve the desired power, NpM, is determined. Fourth, the actual power rating of the

system that will be integrated with the power flow is calculated.

This scheme builds a PHV array in a discrete manner, as the power rating of the

overall array can only be increased in steps corresponding to the size of the power rating

of a single string of modules. Thus, for an arbitrary desired power rating, this scheme

designs an array with a rating greater than or equal to the rating input by the user. The

method is considered conservative for system studies since the effects of PHV

generation, such as phase imbalance, and overall demand curve changes increase as the

PHV system rating increases. It is noted that the power rating used in the power flow

solution is bounded because the over estimate does not exceed the rated power of a single

string of modules. The size of a single string of modules is limited because it is based on

the input DC voltage range of the power conditioning device. The effect of
 33

overestimation is significantly reduced by the fact that in utility-based distribution power

flows each bus typically represents an aggregate bus such as a pole top transformer which

services a number of loads. Therefore the PHV system on a given bus represents the

aggregate of all the PHV systems serviced by that bus, and for this aggregate system the

overshoot is limited to a single string of module.

3.3.2. Convert the Cell Characteristic Equation to an Array Characteristic Equation

Once the connection scheme is determined the characteristic equation must be

scaled up to describe the behavior of the entire array. This is done according to (3.5) and

(3.6) however it is important to note that values for Isc, Voc, and Pr are given at the

module level. Therefore, the procedure is modified to scale appropriately from the

module level to the array level:

I scA = I scM i N sM
VocA = VocM i N sP (3.8)
PrA = N sM i N sP i PrM

The number of cells in series is the product of the number of series cells in the module,

Ncs, and the number of modules in the string, N sM . The number of parallel connections

of cells is simply the number of parallel strings of modules:

N s = N cs i N sM
(3.9)
N p = N sP
 34

Once these calculations have been performed the appropriate substitutions can be made in

(3.10). This is the form of the I-V characteristic equation that will be used to model the

photovoltaic array:

⎛ (VAnN+ I kT
A RsA ) q
⎞
I A = I scA − N p I 0 ⎜ e s cK
− 1⎟ (3.10)
⎜ ⎟
⎝ ⎠

3.3.3. Solve for parameters of the characteristic array equation

• Calculate the series resistance of the array RsA

• Calculate the array open circuit voltage and short circuit current at the given

environmental conditions

• Calculate the saturation current of the diffusion diode I0

Calculate the series resistance of the array RsA:

The series resistance of the array is modeled as a single lumped resistor as is

appropriate given assumption A2. While specifications sheets for solar modules do not

provide a value for the series resistance a value for the series resistance can be calculated

from the parameters that are provided, the key parameter being the maximum power

production at standard test conditions. Parameters needed from previous calculations to

calculate series resistance are (VocA, IscA, PrA). Note that these parameters refer to the

entire array. Thus the series resistance of the entire array will be calculated directly.

In addition, the following calculations will make use of the parameter fill factor

defined as follows:
 35

VrA I rA
FFA = (3.11)
VocA I scA

The fill factor of a given array describes the relationship between the rated power output

and the product of the short circuit voltage and open circuit voltage. While the fill factor

is not strictly a measure of efficiency it is indicative of the efficiency of the array. This

measure is useful because it is possible to approximate FF0A defined as the fill factor of

an array where RsA = 0, using the empirical relationship in (3.12):

voc − ln(voc + 0.72)

FF0 A =
1 + voc
(3.12)
V q
voc = ocA
N s nkTcK

where:
FF0 : Fill Factor of the array without a series resistance

The key to the following derivation is that (VocA, IscA) remain the same in the two cases of

an array with and without series resistance. This allows us to define a relationship

between (3.11) and the empirically determined [5] (3.12) which involves the parameter of

series resistance. The derivation for the value of RsA will now be described:

PrA = VrA I rA
PrA = PrAwoR − I rA2 RsA
⎛ I ⎞
PrA = PrAwoR ⎜1 − rA RsA ⎟ (3.13)
⎝ VrA ⎠
I rA I scA
≈
VrA VocA
 36

where:
PrAwoR = Denotes power the array would output if RsA = 0
(VrA , I rA ) = Voltage and current of the array at rated power

The approximation shown in (3.13) is an assumption based on the shape of the I-V

characteristic of the solar cell.

⎛ I ⎞
PrA = PrAwoR ⎜1 − scA RsA ⎟
⎝ VocA ⎠
⎛ I ⎞ (3.14)
PrAwoR ⎜1 − scA RsA ⎟
FF =
PrA
= ⎝ VocA ⎠ = FF ⎛1 − I scA R ⎞
0⎜ sA ⎟
VocA I scA VocA I scA ⎝ VocA ⎠

Now solve for RsA:

VocA PrA
RsA = − 2
(3.15)
I scA FF0 A I scA

Alternative methods for the calculation of series resistance do exist [29]. This

method requires the use of many empirically determined I-V characteristics of the solar

module or array and therefore is impractical for use in this algorithm because the detailed

information required is often unavailable.

Calculate the array open circuit voltage and short circuit current at the given

environmental conditions:

This section describes how the short circuit current and open circuit voltage of the

array are transformed from rated values (IscAr, VocAr) to the appropriate values given the

environmental state (IscAE, VocAE). IscAE is substituted directly into (3.10) and both (IscAE,
 37

VocAE) are used to calculate the saturation current of the diode I0 later in this section.

While the short circuit current and open circuit voltage do not describe operating points

of a solar generator during normal operation, the operating points are determined from

these parameters. Therefore, for purpose of discussion it is appropriate to assume that the

impact environmental conditions have on the short circuit current and open circuit

voltage will be similar to the effect on the operating current and voltage respectively of

the array.

The rated values (IscM, VocM), are determined empirically at standard test

conditions (3.16) and published in the specification sheet of the module:

G = 1000 W/m 2
TA = 25°C (3.16)
Atmospheric Spectrum = AM 1.5

The current produced by a solar cell is directly dependant on the solar irradiance incident

upon its surface. The current produced by the array is proportional to the energy input in

the form of solar irradiance. The array output voltage is heavily impacted by the

temperature of the solar cell because the voltage handling capability of the semiconductor

material is dependant on the temperature of the material.

The nominal operating cell temperature or NOCT is a value determined

empirically under a separate set of standard test conditions of 800 W/m2, 20ºC, AM 1.5.

