You are on page 1of 13

CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO.

1, MARCH 2017 101

Case-specificity and Its Implications in Distribution


Network Analysis with Increasing Penetration of
Photovoltaic Generation
Kalpesh A. Joshi, Student Member, IEEE and Naran M. Pindoriya, Senior Member, IEEE

Abstract—Distribution system analysis (DSA) currently faces penetration of distributed energy resources (DERs) coupled
several challenges due to inclusion of distributed energy resources with “smart grid” developments [1]. The fact that the rear
(DERs), which have many characteristics, such as inherent vari- end of the power systems has been the least observable and
ability, uncertainty, possibility of flexible four quadrant converter
operations with distributed generation (DG), and the need for controllable adds to the difficulties in conducting realistic
efficient operations to improve reliability of the supply system. simulations and analytical studies. Steady state operational
This article argues for a high degree of case-specificity and issues in unbalanced distribution networks (UDNs) have also
discusses its implications in distribution networks with increasing become increasingly complex as a result of the continued
DG penetration. The research is based on the exhaustive yearly rise in integration of distributed generators (DGs). There are
simulation analyses of 132 candidate scenarios and investigates
the effects of feeder-specific factors, such as geo-electric size and several operational issues due to penetration of distributed
feeder spread, load density, and phase unbalancing. Nineteen resources in the distribution grid as elaborated in [2]. Present
(19) feeder variants—with phase-domain detailed modeling of all day DSA challenges are considered to be two-fold: 1) Each
feeder components, including DGs, are subjected to increasing study scenario becomes unique with levels of DER penetration
penetration of photovoltaic generation without altering the type being different; and 2) inherent intermittency and variability
and location of DGs. The objective is to analyze the role of feeder-
specific factors on feeder response characteristics in terms of in wind and solar energy based DGs interact in complex
annualized operational parameters, such as energy losses, feeder ways with voltage regulators and protection equipment [3].
voltage profile, average power factor, and peak demand at a These challenges demand an integrated approach for DSA with
substation node, as well as tap-changer operations of voltage increasing presence of DGs [4].
regulating equipment and their interaction with shunt compen- In this context, considerable research efforts have been made
sation. Recorded annual load profiles—industrial, commercial,
and residential—as well as location specific weather data are in analyzing the impact of photovoltaic distributed generators
used to simulate the candidate scenarios based on three IEEE test (PVDGs) in the distribution networks. A comprehensive steady
feeders and one actual spot network in India. Results signify the state analysis of typical distribution feeders with varying
consideration of feeder-specific factors in the planning exercise degrees of PVDG penetration is presented in [5]. This study
of grouping “similar” feeders for formulating the strategies that analyzes the effect of size and location (near, mid-point, and
can improve daily operations of distribution feeders. The demon-
strated case-specificity also implies that optimization algorithms rear with respect to substation node) of PVDG in distribution
for improved operations with DGs will need to be based on an feeders by performing hourly simulations in GridLAB-D over
integrated approach that accounts for feeder-specific factors as a period of one year. The effect of residential PVDGs on
well as cyclic variability of DERs. feeder voltage profile is also analyzed in [6] with varying
degrees of penetration. The analysis in [6] also considers
Index Terms—Case-specificity, distribution system analysis, contributing factors, such as feeder impedance, feeder length,
distributed generation, photovoltaic generation, time-series power
flow analysis, unbalanced distribution networks.
and transformer short-circuit resistance to determine voltage
rise caused by PVDGs. Effects of photovoltaic (PV) plant
locations with reference to substation is investigated in [7]
where the tap changing operations (TCOs) of the voltage
regulator (VR) are observed against the intermittency of PV
I. I NTRODUCTION generation. This study also emphasizes the proper modeling of
VR configurations and unbalance in DSA for realistic results.
D ISTRTIBUTION system analysis (DSA) is expected to
present many challenges, particularly due to increased
Impact investigation exercise is presented by detailed 3-Φ
modeling of a low-voltage actual network with a roof-top PV
plant in [8]. Impact assessment study with PVDG in a reduced
Manuscript received February 10, 2016; revised May 7, 2016 and July IEEE 8500 node network is presented in [9] to compare the
24, 2016; accepted September 2, 2016. Date of publication March 30, 2017;
date of current version October 12, 2016. This work was supported in part performance of transient and steady state simulations while
by Indian Institute of Technology Gandhinagar in the form of “Additional considering the time-delays in operation of voltage regulating
Fellowship” to Kalpesh Joshi. equipment. Effect of cloud movement over a grid-connected
K. A. Joshi (corresponding author, e-mail: kalpeshjoshi@ iitgn.ac.in) and
N. M. Pindoriya are with Indian Institute of Technology Gandhinagar, Gujarat, PV plant, in [10], has considered the unbalanced operation
India. of IEEE 13 node test feeder with quasi-static analysis. The
DOI: 10.17775/CSEEJPES.2017.0013 exercise in [10] analyzes the voltage profile of UDN with high
2096-0042
c 2017 CSEE
102 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