It is used to calculate the cell temperature as a function of ambient temperature:

NOCT − 20
Tc = Ta + iG (W/m 2 ) (3.17)
800
 38

As shown in (3.18) the energy output in the form of current is directly

proportional the energy input in the form of solar irradiation. There is a small

⎛ ∂I ⎞
temperature coefficient, ⎜ scA ⎟ , on the order of a few milliamps per degree Celsius to
⎝ ∂Tc ⎠

account for temperature differences recognized empirically.

I scAr ⎛ ∂I ⎞
I scAE = G + ⎜ scA ⎟ (Tc − Tc 0 ) (3.18)
1000 ⎝ ∂Tc ⎠

where:

Tc = Cell temperature (°C)

Tc 0 = Cell temp at standard test conditions (usually 25°C)
TA = Ambient temperature (°C)
NOCT = Nominal Operating Cell Temperature
Ar = Denotes rated (max power point at standard test conditions) value of the array

AE = Denotes value for the array at given environmental conditions

The open circuit voltage of the array under given environmental conditions is

calculated in (3.19):

⎛ ∂V ⎞ kT ⎛I ⎞
VocAE ≈ VocAr + ⎜ ocA ⎟ (Tc − Tc 0 ) + cK ln ⎜ scAE ⎟ (3.19)
⎝ ∂Tc ⎠ q ⎝ I scAr ⎠

As the temperature of the cell rises, the voltage capability of the semiconductor material

in the cell is diminished and thus the voltage across the cell decreases. This can be

determined through manipulation of the PHV characteristic equation. A temperature

 39

coefficient is inserted to account for empirically determined effects. A derivation of this

equation is included in Appendix B.

The values for open circuit voltage and short circuit current at the given

environmental conditions are thus determined [5]. These values will now be used to

solve for the saturation current of the diffusion diode.

Determination of the saturation current of the diffusion diode I0:

With substitutions made for operation under given environmental conditions

(3.10) can be written as:

⎛ (VAnN+ I kT
A RsA ) q
⎞
I A = I scAE − N p I0 ⎜ e s cK
− 1⎟ (3.20)
⎜ ⎟
⎝ ⎠

The parameter I0 is determined by setting the equation to open circuit conditions, and

solving for I0:

IA = 0
VA = VocAE (3.21)
I scAE
I0 =
⎛ VocAE q ⎞
N p ⎜ e nN s kTcK − 1⎟
⎜ ⎟
⎝ ⎠

3.3.4. Calculate the I-V Curve of the Characteristic Equation and Determine PGDC

The characteristic equation that is used to model the PHV array is repeated in

(3.22):
 40

⎛ (VA + I A RsA ) q ⎞
I A = I scAE − N p I 0 ⎜ e nN s kTcK − 1⎟ (3.22)
⎜ ⎟
⎝ ⎠

The variable IA occurs twice in (3.22), and therefore there is no unique solution to

this equation. In order to find the correct solution a known initial condition is required.

This model uses the open circuit point (0, VocAE) for this purpose. It can be determined by

setting the value of IA to zero in (3.22):

⎛ nN kT ⎞ ⎛ I scAE ⎞
VocAE = ⎜ s cK ⎟ ⎜
ln ⎜ + 1 ⎟⎟ (3.23)
⎝ q ⎠ ⎝ Io N p ⎠

With (0, VocA) as the initial solution the iterative technique in (3.24) is used to calculate

the I-V characteristic of the photovoltaic array:

VA = VocAE ;
I A1 = 0;
while VA > 0
VA = VA − small _ step;
⎛ (VA + I A1RsA ) q ⎞
I A = I scA − N p I 0 ⎜ e nN s kTcK − 1⎟ ;
⎜ ⎟
⎝ ⎠
while abs(I A1 − I A ) > epsilon (3.24)
I A1 = I A1 + ( I A − I A1 ) /10;
⎛ (VA + I A1RsA ) q ⎞
I A = I scA − N p I 0 ⎜ e nN s kTcK − 1⎟ ;
⎜ ⎟
⎝ ⎠
end
I A1 = I A ;
end
 41

As can be seen from (3.24) the initial solution for IA is input to the equation as IA1. The

value of VA is reduced by small step and iterations are performed until the values of IA

and IA1 are within a given tolerance epsilon. Once a solution is determined the procedure

is repeated until the short circuit point is reached. I-V curves are thus formed for a solar

array given any environmental conditions.

A power vector is calculated from the I-V curve by multiplying the stored vectors

of I and V. The DC power output of the array is then defined as the maximum element of

the power vector:

Pˆ = IV
ˆˆ
(3.25)
PGDC = Pmax AE = max( Pˆ );

It is assumed that the PHV array operates at the maximum power point. In the

physical system the output of the array may not be the maximum power point however

the PHV output is fed to the MPPT which assures the maximum power point is fed into

the DC to AC inverter. Therefore the described model accounts for the behavior of both

the PHV array and the MPPT. The power loss in the MPPT is taken into account using

the efficiency parameter of the PCU as described in Chapter 4.

 42

3.4. Flow Diagram of Procedure for Calculating Power Output of PHV Array

The flow diagram shown in Figure 3.5 describes the PHV array model by defining

the equations that are implemented in the model and their input output relationships.

⎛ ∂I scM ⎞ ⎛ ∂VocM ⎞
⎜ ⎟, ⎜ ⎟
⎝ ∂Tc ⎠ ⎝ ∂Tc ⎠

PGDC
Figure 3.5: Flow diagram of PHV array model
 43

3.5. Implementation and Example

This model has been implemented in MATLAB in order to interface with the

software developed in [35]. The output of the model for an example PHV array is

presented here in order to describe a typical residential system under a given set of

environmental conditions. Figures 3.6 and 3.7 show a typical PHV array I-V curve and

PV curve respectively. System specifications are described below:

• 4500 Watts rated PHV Array

• Shell SQ 150 W Modules [4]
• 2 series strings of 15 modules each
• Ambient temperature: 22ºC
• Solar irradiance: 850 W/m2
• 26.9 m2 total solar module area
• Xantrex GT-3.8 PCU, η = 95.7%

Voltage vs. Current

9

6
DC Amperes

0
0 100 200 300 400 500 600 700
DC Volts

Figure 3.6: I-V Characteristic of example PHV Array (circle denotes max power point)
 44

Voltage vs. Power

4000

3500

3000

2500
Watts

2000

1500

1000

500

0
0 100 200 300 400 500 600 700
Volts

Figure 3.7: P-V Characteristic of example PHV Array (circle denotes max power point)

Notice from Figure 3.5 that the operating voltage of the array is ~470 V, which is

within the ideal operating range of the inverter (195-550 V) [7]. The power output of this

array under the given conditions is 3544.3 Watts DC. Power output of a PHV array

varies based on the environmental conditions, and is most often less than the rated output

which describes laboratory testing conditions. Chapter 5 contains extensive simulations

using various environmental conditions.