PV penetration and proposes a phase-shifting algorithm to II. P ROBLEM C ONTEXT


mitigate the adverse effect of cloud movement. Another at- It is evident that different levels of penetration of DGs and
tempt in impact analysis with single-phase PVDG in UDNs their intermittent nature of generation greatly influence the
is found in [11] wherein analysis is also extended to neutral- steady state daily operations of distribution networks. It is also
wire by formulating a power flow for four-wire multi-ground acknowledged that vast diversity in distribution networks leads
low-voltage networks. Hourly analysis is generally performed to their unique response when subjected to increasing DG
over a period of one year to assess the impact of high PV pen- penetration. Diversity in distribution networks is reflected by
etration on steady state operations with concerns such as peak many factors, such as load compositions leading to unique load
demand, over/under voltages and losses; on the other hand, profiles, typical network sizes and topologies, load density, and
high resolution data in quasi-static simulations can be used phase-wise uneven distribution of load. Feeder length, feeder
to undertake studies related to voltage flickers/fluctuations, impedance, electrical load at substation node, unbalanced node
control of voltage regulating equipment, and a range of other voltages are also some of the parameters that can partly and
issues [7]. inadequately quantify either the feeder-specific factors or their
Effects of factors such as size, type, and location of PVDG effects. These feeder-specific factors are mostly qualitative,
in active distribution networks (ADNs) have been thoroughly and therefore, it is difficult to explicate their effects by
analyzed. However, the focus of these attempts has been feeder functional analysis, wherein the relationships among variables
voltage profile and voltage regulating equipment only. Effects are defined by functions and sub-functions until they reach
of PV penetration on network losses, unbalanced voltages, a level of solvability. In this context, it is imperative and
maximum demand, TCOs of voltage regulating equipment, prudent to use exhaustive analysis with multiple networks and
power factor at substation node are all factors that have gained scenarios created deliberately to represent these factors so as to
less attention. It should also be noted that the unbalance of capture their effects and role in daily operations management.
distribution networks is not always represented in sufficient While analysis with worst case scenarios and single line
details in these studies, whereas the effect of load unbalance equivalent networks can be used effectively for system re-
on line losses [12] and node voltage unbalance [13] need to liability and security studies for medium to high voltage
be accounted for. While the uniqueness of feeder response to networks, complete disregard of feeder-specific factors in low-
the increasing PV penetration is generally anticipated and ac- voltage networks and their interplay with intermittent nature of
knowledged, there has been no elaborate attempt to analyze the generation from DERs can lead to grossly ineffective solutions
effect of feeder-specific factors such as geo-electric size and for optimization of daily operations. It is therefore important
spread of the feeder, topological and temporal phase unbal- to choose an analysis approach that encompasses the above
ance, and load density. Urban, semi-urban, and rural networks two aspects, namely, effect of feeder-specific factors and slow
greatly differ in their size, spread, topological unbalance, load dynamics of DG over a long period. The objective of this
unbalancing, and load density. Clustering or grouping of these exhaustive exercise is thus restricted to steady state analysis
various distribution feeders is required in planning exercises of distribution networks that takes into account the diurnal and
to formulate strategies for mitigating the unwanted effect of seasonal cyclic variations (slow dynamics) caused by PVDGs
increasing DERs. It is, therefore, important to know the effect and assess the effect of feeder-specific factors in the feeder’s
of feeder-specific factors on feeder response characteristics to response. While analysis of multiple networks outlines the
increasing DERs [4], before attempting classification of these response of different networks to a similar stimulation, this
feeders. analysis is also aimed at investigating the response of similar
feeders (same geo-electric size but different load density and
This paper’s contribution is as follows: 1) analyze the effect phase-unbalancing) to increasing penetration of PVDGs. It
of feeder-specific factors, namely, geo-electric size and spread therefore explores the case-specificity argument for different
of the feeder, phase unbalancing, and load density on feeder as well as similar feeders.
response characteristic to increasing PV penetration, 2) argue
case-specificity not only among the dissimilar feeders but III. T HE S OLUTION A PPROACH
also among the similar feeders’ response when subjected to
high PV penetration, and 3) provide implications of case- In order to capture the effects of PV intermittency and
specificity that emphasize the need for due consideration of variability—both diurnal and seasonal—annual time series
feeder-specific factors in formulating control and optimization analysis is preferred, as suggested and applied in [3]. The
strategies for daily operations management. sequential time simulations (STS) for a data set of one year—
including recorded annual load profiles and PV generation,
Section II sets the problem in the appropriate context and are performed with an unbalanced 3-Φ power flow program.
argues legitimacy of the approach used for analysis. This is fol- The distribution power flow for unbalanced radial distribution
lowed by Section III with the solution approach that includes systems is found in [14], [15]. A backward forward sweep
creation of feeder variants and a sequential time simulations (BFS) algorithm based on [14] is implemented in [16] using
(STS) framework. Section IV presents the simulation results MATLABTM . This implementation uses detailed 3-Φ modeling
with discussions followed by Section V with summary and of all the components from [17], except the model of distri-
implications of high degree of case-specificity. Concluding bution transformers and PV arrays. It is worth a mention here
remarks are included in Section VI. that GridLAB-D implementation is not used in this study for
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 103

two reasons: 1) Modeling of VRs is not done to handle open- created for each feeder to represent an alternative load density
delta connections in non-manual modes. This is particularly whereas case C and D reflect different local phase unbalancing
important for IEEE 37 node feeder, which has an open delta in the same feeder. These feeder variants can effectively
connected VR. 2) Modeling of PV plant in GridLAB-D is represent the network with load centers such as recreational
as of now only based on solar radiation and the gross area of centers or academic campuses in which seasonal changes in
reception that can cumulatively increase the error in electricity demand are quite large. Case E is created to represent a near-
production to unacceptable levels in time-series power flow similar topology—but not the same topology—of a feeder by
studies. This error gets reflected in aggregate energy losses, disconnecting a part of the feeder. Aspect 3) is incorporated
TCOs and occurrences of node voltage limit violations. On by having four feeders with different geo-electric size.
the other hand, the MATLABTM implementation has been Thus, six feeder variants, including five cases and one base-
validated against the published base-case results of IEEE case feeder configuration, are analyzed for each of the three
test feeders [18], for modeling of VR. Models for PVDGs IEEE test feeders (IEEE 13 node, IEEE 34 node and IEEE 37
are validated with published results in [19] as well as with node) with six levels of PV penetration. This is supplemented
field tests carried out on actual PV plants in IITGN-VGEC. by an analysis of an actual low-voltage feeder with a rooftop
Numbered blocks from 1 to 7 in Fig. 1 provide all necessary PV plant, called IITGN-VGEC feeder, in India with different
data and mathematical modeling for executing the BFS power PV penetration levels.
flow for 3-Φ unbalanced networks. Data of solar irradiance It should be noted that the following conditions are strictly
(W/m2 ), ambient temperature (◦ C) and wind speed (m/s) are adhered to while creating the above case study scenarios:
collected from [20]. These data are forwarded to block 2 to 1) Only local phase-unbalancing is changed, which does not
model PV arrays. Blocks 1 and 2 are used for modeling the alter overall phase-unbalancing at the substation node. This
PV generation plants for IEEE networks. However, recorded is to make sure that the resulting feeder variant is otherwise
measurements of PV generation data is available for the Indian same as the base-case feeder except the change in local
network and hence is directly used from block 3. Similarly, phase-unbalancing (overall change in phase-unbalancing is
block 4 has metered real power demand data, which is directly anyway different in different feeders). 2) Demand at substation
used, whereas block 5 has a derived set of load profiles node remains unchanged while altering the location of major
obtained from [21]. load centers in order to capture only the effect of change
in load density rather than load magnitude itself. 3) Any
Solar insolation Recorded Metered Derived load
& weather data measurements measurements profiles
change in load density does not exceed the rated limits of
1 3 4 5
associated components such as transformers and conductors
Modeling of PV generation Real & reactive making sure the feasibility of such variations. 4) Real power
PV arrays data power demand injection from PV plants should not exceed the rated limits
2

6
of sections through which additional power flows to account
Mathematical models STS approach
for the branch current limits. 5) Ratio of real to reactive
7
for cables, O/H lines,
transformers, voltage
based on BFS Standard power demand remains unaltered to avoid inadvertent change
algorithm for 3-Φ load models
regulators, shunt unbalanced ADNs in reactive power flows in the networks. These conditions
capacitors etc.
ascertain that each case exclusively corresponds to only one
feeder-specific factor and that the cases are mutually exclusive
and feasible for a given feeder. Thus, the argument of case-
Energy Monthly average Effect of Voltage regulator
specificity and its implications are based on an exhaustive
losses power factor & daily capacitors switching annual analysis of 132 scenarios performed by:
maximum demand
1) Creating 6 feeder variants for each of the 3 IEEE test
Fig. 1. Overview of STS approach. feeders and one actual feeder in India;
2) using detailed phase-domain modeling for each feeder
STS is performed for a total of 132 scenarios. These with all its components including PVDGs;
scenarios are created to reflect three aspects of analysis: 3) performing yearly simulations at 15-minutes interval
1) effect of increasing PV penetration on feeder response with recorded annual load profiles (industrial, commer-
characteristics, 2) effect of feeder-specific factors such as load cial, and residential load profiles at different nodes) for
density and phase-unbalancing for a given base-case feeder each case using BFS algorithm;
configuration, and 3) comparison of feeder response character- 4) analyzing 132 case studies based on 4 feeders to exclu-
istics for distribution feeders with different geo-electric size. sively represent each feeder-specific factor with different
For aspect 1), i.e., the effect of increasing PV penetration, cases;
six levels of PV penetration—from 0% to 100% of maximum 5) summarizing the effects on energy losses, feeder voltage
demand (at substation node), in steps of 20%—are considered. profile, TCOs of voltage regulating equipment, power
In order to incorporate aspect 2) of this analysis, variants of factor at substation node, maximum real power demand
each feeder are created by deliberately shifting locations of and hosting capacity of the feeder;
major load centers either in all the three phases simultaneously 6) drawing inferences from the response characteristics
(to represent different load density) or in one or two phases of feeders about the implications of these studies in
(to represent local phase unbalancing). Case A and B are planning exercise for counter-actions.
104 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

TABLE II
A. Feeders and Variants VARIANTS OF IEEE 13 N ODE F EEDER

IEEE test feeders, namely IEEE 13 node, IEEE 34 node, Phase A Phase B Phase C
Case Node
and IEEE 37 node test feeders [16], along with an actual kW kVAr kW kVAr kW kVAr
low voltage distribution network, called IITGN-VGEC net- 634 60 41.25 20 15 20 15
Case A 671 302 173 351 200 402 230
work [8] are selected for analysis. The geo-electric size and 680 200 125 200 275 200 133
spread of these networks is different in each case along 633 201 115 130 75 161 92
with their unique phase-unbalanced topology. Table I gives Case B 646 0 0 325 187 0 0
671 201 115 225 129 251 144
a comparison of these feeders in term of their feeder-specific 633 0 0 400 228 0 0
factors. Phase-wise reach in Table I refers to the topological Case C
671 402 230 51 30 502 288
phase-unbalancing. Base-case load profiles of these networks 633 400 157 400 228 0 0
Case D 671 402 230 51 30 502 288
are provided in the Appendix (Table AI, Table AII, and 675 85 33.3 68 60 290 212
Table AIII, respectively, for IEEE 13, 34, and 37 node test Case E Switch between nodes 671 and 692 opened
feeders) whereas the details of changes made to these base-
case network profiles to create feeder variants are listed in TABLE III
Tables II, III, and IV, respectively, for IEEE 13, 34, and 37 VARIANTS OF IEEE 34 N ODE F EEDER
node feeders. Modified IEEE test feeders are shown in Fig. 6, Phase A Phase B Phase C
Fig. 7, and Fig. 8 in the Appendix, respectively, for IEEE 13 Case Node
kW kVAr kW kVAr kW kVAr
node, 34 node, and 37 node feeders. IITGN-VGEC network Case A
830 142 63 135 68 150 73
is given in Fig. 2. 890 25 12 25 12 25 12
828 125 95 125 97 129 99
830 142 63 135 68 150 73
TABLE I Case B
844 19 15 10 8 10 8
C OMPARISON OF F EEDER - SPECIFIC FACTORS 890 25 12 25 12 25 12
844 19 15 135 105 135 105
Case C
Trunk Max. Demand Phase- 890 25 12 150 75 150 75
Load Load
Test Feeder Feeder at Source wise 808 0 0 16 8 250 160
Density Unbalancing
(km) Node (MW) Reach Case D 844 19 15 135 105 10 8
1-phase, 890 25 12 150 75 25 12
IEEE 13 1.52 3.5 2-phase, High Low Case E Feeder from node 834 disconnected
3-phase
1-phase, TABLE IV
IEEE 34 93.34 2.04 2-phase, Low Low VARIANTS OF IEEE 37 N ODE F EEDER
3-phase
IEEE 37 5.49 2.5 3-phase Moderate High
IITGN-VGEC 0.37 0.3 3-phase Low Moderate Phase A Phase B Phase C
Case Node
kW kVAr kW kVAr kW kVAr
701 15 7 15 7 225 112
Case A
728 167 85 167 85 167 85
701 15 7 15 7 225 112
Case B
775 125 63 125 63 125 63
3-phase overhead Δ Δ Distribution 701 140 70 140 71 175 88
line from utility transformers Case C
741 0 0 0 0 217 108
11,000/415 V
250 kVA each 701 140 70 140 70 0 0
Source node-1 Case D 741 0 0 0 0 217 108
775 0 0 0 0 175 88
Case E Feeder from node 737 disconnected

E-3 F-4 G-5 SL-6 S3-7


Pumpus 3-phase
capacitor
DB1-2 DB2-8 bank B. Modeling and Data Sets
All the major components such as transformers, VRs, shunt
capacitors, PV plants, underground cables, and overhead lines,
B-10 C-11 D-12 DH-14 J-16
and loads are modeled as 3-Φ components. Modeling of
RPV 1
RPV-1
conductors, cables and lines, and loads is based on [17]. A
A-9 S1-13 unified model for distribution transformers is used as proposed
H-15
in [22]. The model validation was done with IEEE 4 node test
S2-18 I-19
K-17
feeder for transformer models with different configurations.
PV array modeling was adopted from [19] and validated with
GH-21 the field tests for IITGN-VGEC PV plants. Modeling details
L-20
RPV-3
RPV 3 of IITGN-VGEC network and load compositions can be found
Spot load at
RPV-22
RPV in [8].
X: Premise name
X-Y Y: Node number
Annual load profiles for IEEE test feeders were used from
publically available datasets listed in [21]. Six distinct load
Fig. 2. IITGN-VGEC feeder with 3 roof-top PVDGs.
profiles were classified as industrial, commercial, and residen-
tial load profiles. These were scaled and tied to each node
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 105

in IEEE test feeders. The large and balanced load centers a consideration for conductors’ capacity to draw the PV power;
were populated with industrial load profiles, whereas relatively 2) real power injection from a single PVDG is phase-balanced;
small and unbalanced loads were populated with commercial and 3) values in percent are derived by comparing it with the
or residential load profiles. value for the corresponding case without any PV penetration,
In any case, the scaling of the load profiles ensures that e.g., percentage rise in energy losses in case C for any feeder
the varying loads at any node never exceed the base-case load is obtained by comparing it with the annual energy loss value
value specified in standard test feeders. This also ensured that of case C without PV penetration in that feeder.
the variation in real to reactive power ratio remained within
± 5% of the specified ratio at each node in IEEE test feeders. A. Feeder Voltage Profile
All these steps were incorporated to ascertain that only the
The feeder voltage range specified in ANSI C84.1 [23]
intended factors affect the feeder response characteristics.
is used as the benchmark for enumerating maximum and
minimum node voltage violations in the IEEE test feeders.
IV. S IMULATION R ESULTS AND D ISCUSSION The range-A in ANSI C84.1 specifies the feeder voltage
The STS results were exhaustive and needed to be collected range as ± 5% of nominal voltage to be maintained by the
and aggregated carefully to make meaningful inferences. The utility supplying the electricity. Apart from the fact that the
effect of feeder-specific factors on feeder response charac- distribution system operator (DSO) is obliged to respect this
teristics to increasing PV penetration is shown by grouping voltage range, the effect of increasing PV penetration on node
the results in two ways: 1) performance parameter of interest voltage limit violations is also important in estimating the
for different feeders e.g., voltage profile of different feeders, feeders’ hosting capacity for DERs. Minimum and maximum
and 2) response of feeder and its variants for the chosen node voltages observed in the IEEE 13 node feeder’s voltage
performance parameter, e.g., losses for feeder variants. This profile in the base-case annual STS results are shown in Fig.
is so done to bring out the variation in response of the same 3(a). Maximum node voltages are quite close to the upper limit
feeder to change in feeder-specific factors. Comparisons are of 1.05 p.u. However, it is difficult to make out the occurrences
also drawn between the response characteristics of different of voltage limit violations from Fig. 3(a).
feeders. Results for feeder voltage profile, energy losses, power Therefore, node voltage limit violations are shown for all
factor and maximum demand at source node, and effect on the variants of IEEE 13 node feeder in Fig. 3(b) for increasing
TCOs of VRs and their interaction with shunt capacitors are PV penetration. The maximum node voltage limit is violated
collected. The following points should be considered while even in the base-case (without PV penetration) for IEEE 13
interpreting the results described in this section: 1) A single node feeder with B-phase voltage of 1.055 p.u at node 675.
PVDG (location of PVDG is indicated in the Appendix) is The reason for this high node voltage is phase-balanced shunt
connected at a node away from source node in each case with compensation at node 675 with lightly loaded phase B at rear

20,000
Min/Max Feeder Voltage (p.u.)