 45

4. Model of the Power Conditioning Unit and Integration with Power Flow

This chapter will present the conversion of the PHV array output power PGDC to an

AC power injection SGAC to the grid dependant on the source model selected:

PQ: SGAC ( PGDC , η , assumed pf)

(4.1)
P|V|: PGAC , QGAC,lim ( PGDC , η )

where:
SGAC = complex power injection by PHV generator
PGAC = real power injection by PHV generator
QGAC,lim = limit of the reactive power inection by the PHV generator
η = efficiency of the PCU
pf = power factor

As discussed in Chapter 2 there are two control schemes available for the inverter of the

PCU which correlate to two distinct generator models, the PQ source model and the P|V|

source model. If the constant PQ source model is selected the complex power injection

SGAC is completely determined by the characteristics of the PHV model. If the constant

P|V| model is selected the real power injection and voltage magnitude of the output bus

are determined while the reactive power injection is determined by power flow

computations. A reactive power limit is also calculated in order to assure the current

production capabilities of the PHV generator are not exceeded. PHV generators are often

connected at the load site as most PHV generators are owned by individual consumers.

Section 4.2 describes how the calculated power injections are integrated with the model

of the existing distribution system for power flow analysis.

 46

4.1. Power Conditioning Unit Modeling

Most PHV generators utilize self commutated voltage source inverters and the

inverter control scheme used in the PCU of the PHV generator determines the generator

model to be implemented. Generators which utilize current controlled inverters are

modeled as constant PQ sources and generators which utilize voltage controlled inverters

are modeled as constant P|V| sources. In each case, the efficiency, η, of the PCU is

considered constant and the output power of the PHV system PGDC , is scaled according to

(4.2):

' = η i PG
PGDC DC
(4.2)

The constant PQ and constant P|V| power flow component models are described in the

following sections.

4.1.1. The Constant PQ Source Model

Most residential and small commercial (<10kW) PHV generators employ

inverters which utilize the current control scheme. In this scheme the phase of the current

to be output by the PCU is controlled to match a voltage reference measured at the output

bus of the inverter. This emphasis on matching the current phase angle to the output bus

ensures a high output power factor for the PHV generator [14]. Most commercially

available PCU’s specify an output power factor close to 1 under normal conditions and

no lower than 0.9 [33]. The generator acts as a current source and does not control the

voltage at its output bus. Thus a constant PQ model is selected for integration with the

power flow module. For this model the real and reactive power injections PGinj and
 47

QGinj are functions of the PHV array model under the given environmental conditions and

the assumed PF of the inverter as follows:

PGinj = pf i PGDC
'

QGinj = (sin(arccos( pf ))i PGDC

' (4.3)
SGinj = PGinj + jQGinj

4.1.2. The Constant P|V| Source Model

It is likely that large, three-phase commercial sized (>10kW) PHV generators

would incorporate an inverter using the voltage control scheme and thus would act as a

voltage source. As such, commercial sized PHV installations will be modeled utilizing a

constant P|V| generator model.

For a given environmental condition the apparent power capacity of the generator

SGinj is determined by the PHV array model (4.4) and the desired | V | is specified:

| SGinj |≤ PGDC
' (4.5)

The current capacity of the PHV generator is limited by SGinj therefore it is necessary to

define another parameter in order to determine the constant PGinj injection and the reactive

power limit QGinj,lim . In this case, it is appropriate to define a minimum power factor for the

power output of the array in order to accomplish this. A power factor of 0.9 will be

selected as the minimum power factor because this is stated to be the minimum power

factor for some commercially available PCU’s [33]. Therefore PGinj and QGinj,lim are

calculated in (4.6):
 48

PGinj = 0.9i PGDC

'

QGinj,lim = (sin(arccos(0.9))i PGDC

'
(4.6)
constraint:
QGinj ≤ QGinj,lim

The P|V| generator is then integrated with the power flow module in the standard

manner a generator bus treated in the Newton Raphson power flow method. When an

iteration is completed a value for QGinj is calculated and compared to QGinj,lim . If the

constraint is violated the power flow is considered invalid because a solution does not

exist where both the constraint and the voltage magnitude requirement are satisfied.

4.2 Power Flow Integration

An unbalanced three phase distribution power flow solver has been developed in

[14] and implemented in MATLAB. This implementation models the steady state

behavior of a three phase distribution system based on information in a case file which

describes the components and topology of the distribution system to be modeled. The

power flow model does include an implementation of PQ and P|V| sources.

The power flow models single phase and two phase networks, as well as

unbalanced loads and generators. In this implementation load busses often represent

aggregate loads such as a pole top transformer which services a number of residential and

commercial units, or an industrial substation feeding large industrial loads. The PHV

generator model is represented as an injection at the load bus therefore the PHV generator

model is also an aggregate of all the PHV generators connected to the given bus. An

unbalanced PHV generator may represent a situation where the number and capacity of

PHV generators connected to a given bus is unbalanced. In general, large PHV

 49

generators (10>kW) and those modeled as constant P|V| sources are balanced three phase

systems and are modeled as such in simulations.

4.2.1. PQ Source Model Integration

As stated in Chapter 2, a PQ source is integrated with the power flow by adding

the generated power to the load injected power at bus k as in (4.4):

S kinj = S Lk
inj
+ SGACk (4.4)

where:
Skinj = total power injected at bus k
inj
S Lk = power injected by the load at bus k

4.2.2. P|V| Source Model Integration

The power injection of the PHV generator must also be combined with the load

injection for the constant P|V| case. The real power injection and reactive constraint at

bus k are thus defined in (4.5) and the voltage magnitude | Vk | is specified:

Pkinj = PLkinj + PGinjk

inj
QGk = Qkinj − QLk
inj

(4.5)
constraint:
inj
QGk ≤ QGk
inj
,lim

where:
PLkinj = total real power injection specified for bus k
Qkinj = total reactive power injection at bus k
calculated from results of the power flow
(PGkinj , QGk
inj
) = real and reactive power injections due to PHV generator
 50

For the Newton Raphson implementation presented in [35] the variable, Jacobian,

and mismatch matrices are created based on the network topology before all P|V| busses

are determined. Therefore it is necessary to remove the appropriate rows and columns

corresponding to the voltage magnitude of P|V| busses. If the constraint is violated the

PHV generator model is converted to the constant PQ type with a given power factor, for

example 0.9. In this case, the voltage requirement | Vk | will not be met in the solution.