Voltage Limit Violations

1.06 Base-case A B C D E
1.05
1.04
1.02
Minimum feeder voltage (p.u.) 10,000
1 Maximum feeder voltage (p.u.) 4,428
4,730
0.98
0.96
0.95 0
0.94 0 20 40 60 80 100
Jan. Feb. Mar. Apr. May Jun. Jul. Aug. Sep. Oct. Nov. Dec. PV Penetration (%)
(a) (b)

16,000 6,000
Base-case AA B CB D CE
Voltage Limit Violations

Base-case A B C D E Base-case D
Voltage Limit Violations

12,000
4,000
8,000
2,000
4,000

0 0
0 20 40 60 80 100 0 20 40 60 80 100
PV Penetration (%) PV Penetration (%)
(c) (d)

Fig. 3. (a) Min-Max node voltages in IEEE 13 node base-case. (b) Maximum node voltage violations in variants of IEEE 13 node feeder. (c) Maximum
node voltage violations in variants of IEEE 34 node feeder. (d) Minimum node voltage violations in variants of IEEE 34 node feeder.
106 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

end of feeder (Table AI). More than half of the load in phase B volatility of demand and generation, power loss minimization
is connected to nodes 634, 645, and 646. With change in load is better replaced by energy loss minimization. In this section,
density in case A and B, wherein the loads are shifted towards responses of feeder variants to increasing PV penetration are
source node (Table II), the reactive power demand reduces at summarized for energy losses.
node 675. As a result, the node voltages increase to cause a These results are presented in absolute terms with annual
reactive power flow from node 675 towards nodes 671, 633, energy loss values as well as in relative terms to capture
646, resulting in high voltage limit violations. With increase the trends in variations. Aggregated annual energy losses and
in local real power injection from PVDG at node 680, real trends in variation of energy losses for IEEE feeders and their
power flow from source to the rear node reduces, resulting variants are shown in Fig. 4. Energy losses (in %) in Fig.
in reduced potential drop across trunk branch 632-671 and 4 are based on the value of annual energy loss with no real
hence, reduced voltage limit violations. This implies that the power injection from PVDG in the corresponding case. All the
maximum node voltage violations are primarily occurring due variants of IEEE 13 node feeder exhibit more or less similar
to local phase-unbalancing and that any steps taken to mitigate decreasing trend in energy loss with increasing PV penetration,
these violations would not be as effective as phase-unbalanced as seen in Fig. 4(a). Small variations in this trend are observed
real or reactive power injections (by community energy storage in case E wherein the network has reduced total load with
or shunt capacitors, respectively). It is observed that the phase- removal of part of the feeder. However, considerable reduction
balanced real power injection from PVDG can only partially in loss is observed in Fig. 4(b) for this case. This is due to
mitigate the problem of node voltage limit violations, Fig. the fact that the substation has to feed less load, and more so
3(b). It is also interesting to note that maximum node voltage because of the local PV generation. The increasing trend in
limit violations have altogether vanished in cases C, D, and E case E is due to the fact that local PV generation exceeds the
because case E removes nodes 692 and 675 from the network, network demand and reverse power flow is observed from the
and case C and D have increased local reactive power demand rear end towards source node during peak PV generation. This,
near shunt capacitors. however, is not the case with other network variants as PV
Quite the opposite trend in IEEE 34 node feeder is observed generation does not exceed total network demand. The highly
in Fig. 3(c) wherein the voltage limit violations increase with loaded IEEE 13 node feeder shows marginal change in energy
increasing PV penetration, especially in case D. This can be loss characteristics with change in major load centers or even
explained as follows: IEEE 34 node has major load centers with the local unbalancing. Not more than 10% of change
towards the rear end of the feeder with shunt capacitors at in annual energy losses is observed among all the network
nodes 844, and 848. Reverse power flow across VR-2 (node variants. But, it should be noted that 5% to 10% improvement
852 [16]) gets established only after more than 50% of PV in annual energy loss with phase-unbalanced real or reactive
penetration at node 890 and base-case configuration starts power injection is worth the effort as indicated by Fig. 4(b).
experiencing voltage limit violations that keep on increasing The difference between losses in case D and case A is caused
with increasing PV penetration. However, case D has a major by local phase unbalancing and case D losses can be reduced
load center shifted to node 808, which is near source node. by unbalanced real power injection from PVDGs.
In this case, high reactive power flows from the rear end Energy loss characteristics of the IEEE 34 node feeder
towards node 808, resulting in voltage limit violations near and its variants are quite different and more sensitive to
shunt capacitors. With increasing real power injection from change in load density as well as local phase-unbalancing. It is
PVDG at the rear node, real power also starts flowing in a observed in Fig. 4(c) that the base-case energy loss continues
reverse direction, worsening the situation even further. The low to decrease down to −22% of its energy losses without PV
voltage limit violations caused by high loads in the C-phase penetration. However, it should also be observed that this base-
remains almost unaffected by increasing PV penetration as case configuration has the highest losses among all its variants,
balanced injection does not alter the unbalancing (Fig. 3(d)). from Fig. 4(d). A shift in load density causes a reduction of
Table V lists the highest unbalanced node voltages—IEEE 34 more than 58% (1.753 GWh to 1.026 GWh) of annual energy
node feeder has the highest among all (case C, D). IEEE 37 loss in the absence of PVDG. While base-case configuration
node feeder has a very healthy voltage profile [24], and can appears to be well behaved with increasing PV penetration,
accommodate 100% PVDG without voltage limit violations. it is in fact the case B variant that has minimum annual
energy losses (Fig. 4(d)). It should be noted that the major
TABLE V
M AXIMUM U NBALANCE A MONG P HASE - VOLTAGES load centers of IEEE 34 node test feeder are shifted from
rear to near substation node. This results in considerably less
Maximum Unbalance Voltage (%) energy losses as the feeder is quite long. It also explains why
Test Feeders
Base-case A B C D E
IEEE 13 1.33 1.50 1.41 1.02 1.26 0.58
energy losses start increasing in this feeder after a minima
IEEE 34 0.73 0.60 0.64 2.36 1.73 0.62 and with increasing PVDG. Excess power from local PVDG
IEEE 37 0.92 0.94 0.96 1.18 1.18 0.92 starts flowing towards far ends in networks and that results in
increasing losses.
All of the above point to the fact that apparently similar
B. Energy Loss Characteristics feeders may have altogether different energy loss values and
Power loss minimization is one of the most important ob- loss characteristics in response to increasing PV penetration.
jectives in distribution network optimization. With increasing It then becomes imperative to update optimization objectives
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 107

PV Penetration (%)
0.8

Annual Energy Loss (GWh)


0% Base-case A B C D E
0 20 40 60 80 100
Annual Energy Loss 0.6

10%
0.4

0.2
20%
Base-case A
B C 0.0
0 20 40 60 80 100
D E
30% PV Penetration (%)
(a) (b)

20% 2.0

Annual Energy Loss (GWh)


Base-case A B 1.75 Base-case A B C D E
C D E
Annual Energy Loss

1.5 1.38
PV Penetration (%)

1.03

0.93
0%
0 20 40 60 80 100 1.0

20% 0.5

0.0
0 20 40 60 80 100
40% PV Penetration (%)
(c) (d)

PV Penetration (%) 0.5


Annual Energy Loss (GWh)

5% Base-case A B C D E
0.40 0.37
0.4
0.34 0.32 0.30
Annual Energy Loss

0 20 40 60 80 100 0.30 0.28 0.26 0.31


5% 0.3 0.25
0.24 0.24
0.2

15% 0.1
Base-case A
B C 0.0
0 20 40 60 80 100
D E
25% PV Pentration (%)
(e) (f)

Fig. 4. (a), (b), (c) Trends in energy losses. (d), (e), (f) Energy loss values for IEEE 13 node, 34 node and 37 node test feeders, respectively.

to exploit the change in feeder response characteristics during an obligation of the bulk power consumers in many countries
different time of year in similar feeders with different load to maintain the power factor within a stipulated range. Penalty
density. for poor power factor and incentives for maintaining healthy
Loss characteristics of IEEE 37 node feeder and its variants power factor are prevalent especially for spot networks. Real
in Fig. 4(e) and (f) appear to be similar to that of IEEE 13 power injection from DGs affects the balance of real and
node feeder, except that they are more sensitive to change in reactive power demands from the source end, which is in turn
feeder-specific factors as is noted from absolute values. This is reflected in the average power factor.
attributed to the fact that the feeder has high load unbalancing
The STS results of base-case feeder configurations are only
i.e., its demand in phase C (1.09 MW) is quite high as
included for comparison as the derived variants of feeders
compared to phase A (0.73 MW) and phase B (0.64 MW)
are created based on condition of keeping the real to reactive
in base-case configuration. Although small, the IITGN-VGEC
power ratio unchanged at source node. However, two sets of
network’s loss characteristics also show an increasing trend
STS simulations—with and without shunt compensation—are
after a minimum loss point as shown in Fig. 5(a) and (b). The
performed with increasing PV penetration for all the feeders.
reason for this trend is reverse power flow with increasing PV
Monthly average power factor at substation node for IEEE 13
penetration. This especially happens when the PV generation
node feeder is shown in Fig. 5(c). With real power injection
peaks with off-peak intervals in local demand.
to the tune of 60% from PV plant, the real power demand
dips with reactive power demand unchanged at source node.
C. Average Power Factor and Peak Demand This leads to poor power factor during sunshine hours—
For maintaining a healthy power factor at source node, the more so with longer sunshine days of the year—resulting
ratio of reactive to active power needs to be kept under 0.5. It is in lower average power factor in the months of April, May
108 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

8.60
1.00
IITGN-VGEC with caps 8.452

Average Power Factor (Lagging)


8.40
IITGN-VGEC without caps 0.95
Energy Loss (MWh)

8.20
8.373 0.90
7.954 7.981
8.00
7.902 0.85
7.726 7.737
7.80
7.875 0.80
7.60
7.647 7.658 0.75
7.40 No PV 20% PV 60% PV
0 20 40 60 80 0.70
PV Penetration (%) Jan. Feb. Mar.Apr. May Jun. Jul. Aug. Sep. Oct. Nov.Dec.

(a) (b)
1 1 1
No PV no caps No PV with caps
Average Power Factor (Lagging)

60% PV no caps 60% PV with caps

Average Power Factor (Leading)


Average Power Factor (Lagging)
0.95
0.95 0.95
0.9

0.85 0.9 0.9

0.8
0.85 No PV no caps 0.85
0.75 No PV with caps
60% PV no caps
60% PV with caps (leading)
0.7 0.8 0.8
Jan. Mar. May Jul. Sep. Nov. Jan. Mar. May Jul. Sep. Nov.
(c) (d)

60,000
57,767
Base-case A B C D E 56%
TCOs with Additional Shunt

Without caps 46%


45,000 45% 44%
Number of TCOs

With caps
Compensation

34% 34%
Additional caps 33% 27%
30,000
23%
22%
15,000 14% 14% 16%
7%
7% 5%
0% 2%
0
0 20 40 60 80 100 0 20 40 60 80 100
PV Peneration (%) PV Penetration (%)
(e) (f)

Fig. 5. (a) Energy loss characteristic for IITGN-VGEC network. Monthly average power factor for IITGN-VGEC network, IEEE 13 node and IEEE 34 node
feeder in (b), (c) and (d), respectively. (e) and (f) show TCO trends in IEEE 34 node network and with additional compensation in IEEE 13 node network,
respectively.

and June. Even with the shunt compensation, the average node voltage limit violations in base-case start from 80% (and
power factor remains below 0.9 for over six months. Increasing higher) PV penetration. This situation becomes all the more
PV penetration worsens this situation. Whereas in IEEE 34 prevalent in case D of this feeder due to rising downstream
node feeder, the downstream node voltages start rising with node voltages. IEEE 37 node feeder does not have any shunt
increasing real power injection from PVDG. This results in compensation and it shows the variation in power factor
increased reactive power injections from two capacitor banks similar to that exhibited by IEEE 13 node feeder. IITGN-
situated at nodes 844 and 846 during sunshine hours. Thus, VGEC network’s power factor variation shows a dip during
the average power factor is observed (in Fig. 5(d)) to become winter season in India. Fig. 5(b) shows the monthly average
leading with 60% PV penetration with all capacitor banks power factor for IITGN-VGEC network with capacitor banks
switched on. Without these capacitor banks, the substation connected at the source node. Unlike IEEE 34 node test feeder,
will notice a lagging power factor less than 0.9 for almost the capacitor banks are not affected by marginal voltage rise
throughout the year. It can be recollected from Fig. 3(c) that with increasing PV penetration as these are situated right next
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 109