The PHV array and PCU models have been developed and integrated with

existing unbalanced distribution power flows. The following chapter presents the

implementation details, simulation results and subsequent analysis performed on large

scale distribution systems. A 394 bus test system is used as an example

 51

5. Simulation Methods and Results

The model described in Chapters 3 and 4 has been implemented in MATLAB and

integrated with an existing power flow program in order to perform system studies on

large scale distribution systems. For each power flow computation, a load level and set

of environmental conditions must be selected and input to the system. Then, a 394 bus

test system of a real world power network is used as an example for the following

studies.

5.1. Test System and Case File Construction

The 394 bus test system with 199 load busses that is used for simulation studies in

this work is shown in Figure 5.1. For a simulation to be performed a case file must be

constructed which describes the system conditions. The case file must contain data for

both the electrical and weather conditions in order to fully describe the test system.

Electrical data includes load data as well as the location and ratings of installed PHV

generators. This section describes the construction of case files which include this

information. Cases studies and corresponding result are summarized in Table 5.1

Table 5.1: Summary of simulation study conditions

Study Conditions
Example Network Parameters Load Locations of Ratings of Environmental
Studies and Configuration Levels PHV Systems PHV Conditions Results
Single
Flow Constant Constant Constant Varied Constant Table 1
Constant Figures
Twenty for each 24 5.2, 5.3,
Four Hour Constant Varied Constant hour trial Varied 5.4, 5.5
 52

3
4

Figure 5.1: One line diagram of 394 bus test system

5.1.1. Electrical System Data

Power flow data includes network, load and generation parameters. This section

will focus on the two most relevant components to this work which are the loading

conditions and the configuration of installed PHV generators. These will now be

discussed respectively.

A base case loading condition was provided for the 394 bus test system, which

represented peak demand. The loads have been divided into categories of residential,

commercial, and industrial loads and each type of load may be scaled separately to

describe a desired loading level.

In order to create 24 hour load profiles from the base case, existing load profiles

may be used to describe demand of the test system for a specific hour of a given load

profile. Seasonal load profiles are available online for free from Itron through their
 53

eShapes database [15]. This database contains load profiles for ten geographic regions

for the years 2001-2005, and specifies the type of load which allows for the use of

residential, commercial and industrial example load profiles. These profiles are used to

construct load data for case files according to the procedure described in Appendix C.

With respect to PHV generators, the case file must include data on the location

and rating of the generator. The 394 bus test system has 199 load buses. It is appropriate

to place PHV generators on these busses because of the previously stated trend that most

PHV generators are owned by the consumer and placed at the load site. It is also

appropriate to place residential sized (<10kW) systems on load busses classified as

residential and larger systems on commercial or industrial load busses. Simulation

studies will use specific schemes to determine the desired locations and capacity for PHV

generators. These schemes will be described in the following sections which outline the

specific simulations performed.

5.1.2. Environmental Data

Measured hourly environmental data is available through the Typical

Meteorological Year 2 (TMY2) database compiled by the National Renewable Energy

Laboratory [27]. This database is a compilation of data measured during the years 1961-

1990. It was constructed by selecting months considered to be average months from

individual years and concatenating them to form a single year of data. The database was

compiled specifically for use in computer simulations of solar energy systems.

The PHV generator model uses two parameters which are available in the TMY2

database, solar irradiation and ambient temperature. Several measurements of irradiance

are included in the TMY2 measured data, however the most appropriate for use is that
 54

labeled in [27] as the Global Horizontal Radiation (W/m2), defined as the total direct and

diffuse radiation incident upon a horizontal surface. The array of most solar generators is

installed at an angle which maximizes the time at which direct rays from the sun are

normal to the surface of the panels. Therefore use of the irradiance measured on a

horizontal surface as an input will create a bias which underestimates the production

capability of a PHV generator. The measurements which are given

The following sections describe specific simulations and their results. In each

simulation environmental data is matched with appropriate load data from the correct

season.

5.2. Single Flow Studies

Single power flow studies are the basic tool to be used in building 24 hour or

longer studies. Thus given a single set of PHV installation locations and ratings,

environmental conditions, network load conditions and network parameters, a power flow

study can be made. Here, arbitrary levels of PHV penetration are considered based on a

percentage of the total substation demand (real power) of the base case (no PHV

generation). Some US state mandates have set a goal of 20% renewable generation and

therefore increments of 5% up to 20% have been performed. In each case the total PHV

generation was spread evenly over 65 of the 199 load busses. This resulted in PQ bus

power flow models of the PHV generators. An arbitrary unbalanced, aggregate PHV

installation at each bus was selected with:

• phase A and phase B at 40% of the PHV generation

• phase C at 20%

For this study the peak load levels of the original distribution system case file are used.
 55

In order to measure the level of imbalance at the substation, the percentage

difference between the individual phase apparent powers is determined with respect to

the total apparent power, |S3Φ|, at the substation. For example, between phase a and b, the

difference is calculated by the following:

abs (| S a | − | Sb |)
| S a −b | % = *100 (5.6)
(| S3φ | / 3)

The per unit difference in voltage between two phases is calculated as:

| Va −b | p.u . = abs(| Va | − | Vb |) (5.7)

The results of the power flow at peak loads for this study are shown in Table 1.

Table 5.2: Imbalance study varying PHV penetration

PHV Penetration max |Sx-y|% |Sx-y| max |Vx-y| |Vx-y|
0% 3.12 a-b 0.00094 c-a
5% 3.63 c-a 0.00139 b-c
10% 6.50 c-a 0.00178 b-c
20% 8.42 c-a 0.00217 b-c

5.3. Twenty Four Hour Simulation Studies

Twenty four hour simulations are designed to investigate the impacts the

intermittent nature of PHV generation has on the distribution system. These simulations

use the methods described in section 5.1 to create load profiles which are then matched

with an appropriate environmental profile in order to simulate the behavior of a

distribution system over the course of a 24 hour period. The environmental data provided

are single measurements taken once an hour. Therefore the simulation results are limited
 56

in their ability to model system behavior for the entire 24 hour period. The results do not

represent an average of the system behavior over time they represent the system behavior

at 24 moments during a 24 hour period.