to the supply transformers at source node. node 646 for phase B (Fig. 6). Increased TCOs in this feeder
Maximum demand at source node is another important is primarily because of the high reactive power flowing from
concern of DSO. However, it is the coincidence of time of node 675 to node 646, refer to Table VII. Similarly, in case
peak demand and peak generation from PVDG that determines of IEEE 34 node feeder, the reverse power flow—both real
the impact of increasing PV penetration on peak demand. and reactive—consistently increases TCOs with increasing PV
Reduction in peak demand of about 15 to 20% is observed in power. With provision for phase-unbalanced reactive power
IITGN-VGEC and IEEE 34 node network. Whereas the IEEE injection, increased TCOs can be substantially reduced which
13 node and 37 node feeder show minimal effect of PVDG on leads to overall reduction in TCOs.
peak demand. The institutional load profile coincides well with
TABLE VI
the PV peak generation and hence maximum peak shaving TCO S IN IEEE 13 N ODE F EEDER AND I TS VARIANTS
is observed in IITGN-VGEC network. However, IEEE 34
node network has its load profile made up of residential and PV Penetration
Base-case A B C D E
(%)
commercial loads that make it more uneven and peaky as 0 12,137 14,085 11,291 12,713 12,056 8,235
compared to the rest of the two IEEE feeders. 20 12,893 14,805 12,011 13,145 12,272 9,270
40 13,658 15,336 13,028 13,991 13,325 10,890
60 14,945 16,389 14,450 14,846 14,387 12,060
D. Trends in TCOs 80 16,331 17,505 15,953 16,160 15,854 13,689
100 17,654 18,540 17,096 17,402 17,006 15,084
Voltage regulating equipment such as VRs, LDCs, and
OLTCs, are mechanically operated devices that need periodic
maintenance based on the number of tap changing operations TABLE VII
they are subjected to. In order to minimize the occurrences of TCO S IN IEEE 37 N ODE F EEDER AND I TS VARIANTS
TCOs and thereby maximize equipment life-time, efforts are PV Penetration
made to optimize network operations so as to cause minimum Base-case A B C D E
(%)
wear and tear to such devices. Annual aggregate TCOs are 0 3,110 3,222 3,173 2,943 3,006 2,079
20 4,361 4,167 4,127 4,149 4,113 2,871
expected to increase with increase in real power injection 40 5,369 5,553 5,360 5,031 5,049 4,581
from PVDGs as diurnal variability of PV generation causes 60 6,800 6,885 6,782 6,471 6,453 6,534
additional TCOs. 80 7,673 7,776 7,862 7,371 7,290 7,812
100 9,050 9,153 9,167 8,631 8,703 9,522
This is also observed from the results listed in Table
VI and VII for IEEE 13 node and IEEE 37 node feeder,
respectively. It is observed that TCOs are more sensitive to TABLE VIII
changes in network topology than load density and local P HASE - WISE TCO S IN IEEE 13 N ODE BASE - CASE
phase unbalancing. But, effect of these factors—load density
and local phase unbalancing—are also considerable in two PV Penetration With Caps Without Caps
(%) Ph. A Ph. B Ph. C Ph. A Ph. B Ph. C
cases. First, the IEEE 34 node test feeder shows over 55,000 0 4,093 3,199 4,393 4,101 2,613 4,425
operations in its base-case with no PV penetration whereas its 20 4,237 3,675 4,607 4,431 2,691 4,583
variants with change in load density and phase-unbalancing 40 4,545 3,873 4,927 4,747 3,057 4,931
60 4,931 4,439 5,171 5,183 3,361 5,315
show less than 30,000 operations (refer to Fig. 5(e)). Second, 80 5,459 4,865 5,561 5,543 3,623 5,523
phase-wise analysis of IEEE 13 node feeder’s TCOs in its 100 5,833 5,213 6,043 6,015 4,029 5,929
base-case, given in Table VIII, shows that the rise in TCOs
with shunt compensation is only due to increased TCOs in
one phase. But this rise in one phase’s TCOs is dominant
V. S UMMARY AND I MPLICATIONS
and results in overall increase in TCOs as compared to
that without shunt compensation. This can lead to a wrong Yearly simulations at 15 minute-intervals capture the diurnal
conclusion that the shunt compensation can cause high TCOs variations as well as seasonal fluctuations in both generation
with increasing PV penetration. Whereas, additional phase- and demand. However, the results for TCOs are conservative
unbalance shunt compensation can bring the TCOs down to with 15-minutes intervals as the VR can respond to changes,
very low value. This is shown in Fig. 5(f). Improvement which are faster than averaged 15-minutes intervals, such as
brought about by additional shunt compensation in one phase cloud movements causing change in PVDG power resulting
is consistently observed in all the feeder variants of IEEE in voltage fluctuations. But it should be noted that the present
13 node feeder. It can be explained as follows: Variability exercise is aimed at capturing effect of feeder-specific factors
of real power injection causes increased TCOs in all networks and for this purpose, simulations with 15-minute intervals are
normally. Shunt compensation does no good in reducing TCOs adequate. It implies that STS with high resolution data will
as phase-balanced reactive power injection cannot mitigate the capture the same trends in TCOs albeit with higher values of
voltage variations already present due to phase-unbalanced TCOs. Effects of each feeder-specific factor on the operational
load. However, if reactive power is injected at the right place parameters are summarized as follows.
and phase, increased TCOs due to PVDG can altogether be 1) Geo-electric Size of Feeder
compensated as shown in Fig. 5(f). It is made possible in Small feeders with PVDGs can have the benefit of improved
IEEE 13 node feeder by providing local reactive power at voltage profile with considerable reduction in energy losses;
110 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

whereas, substantial saving in energy losses can be obtained is observed in IEEE 34 node network and its variants that
in long and sparse feeders with an optimum level of DG they remain unaffected by change in load density (voltage
penetration. However, the same benefit is lost with excessive profile, TCOs) whereas phase unbalancing severely affects the
reverse power flow due to local generation exceeding the voltage profile. Both minimum and maximum node voltage
network demand. It is also noted that the long feeders exhibit limit violations considerably increase (Fig. 3(c), (d)) in case D.
more complex response characteristics in terms of feeder This implies severe limitation on hosting capacity of the feeder
voltage profile, TCOs and power factor. This is attributed to for DGs. On the other hand, energy losses are also sensitive to
highly case-specific interaction between shunt compensation, phase unbalancing and show considerable reduction in cases
VRs and diurnal variability of PVDGs. It is observed that C and D in IEEE 34 node network (Fig. 4(d)).
the long feeder with a VR away from a source node can It is also deduced from the exhaustive results that a dense
have adverse effect on voltage profile and reactive power flow. feeder such as the IEEE 13 node network remains unaffected
One such instance is explained here: IEEE 34 node feeder’s (by phase unbalancing) in all its operational parameters such
VR at rear end will try to raise the tap position while the as voltage profile, losses, and TCOs. IEEE 37 node network,
real power flow increases from end of feeder towards source which is a moderately loaded, and highly unbalanced feeder
node. Sensing only the magnitude of the current—and not (C-phase is highly loaded as compared to other phases),
its direction—it tries to maintain the remote node voltage by shows no ill-effects of phase unbalancing. However, it does
compensating line drop. This leads to higher voltages at the register improved losses and TCOs with reduction in phase
remote node which is monitored by VR. This rise in voltages unbalancing.
accompanied by reduced real power demand of the network Another important takeaway of the STS analysis with regard
and increased reactive power injection (due to increased node to phase unbalancing is that phasing and placement of shunt
voltages) causes leading power factor even at the source node. capacitors should be revised for UDNs with high DG penetra-
This can typically happen in long feeders with VR which is tion. This can be a very cost-effective and efficient solution for
used to maintain downstream rear node voltages. In small reducing TCOs and keeping the feeder voltage profile within
feeders, the VRs are either not required or they are placed stipulated range. This is demonstrated by Fig. 5(f). It can also
near source end which does not result in the adverse effects be crucial in reducing energy losses.
discussed earlier. At the risk of generalizing the case-specific attributes of
2) Load Density feeder response characteristics, it can be said that the long
Change in load density—which might be caused by seasonal and sparse feeders are more sensitive to the changes in
variation in load demand from major load centers—establishes phase-unbalancing while small dense feeders are prone to the
new equilibrium of power flows and voltage profile. If the changes in load density. Each feeder stands unique with its
feeder is already dense, any change in load density conspicu- geo-electric size, load density and phase unbalancing. Thus,
ously responds in the feeder’s voltage profile and TCOs. This feeder-specific factors produce a unique composition of effects
is typically observed in IEEE 13 node feeder. Altered load discussed so far on all operational parameters.
density in case B causes very high numbers of maximum node Therefore, any exercise of optimizing daily operations
voltage limit violations (Fig. 3(b), case B) whereas no such of “similar” or different feeders without accounting for
adverse effect is noticeable in variants of IEEE 34 node (Fig. the feeder-specific factors would not be efficient and cost-
3(c)) and IEEE 37 node feeder. TCOs also expectedly increase effective. Seasonal variation in load density and local phase-
in IEEE 13 node feeder due to violent effects on voltage unbalancing (e.g. due to seasonal recreational centers or
profile. Feeder voltage profile is one of the prime concern academic institutions) can cause benefits in one aspect and
in determining hosting capacity for DGs. These results have raise concerns for another. Thus, the approach that takes into
direct implications of reduced DG hosting capacity. account the effects of these feeder-specific factors, and cyclic
Shift in load density brings noticeable change in energy variability of generation and load profile can be the ideal
losses in long feeders like IEEE 34 node feeder. Considerable means for classification exercise. However, such exhaustive
reduction in losses is observed (Fig. 4(d), case A, B) as analysis of thousands of feeders may not be feasible. In that
the shift in load centers can cause change in length covered case, carefully selected samples of daily load and generation
by branch currents. It should be noted carefully that this profiles can be used to run the STS for many feeders be-
change could be bi-directional (increase or decrease). But, it is longing to the same geographic region. These results can then
plausible to state that adverse effect of change in load density be analyzed to identify different responses and classify the
can be expected in dense networks whereas sparse networks feeders based on allowable range of variation in operational
can remain immune to its ill effects. parameters. Feeder classification based on such an integrated
approach can be of immense help in planning exercise for
3) Phase Unbalancing improving daily operations.
While dense networks are sensitive to change in load
density, they remain indifferent to the change in phase unbal-
VI. C ONCLUSION
ancing. The branches carrying currents close to their capacities
show less sensitivity to shift in currents from one phase to Seasonal and diurnal variability of real power injections
the other over a short to medium distances. Whereas, sparse from DERs necessitate careful efforts in investigating impacts
networks are prone to the change in phase unbalancing. It of these sources on distribution network operations. Many
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 111