5.3.1 Example Twenty Four Hour Simulation Study

An example is provided here which studies the behavior of the 394 bus test

system on a peak load day with peak irradiance conditions. The example is similar to

that of the single flow study presented above in that the total PHV generation was spread

evenly over 65 of the 199 load busses, which results in PQ bus power flow models.

Again the arbitrary unbalanced, aggregate PHV installation at each bus was selected

with:

• phase A and phase B at 40% of the PHV generation

• phase C at 20%

For these simulations loads were categorized and scaled appropriately according

to the procedure described above. The base case which contains no PHV generators and

thus describes the load profile of the distribution system at the substation is shown in

Figure 5.2.
 57

Real Power Demand of 394 bus System over 24 Hours

30000.00

25000.00

20000.00
P Watts

15000.00

10000.00

5000.00

0.00
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Hour

Figure 5.2: Real power demand of the 394 bus test case for peak load

Environmental data corresponding to Binghamton, NY was selected because of its

geographical proximity to the 394 bus test system located in Elmira, NY. The day

containing the highest irradiance measurement in the TMY2 database was selected for the

environmental input as an appropriate pairing for the peak loading day of the distribution

system. The corresponding irradiance and temperature curves are show in Figures 5.3

and 5.4 respectively.

 58

Irradiance Binghamton 7-16

1000
900
800
700
G (W/m^2)

600
500
400
300
200
100
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
Hour

Figure 5.3: TMY2 Irradiance profile for binghamton NY, July 16

Temperature Binghamton 7-16

25

20
Degrees Celcius

15

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
Hour

Figure 5.4: TMY2 Ambient temperature profile for Binghamton NY, July 16

The results of the 24 hour simulations are shown in Figure 5.5. As in the single flow

simulation the overall rated capacity of installed PHV generation is varied from 5% to

20% of the overall power demand.

 59

Real Power Demand of 394 bus System over 24 Hours

30000

25000

20000 0%
5%
P Watts

15000 10%
15%
10000 20%

5000

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
Hour

Figure 5.5: Real power demand at the substation under varying capacity of PHV
generation

5.3.2. Observations and Remarks

In order to gain insight regarding the nature of PHV generation it is valuable to

simulate the power output of an example PHV generator over a 24 hour period using an

available environmental profile. Figure 5.6 shows the power output of a 1000W PHV

generator under the same environmental conditions used in the previous section

(Binghamton NY, 7-16) [27].

 60

24 Hour Real Power Output Profile

900

800

700

600
P (Watts)

500

400

300

200

100

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
Hour

Figure 5.7: Real power output of a 1000W rated PHV generator

By comparing Figure 5.6 to Figures 5.3 and 5.4 it can be seen that the power

output of a PHV generator is most heavily impacted by the irradiance incident upon its

surface. This conclusion is in agreement with Chapter 3 which states that the electrical

energy output of a PHV generator is most directly proportional to the energy input in the

form of solar irradiance incident on the surface of the solar array.

Figures 5.3 and 5.5 show one of the potential impacts a high penetration of PHV

generation may have on distribution systems. Solar irradiance and thus PHV generator

production peaks at 1:00pm followed by a sharp drop off while the load demand peaks at

3:00pm and drops off more slowly towards the end of the day. Therefore overall system

demand changes rapidly during this afternoon drop off in solar radiation. This change in

load profile could present a new challenge for distribution system planners and operators.
 61

6. Conclusion and Future Work

6.1. Conclusions

This thesis provided a review of PHV generators and presented the development

of a mathematical model of a grid connected PHV generator practical for use in large

scale distribution systems studies. The model was implemented and integrated with an

existing unbalanced distribution system power flow analysis. Simulations were

performed using a 394 bus test system with the purpose of identifying and characterizing

distribution system impacts of grid connected PHV generators.

The presented PHV generator model was designed specifically for distribution

system studies. Therefore, environmental and generator input parameters were limited to

those readily available to distribution system planners and operators. The PHV generator

model was created using the generally accepted I-V characteristic equation of a PHV cell

as its basic building block. The model accounted for the two most common inverter

control schemes by allowing for integration with distribution power flow as either a

constant PQ or constant P|V| source. The model can be integrated with several methods

of distribution power flow and specific schemes for integration with the Newton

Raphson, and backward/forward sweep methods were described.

The model was implemented in MATLAB using an existing Newton Raphson

unbalanced distribution power flow solver. Simulation methods were developed to

capture the intermittent nature of PHV generation. Hourly load levels were calculated

using test system peak load levels and publicly available distribution load profiles. These

methods were then used along with hourly environmental data to develop case files for 24

hour simulation of distribution systems with installed PHV generation. Simulations were
 62

performed with a 394 bus test system using both single flow and 24 hour simulation

methods in order to investigate distribution system impacts of grid connected PHV

generation. Specifically phase balancing, feeder balancing, system demand profiles, and

voltage rise were addressed.

6.2. Summary of Research Contributions

The work presented in this thesis is a continuation of research in the field of

steady state distribution system modeling. The following contributions to this research

field have been made:

• Practical PHV generator model for systems engineers requiring limited input data

• Matlab based simulation tool which integrates the PHV model with an unbalanced

distribution power flow solver for system studies

• Simulation methods, results, and analysis for 24 hour system studies using a 394

bus test case.

• Demonstrated that increased penetration of single phase PHV generators may

increase distribution system imbalance levels which would be a cause of concern

for system planners and operators.

• Demonstrated that a 20% penetration of PHV generation will cause significant

changes to the shape of daily load profiles.

• Identification of early afternoon (12:00pm-2:00pm) as a time period of concern

due to rising demand and sharply decreasing PHV power production.

This thesis provides an additional component model which increases the breadth of cases

available for steady state distribution system studies, including an example of its

implementation and simulation results.

 63

6.3. Future Work

This thesis can be seen as a first step towards the comprehensive study of grid

connected PHV generation, its system impacts and possible control schemes to enhance

reliability, power quality, and economic operation of distribution systems. As penetration

levels of grid connected PHV generators increase the behavior of distribution systems

will become increasingly impacted by the intermittent nature, and single phase

characteristics of this renewable energy source. It may be necessary for new schemes to

be developed for applications such as load balancing, capacitor placement and switching,

service restoration, and distribution automation for economic operation. This work

provides engineers with a model which can be used to perform the system studies for

investigations into these fields.

This model could be used as a predictive tool where predictions of load and

environmental data could be used to predict the steady state behavior of a given

distribution system with grid connected PHV generation. Such an application would

require additional investigation regarding the PHV model in order to assure the model

accurately predicts the behavior of a physical system within a certain margin of error.

Such investigations are problematic because they would require a physical PHV

generator of the appropriate power rating, as well as the instrumentation necessary to

measure environmental inputs and system conditions. The cost of a residential sized

PHV generator is on the order of $10,000 and proper installation requires an appropriate

space with unobstructed southern exposure that is also accessible for measurement.

The model presented in this thesis can be utilized in a multitude of studies and

investigations. It provides engineers with a steady state simulation engine for distribution
 64

system studies with grid connected PHV generation. It can be used to simulate and

analyze distribution systems with varied PHV connection schemes and varied levels of

installed PHV capacity only a few of which are included in this work. It can also be used

to test and optimize control schemes designed for the case of distribution systems with

grid connected PHV generation. This work presents only a sample of the applications for

which this model may be used to improve planning and operations for distribution

systems with grid connected PHV generation.

 65

References

[1] K. Aoki, T. Ichimori and M. Kanezashi, "An efficient algorithm for load balancing of
transformers and feeders by switch operation in large scale distribution systems," IEEE
Trans. Power Del., vol. 3, pp. 1865–1872, Oct. 1988. 1988.

[2] Austin Energy, "Solar Rebate Program,"

http://austinenergy.com/Energy%20Efficiency/Programs/Rebates/Solar%20Rebates/inde
x.htm. 2006.

[3] M. E. Baran and F. F. Wu, "Network reconfiguration in distribution system for loss
reduction and load balancing," IEEE Trans. Power Del., vol. 4, pp. 1401–1407, Apr.
1989. 1989.

[4] California Public Utilities Commision, "California Solar Incentive Program,"

http://www.cpuc.ca.gov/static/Energy/solar/index.htm. 2006.

[5] L. Castaner and S. Silvestre, Modeling Photovoltaic Systems using PSpice. West
Sussex, England: Wiley, 2002,

[6] T. -. Chen, M. -. Chen, T. Inoue, P. Kotas and E. A. Chebli, "Three-phase cogenerator

and transformer models for distribution system analysis," IEEE Trans. on Power
Delivery, vol. 6, pp. 1671 - 1681, Oct. 1991. 1991.

[7] H. Chihchiang and S. Chihming, "Study of maximum power tracking techniques and
control of DC/DC converters for photovoltaic power system," Power Electronics
Specialists Conference, 1998. PESC 98 Record. 29th Annual IEEE, vol. 1, pp. 86, May
17-22. 1998.

[8] Y. C. Chuang, Y. L. Ke, C. S. Chen and Y. L. Chen, "Distribution system temperature

adaptive load transfer using colored petri net approach," IEEE Trans. on Power Systems,
vol. 8, pp. 2931, Aug. 2006. 2006.

[9] Energy Information Administration, "State Renewable Energy Requirements and

Goals: Status Through 2003," http://www.eia.doe.gov/oiaf/analysispaper/rps/index.html.
2004.

[10] Go Solar Company, "Shell Solar Panes SQ Series,"

http://www.solarexpert.com/Shell/Shell-SQ.html. 2006.
 66

[11] J. Grainger and W. Stevenson, Power System Analysis. New York: McGraw-Hill,
1994, pp. 787.

[12] F. Groppi, "Grid-connected photovoltaic power systems: Power value and capacity
value of PV systems," International Energy Agency, Milano, Italy, Tech. Rep. IEA PVPS
T5-11: 2002, February, 2002.

[13] T. Hoff and D. S. Shugar, "Value of grid-support photovoltaic’s in reducing

distribution system losses," IEEE Trans. on Energy Conversion, vol. 10, Sept. 1995.
1995.

[14] T. Ishikawa, "Grid-connected photovoltaic power systems: Survey of inverter and

related protection equipments," International Energy Agency (IEA), Tokyo Japan, Tech.
Rep. IEA-PVPS T5-05: 2002, December, 2002.

[15] Itron, "eShapes,"

http://www.itron.com/pages/products_detail.asp?id=itr_000491.xml. 2006.

[16] S. B. Kjaer, J. K. Pedersen and F. Blaabjerg, "A review of single-phase grid-

connected inverters for photovoltaic modules," Industry Applications, IEEE Transactions
on, vol. 41, pp. 1292, Sept.-Oct. 2005.

[17] F. Ma and D. Curtice, "Distribution Planning and Operations with Intermittent

Power Production," Ieee Transactions on Power Apparatus and Systems, vol. 8, pp. 2931,
August. 1982.

[18] F. Ma and L. Isaken, "Impact of Dispersed Supply Management on Electric

Distribution Planning and Operations," IEEE Transactions on Power Apparatus and
Systems, vol. 5, pp. 1531, Sept/Oct. 1979.

[19] A. Makinen, M. Parkki, P. Jarventausta, M. Kortesluoma, P. Verho, S. Vehvilainen,

R. Seesvuori and A. Rinta-Opas, "Power quality monitoring as integrated with
distribution automation," Electricity Distribution, 2001. Part 1: Contributions. CIRED.
16th Int. Conf. and Exhibition on (IEE Conf. Publ no. 482), vol. 2, pp. 5, 18-21 June
2001. 2001.

[20] T. Markvart, Solar Electricity. Chichester, England: John Wiley & Sons, 1994,

[21] R. Messenger and J. Ventre, Photovoltaic Systems Engineering. USA: CRC Press,
2004,
 67

[22] K. N. Miu and H. D. Chiang, "Service restoration for unbalanced radial distribution
systems with varying loads: solution algorithm," IEEE Pwr. Eng. Soc. Summer Meeting,
vol. 1, July 1999. 1999.

[23] K. N. Miu, H. -. Chiang and G. Darling, "Capacitor placement, replacement and

control in large-scale distribution systems by a GA-based two-stage algorithm," IEEE
Trans. on Power Systems, vol. 12, pp. 1160 - 1166, Aug. 1997. 1997.