TABLE AII
efforts have been based on size, type, and location of DGs L OAD P ROFILE FOR IEEE 34 N ODE F EEDER
in distribution networks to evaluate their influence. However,
the response of distribution networks to the variability of DGs Phase A Phase B Phase C
Node Load Profile
kW kVAr kW kVAr kW kVAr
largely depends on feeder-specific factors as well. In this 840 27 16 31 18 9 7 Industrial 1
work, effects of feeder-specific factors, namely, geo-electric 844 144 110 135 105 135 105 Industrial 2
size, phase-unbalancing, and load density are analyzed with 848 20 16 43 27 20 16 Industrial 1
890a 150 75 150 75 150 75 Industrial 2
132 candidate scenarios based on three IEEE test feeders and 830 17 8 10 5 25 10 Commercial 1
one actual operating spot network in India. Simulation results 806 0 0 30 15 25 14 Commercial 2
imply a high degree of case-specificity in feeder response to 810 0 0 16 8 0 0 Residential 2
820 34 17 0 0 0 0 Residential 1
increasing PV penetration. Feeder voltage profile is found to 822 135 70 0 0 0 0 Commercial 2
be very sensitive to change in load density in case of highly 824 0 0 5 2 0 0 Residential 2
dense feeders whereas lightly loaded long feeders are prone 826 0 0 40 20 0 0 Residential 2
828 0 0 0 0 4 2 Residential 2
to phase unbalancing. In either case, DG hosting capacity of 856 0 0 4 2 0 0 Residential 1
feeders gets severely reduced due to increasing numbers of 858 7 3 2 1 6 3 Residential 1
voltage limit violations. Apart from voltage profile, results 864 2 1 0 0 0 0 Residential 1
834 4 2 15 8 13 7 Residential 1
also indicate that energy losses and tap-changer operations 860 36 24 40 36 130 71 Commercial 2
of voltage regulating equipment are also vulnerable to the 836 30 15 10 6 42 22 Residential 2
said factors. Diverging response characteristics are observed 838 0 0 28 14 0 0 Commercial 2
846 0 0 25 12 20 11 Residential 2
with variants of the same feeder, which implies that similar a PVDG is connected.
treatment to mitigate the unwanted effects of DGs can prove
grossly ineffective or simply irrelevant. Thus, the approach that
takes into account the effects of these feeder-specific factors, TABLE AIII
cyclic variability of DG, and load profile can be the ideal L OAD P ROFILE FOR IEEE 37 N ODE F EEDER
means for feeder classification exercise. Carefully selected
Phase A Phase B Phase C
samples of daily load and generation profiles can be used to Node
kW kVAr kW kVAr kW kVAr
Load Profile
run sequential time simulations for many feeders belonging to 701 140 70 140 70 350 175 Industrial 1
the same geographic region. Such an exercise can bring out the 712 0 0 0 0 85 40 Commercial 1
713 0 0 0 0 85 40 Commercial 2
effects of feeder-specific factors and can be of immense help 714 17 8 21 10 0 0 Residential 2
in formulating strategies to improve daily network operations. 718 85 40 0 0 0 0 Commercial 2
This study has used only phase-balanced injections from 720 0 0 0 0 85 40 Commercial 1
722 0 0 140 70 21 10 Residential 2
PVDGs to assess the effect of feeder-specific factors. However, 724 0 0 42 21 0 0 Residential 2
implications of this study are all the more applicable to single- 725 0 0 42 21 0 0 Residential 1
phase PVDGs and also for any type of DGs with uncertainty 727 0 0 0 0 42 21 Residential 1
728 42 21 42 21 42 21 Industrial 2
and intermittency. 729 42 21 0 0 0 0 Residential 2
730 0 0 0 0 85 40 Commercial 2
A PPENDIX 731 0 0 85 40 0 0 Commercial 1
732 0 0 0 0 42 21 Residential 1
733 85 40 0 0 0 0 Commercial 1
TABLE AI 734 0 0 0 0 42 21 Residential 1
L OAD P ROFILE FOR IEEE 13 N ODE F EEDER 735 0 0 0 0 85 40 Commercial 2
736 0 0 42 21 0 0 Residential 2
Phase A Phase B Phase C 737 140 70 0 0 0 0 Commercial 2
Node Load Profile
kW kVAr kW kVAr kW kVAr 738 126 62 0 0 0 0 Commercial 1
611 0 0 0 0 170 80 Residential 1 740 0 0 0 0 85 40 Commercial 1
634 160 110 120 90 120 90 Commercial 1 741 0 0 0 0 42 21 Residential 2
645 0 0 170 125 0 0 Residential 2 742 8 4 85 40 0 0 Commercial 2
646 0 0 230 132 0 0 Commercial 1 744 42 21 0 0 0 0 Commercial 2
652 128 86 0 0 0 0 Residential 1 775a 0 0 0 0 0 0 Commercial 1
671 402 230 451 258 502 288 Industrial 1 709 0 0 0 0 0 0 Industrial 2
675 485 190 68 60 290 212 Commercial 2 a PVDG is connected.
692 0 0 0 0 170 151 Residential 2
680a 0 0 0 0 0 0 Industrial 2
633 0 0 0 0 0 0 Industrial 1
a PVDG is connected.
112 CSEE JOURNAL OF POWER AND ENERGY SYSTEMS, VOL. 3, NO. 1, MARCH 2017

Substation transformer 799


A
650 B 724
C Voltage regulator
722
D-Yg, 5000 kVA 701 707
712
Voltage regulator 115/4.16 kV 713 704
705 720
742
632 633 634 702
646 645 744
703 714 706
729 727
728 725
730
684 671 692 675 708 709
611 731
732

652 733
775
680
~
~ PV Inverter 736 PV Inverter

710
Multi-crystalline Si PV 734
modules (STP260-20/Wd)
740
735
737 738 711 741
Fig. A 1. Modified IEEE 13 node test feeder with PVDG and additional
capacitor at node 646.
Fig. A3. Modified IEEE 37 node test feeder with PVDG at node 775.