[24] K. N. Miu, H. D. Chiang and R. J. McNulty, "Multi-tier service restoration through

network reconfiguration and capacitor control for large-scale radial distribution
networks," IEEE Trans. on Power Systems, pp. 1001 - 1007, Aug. 2000. 2000.

[25] K. N. Miu, H. D. Chiang, B. Yuan and G. Darling, "Fast service restoration for
large-scale distribution systems with priority customers and constraints," 20th Int. Conf.
on Power Industry Computer Applications, pp. 3-9, 11-16 May 1997. 1997.

[26] S. Naka, T. Genji and Y. Fukuyama, "Practical equipment models for fast
distribution power flow considering interconnection of distributed generators," IEEE
Pwr. Eng. Soc. Summer Meeting, vol. 2, pp. 1007 - 1012, 15-19 July 2001. 2001.

[27] National Renewable Energy Laboratory (NREL), "Typical Mean Year 2 Database,"
http://rredc.nrel.gov/solar/old_data/nsrdb/tmy2/.

[28] Renewable Resource Data Center, "Solar Radiation Resource Information," 2006,
http://rredc.nrel.gov/solar/. 2006.

[29] D. Staszesky, "Application of distributed intelligence to automatic distribution

reconfiguration for reliability improvement," IEEE Pwr. Eng. Soc. General Meeting, vol.
3, pp. 2621, 12-16 June 2005. 2005.

[30] G. N. Tiwari, Solar Energy Fundamentals, Design, Modelling and Applications.

India: CRC Press, 2002,

[31] B. Verhoeven, "Utility aspects of grid connected photovoltaic power systems,"

International Energy Agency, Arnhem, The Netherlands, Tech. Rep. IEA PVPS T5-
01:1998, December, 1998.

[32] J. C. Wang, H. D. Chiang, K. N. Miu and G. Darling, "Capacitor placement and real
time control in large-scale unbalanced distribution systems: loss reduction formula,
 68

problem formulation, solution methodology and mathematical justification," IEEE Trans.

on Power Delivery, vol. 12, pp. 953 - 958, April 1997. 1997.

[33] Xantrex, "Xantrex GT Series Grid Tie Solar Inverters,"

http://www.xantrex.com/web/id/172/p/1/pt/23/product.asp. 2006.

[34] Q. Zhou, D. Shirmohammadi and W. H. E. Liu, "Distribution feeder reconfiguration

for operation cost reduction," IEEE Trans. on Power Systems, vol. 12, pp. 730 - 735, May
1997. 1997.

[35] R. D. Zimmerman, "Comprehensive Distribution Power Flow: Modeling,

Formulation, Solution Algorithms and Analysis," Doctoral Dissertation, Cornell
University, Jan. 1995.

[36] R. D. Zimmerman and H. D. Chiang, "Fast decoupled power flow for unbalanced
radial distribution systems," IEEE Trans. on Power Systems, vol. 10, pp. 2045 - 2052,
Nov. 1995. 1995.
 69

APPENDIX A: List of Symbols by Section

Section 2.1.1. The Photovoltaic Cell

I D = dark current (A)

I o = saturation current of the diode (A)
Vc = cell voltage (V)
q = 1.6*10−19 (coul )
k = 1.38*10−23 ( j / K )
TcK = Cell Temperature (K)
I c = cell current (A)
I l = photo current (A)
G = solar irradiance (W/m 2 )
TA = ambient temperature (°C)

Section 2.2. Unbalanced Distribution Power Flow

Vk −1 = steady state voltage at bus k-1

Vk = steady state voltage at bus k
I k = steady state current flow from bus k-1 to bus k
I k = steady state current flow from bus k to bus k-1
I (G ,C , L ) k = steady state current injection at bus k
S(G ,C , L ) k = steady state power injection at bus k

2.2.2. Constant P|V| Sources

Qk = specified value of reactive power injection at bus k

Pk = specified real power output of the generator at bus k
Vkt = calculated voltage value in bus k at iteration t
V fict , k = calculated voltage of fictitious node of bus k at iteration t
X fic ,k = fictitious branch impedance of bus k

3. Photovoltaic Array Model

 70

PGDC = calculated real power output of the array at given conditions (W)
Pra = rated DC power output of the array (W)
VocM = module open circuit voltage (V)
I scM = module short circuit current (A)
( I mM , VmM )= module maximum power point (A, V)

PrM = module rated power output (W)

NOCT = nominal operating cell temperature (°C)
⎛ ∂I scM ⎞ ⎛ ∂VocM ⎞
⎜ ⎟, ⎜ ⎟ = temperature coefficients
⎝ ∂Tc ⎠ ⎝ ∂Tc ⎠
N cs = number of cells in series in the module
G = solar irradiance (W/m 2 )
TA = ambient temperature (°C)

3.1. Characteristic Equation of the Non-Ideal Photovoltaic Cell

I c = cell current (A)

I sc = cell short circuit current (A)
I l = cell photo-current (A)
I o = saturation current of the diode (A)
Vc = cell voltage (V)
Rs = series resistance
Rsh = shunt resistance
q = 1.6*10−19 (coul )
k = 1.38*10−23 ( j / K )
TcK = Cell Temperature (K)
n = nonideality constant of the diode
Note: I l =I sc

3.2. Conversion of Cell Characteristic Equation to Array Characteristic Equation

I A = array current (A)

VA = array voltage (A)
 71

I A = N p Ic
I scA = N p I sc
VA = N sVc
VocA = N sVoc
N s = number of cells connected in series
N p = number of cells connected in parallel
I scA = short circuit current of the array (A)
VocA = open circuit voltage of the array (V)

3.3.1. Determine the Connection Scheme of the Array

ciel () = function that rounds to the next highest integer

VDC = The desired nominal DC voltage (usually 500V)
N sM = Number of modules in series required to achieve VDC
PCol = Power rating of a single string of N sM modules
N pM = number of series strings connected in parallel required to achieve PrA

3.3.3. Solve for parameters of the characteristic array equation

PrAwoR = Denotes power the array would output if RsA = 0

(VrA , I rA ) = Voltage and current of the array at rated power
FF0 : Fill Factor of the array without a series resistance
Tc = Cell temperature (°C)
Tc 0 = Cell temp at standard test conditions (usually 25°C)
TA = Ambient temperature (°C)
NOCT = Nominal Operating Cell Temperature
Ar = Denotes rated (max power point at standard test conditions) value of the array

AE = Denotes value for the array at given environmental conditions

4. Model of the Power Conditioning Unit and Integration with Power Flow

SGAC = complex power injection by PHV generator

PGAC = complex power injection by PHV generator
QGAC,lim = limit of the reactive power inection by the PHV generator
η = efficiency of the PCU
pf = power factor
 72

4.2.1. PQ Source Model Integration

Skinj = total power injected at bus k

inj
S Lk = power injected by the load at bus k
 73

APPENDIX B: Derivation of VocAE

It will now be shown how equation (3.19) has been developed. Given the I-V

characteristic equation of the solar array at standard test conditions (1000 W/m2, 25ºC):

⎛ (VArnN+ I kT
Ar RsA ) q
⎞
I Ar = I scAr − N p I0 ⎜ e s cK
− 1⎟ (B.1)
⎜ ⎟
⎝ ⎠

Set the current to zero in order solve for the open circuit voltage:

⎛ nNVocArkTq ⎞
0 = I scAr − N p I 0 ⎜ e s cK − 1⎟ (B.2)
⎜ ⎟
⎝ ⎠

kTcK ⎛ I scAr ⎞
VocAr = ln ⎜1 + ⎟ (B.3)
q ⎝ Io ⎠

Now the empirically determined temperature constant is inserted in equation (B.3):

⎛ ∂V ⎞ kT ⎛ I ⎞
VocAr + ⎜ ocA ⎟ (Tc − Tc 0 ) = cK ln ⎜ 1 + scAr ⎟ (B.4)
⎝ ∂Tc ⎠ q ⎝ Io ⎠

⎛ ∂VocA ⎞ kTcK ⎛ I scAr ⎞

⎜ ⎟ (Tc − Tc 0 ) = ln ⎜1 + ⎟ − VocAr + VocAE − VocAE (B.5)
⎝ ∂Tc ⎠ q ⎝ Io ⎠

Similar to equation 6.

kTcK ⎛ I scAE ⎞
VocAE = ln ⎜1 + ⎟ (B.6)
q ⎝ Io ⎠
 74

Substitute for VocAr and one occurrence of VocAE:

⎛ ∂V ⎞ kT ⎛ I ⎞ kTcK ⎛ I scAr ⎞
VocAE = VocAr + ⎜ ocA ⎟ (Tc − Tc 0 ) + cK ln ⎜1 + scAE ⎟− ln ⎜ 1 + ⎟ (B.7)
⎝ ∂Tc ⎠ q ⎝ Io ⎠ q ⎝ Io ⎠

Using logarithm rules:

⎛ I ⎞
⎜ 1 + scAE ⎟
⎛ ∂V ⎞ kT Io
VocAE = VocAr + ⎜ ocA ⎟ (Tc − Tc 0 ) + cK ln ⎜ ⎟ (B.8)
⎝ ∂Tc ⎠ q ⎜ I scAr ⎟
⎜ 1+ I ⎟
⎝ o ⎠

In general unity can be neglected in the numerator and denominator of the log term in

equation 11 and thus it reduces to:

⎛ ∂V ⎞ kT ⎛I ⎞
VocAE ≈ VocAr + ⎜ ocA ⎟ (Tc − Tc 0 ) + cK ln ⎜ scAE ⎟ (B.9)
⎝ ∂Tc ⎠ q ⎝ I scAr ⎠
 75

APPENDIX C: Calculating the Load Level for a Given Power Flow

In the case file load data is included for each load bus at conditions of peak

demand. Many combinations are possible for the type of load and model used. Balanced

or unbalanced, and grounded or ungrounded load types are possible and each load can be

modeled using the constant power, constant impedance or constant current load models.

Regardless of the model used however the load data which is input to the model is in the

form of power consumption at the given bus (C.1):

(S A
P,k , S PB,k , S PC,k ) (C.1)

where:

S PA, k = peak power consumption on phase A of bus k (kVA)

In order to model the behavior of a distribution system for a specific situation

such as a given hour of a given day of the year it is necessary to scale this peak load in an

appropriate manner to describe the situation which is desired for test. Load scalars are

defined in order to perform this function. Because different types of loads exhibit

different loading profiles three types of loads are considered separately, these are

industrial, commercial and residential loads. Scalars are calculated for each type of load

in order to model the load behavior of the entire system more accurately. Two scalars are

used to calculate the load level for a specific hour of a specific day of the year. The first

scalar Sd d referred to as the day scalar is used to calculate the peak load on a given day

(C.2):
 76

| S Pd ,k |=| S P ,k |i Sd d( I ,C , R ) | (C.2)

where:
Sd d = day scalar for day d
( I ,C , R )
= denotes (industrial, commercial, residential) load type
S Pd ,k = peak power consumption on day d on bus k (kVA)

The parameter Sd d is calculated from a given set of loading data. It is not necessary for

the load data to be specific to the distribution system under test, it is only necessary that

the load data be from a distribution system load of the same type (I,C,R). Sd d is

calculated as follows:

| SlPd |
Sd d = (C.3)
| SlP |

where:
SlPd = peak load on day d from given load data
SlP = peak load for the entire year from given load data

Once the peak load on a given day is determined the load for any hour of that day

can be calculated using the scalar Shd , h referred to as the hour scalar, as follows:

| S d ,h ,k |=| S Pd ,k |i Shd( I,h,C , R ) (C.4)

where:
Shd , h = hour scalar for day d, hour h
S d ,h ,k = power consuption on day d, hour h, at bus k

The parameter Shd ,h is calculated from load data on the given day in a similar manner as

the day scalar:

 77

| Sld , h |
Shd ,h = (C.5)
| SlPd |

where:
Sld ,h = power consumption for day d, hour h, from given load data

In this manner given the necessary load data the power consumption for all load

busses on the 394 bus test system can be determined for any day of the year at any hour

of the day. This describes the ideal case where daily load data is available for the period

of a year. This specific data is not always available, as often load data is grouped by

season where four sets of data are used to describe average days for the four seasons. In

this case the calculations above are valid however it would be appropriate for notation

purposes to replace the d which denotes day with an s to denote season.

 78

APPENDIX E: EXAMPLE SOLAR MODULE SPEC SHEET