Substation transformer
A ACKNOWLEDGMENT
800 B
C Kalpesh Joshi acknowledges the financial support of Indian
Institute of Technology Gandhinagar for this research work
D-Gr.W, 2500 kVA in the form of additional fellowship. Authors also gratefully
69/24.9 kV
acknowledge the support of officials from Torrent Power
810 808
Limited, Ahmedabad, Gujarat, INDIA for providing access to
the metered measurements of IITGN-VGEC campus.
Voltage regulator 1

816 822
R EFERENCES
[1] R. F. Arritt and R. C. Dugan, “Distribution system analysis and the
824 826 future smart grid,” IEEE Transactions on Industrial Application, vol.
846 47, no. 6, pp. 2343–2350, Nov. 2011.
856 854 [2] R. A. Walling, R. Saint, R. C. Dugan, J. Burke, and L. A. Kojovic,
“Summary of distributed resources impact on power delivery systems,”
IEEE Transactions on Power Delivery, vol. 23, no. 3. pp. 1636–1644,
Jul. 2008.
Voltage regulator 2 [3] R. J. Broderick, J. E. Quiroz, M. J. Reno, A. Ellis, J. Smith, and R.
XFM-1 Dugan. (2013, Jan.). Time series power flow analysis for distribution
890 connected PV generation. Sandia Corporation, U.S. [Online]. Available:
24/4.16 kV
http://energy.sandia.gov/wp-content/gallery/uploads/SAND Time-
844
832 Series-Power-Flow-Analysis-for-Distribution-Connected-PV-
Generation.pdf
[4] J. Smith, M. Rylander, L. Rogers, and R. Dugan, “It’s all in the plans:
Maximizing the benefits and minimizing the impacts of DERs in an
858 integrated grid,” IEEE Power Energy Magazine, vol. 13, no. 2, pp. 20–
~ 29, Mar. 2015.
PV inverter 834 [5] A. Hoke, R. Butler, J. Hambrick, and B. Kroposki, “Steady-state analysis
of maximum photovoltaic penetration levels on typical distribution
feeders,” IEEE Transactions on Sustainable Energy, vol. 4, no. 2, pp.
350–357, Apr. 2013.
840 [6] R. Tonkoski, D. Turcotte, and T. H. M. El-Fouly, “Impact of high
864 836
PV penetration on voltage profiles in residential neighborhoods,” IEEE
Multi-crystalline Si PV 838 Transactions on Sustainable Energy, vol. 3, no. 3, pp. 518–527, Jul.
2012.
Modules (STP260-20/Wd)
[7] R. C. Dugan. (2012, Jul.). The value of quasi-static time series
simulation. [Online]. Available: http://www1.eere.energy.gov/solar/pdfs/
Fig. A2. Modified IEEE 34 node test feeder with PVDG at node 890. distribution modeling wkshp2012 dugan.pdf
JOSHI et al.: CASE-SPECIFICITY AND ITS IMPLICATIONS IN DISTRIBUTION NETWORK ANALYSIS WITH INCREASING PENETRATION OF PHOTOVOLTAIC GENERATION 113

[8] K. A. Joshi and N. M. Pindoriya, “Impact investigation of rooftop solar Kalpesh A. Joshi (S’10) received his M.Tech.
PV system: A case study in India,” 2012 3rd IEEE PES International degree in power systems from Indian Institute of
Conference and Exhibition on Innovative Smart Grid Technologies Technology Delhi, India, in 2008. He has recently
(ISGT Europe), 2012, pp. 1–8. defended his Ph.D. thesis in electrical engineering
[9] D. Paradis, F. Katiraei, and B. Mather, “Comparative analysis of time- (December 2016) at Indian Institute of Technol-
series studies and transient simulations for impact assessment of PV ogy Gandhinagar. India. He was a visiting research
integration on reduced IEEE 8500 node feeder,” in 2013 IEEE Power scholar with Washington State University for six
& Energy Society General Meeting, 2013, pp. 1–5. months in 2015.
[10] R. F. Yan and T. K. Saha, “Voltage variation sensitivity analysis for un- Joshi’s research interests include modeling, anal-
balanced distribution networks due to photovoltaic power fluctuations,” ysis and optimization of distribution networks with
IEEE Transactions on Power Systems, vol. 27, no. 2, pp. 1078–1089, distributed generation and energy storage system.
May 2012. He is also interested in protection and control aspects of active distribution
[11] M. J. E. Alam, K. M. Muttaqi, and D. Sutanto, “A three-phase power networks and micro grids.
flow approach for integrated 3-wire MV and 4-wire multigrounded LV
networks with rooftop solar PV,” IEEE Transactions on Power Systems,
vol. 28, no. 2, pp. 1728–1737, May 2013.
[12] T.-H. Chen, “Evaluation of line loss under load unbalance using the
complex unbalance factor,” IEE Proceedings – Generation, Transmis- Naran M. Pindoriya (M’10–SM’15) received the
sion, and Distribution, vol. 142, no. 2, pp. 173–178, Mar. 1995. M.E. degree in electrical engineering from M. S.
[13] Z. Liu and J. V. Milanovic, “Probabilistic estimation of voltage un- University of Baroda, Gujarat, India in 2003 and
balance in MV distribution networks with unbalanced load,” IEEE Ph.D. in electrical engineering from Indian Institute
Transactions on Power Delivery, vol. 30, no. 2, pp. 693–703, Apr. 2015. of Technology Kanpur, Kanpur, India in 2009.
[14] C. S. Cheng and D. Shirmohammadi, “A three-phase power flow method He was a Research Fellow in the Department
for real-time distribution system analysis,” IEEE Transactions on Power of Electrical and Computer Engineering at National
Systems, vol. 10, no. 2, pp. 671–679, May 1995. University of Singapore, Singapore, in 2010. He
[15] P. A. N. Garcia, J. L. R. Pereira, S. Carneiro, V. M. da Costa, and is presently an Assistant Professor in electrical en-
N. Martins, “Three-phase power flow calculations using the current gineering at Indian Institute of Technology Gand-
injection method,” IEEE Transactions on Power Systems, vol. 15, no. 2, hinagar, Gandhinagar, India. His research interests
pp. 508–514, May 2000. include smart distribution grid, distributed energy resources, load forecasting,
[16] K. A. Joshi, “On some aspects of distribution system analysis and and energy management.
optimization,” M. Tech. thesis, Department of Electrical Engineering, He is a member of Institution of Engineers (India) and Life member of
IIT Delhi, 2008. Avaiable: http://www.kalpeshjoshi.co.in Indian Society for Technical Education (ISTE) India.
[17] W. H. Kersting, Distribution System Modeling and Analysis, Third
Edition. Boca Raton, FL: CRC Press, 2012.
[18] W. H. Kersting, “Radial distribution test feeders,” IEEE Transactions on
Power Systems, vol. 6, no. 3, pp. 975–985, Aug. 1991.
[19] M. G. Villalva, J. R. Gazoli, and E. R. Filho, “Comprehensive approach
to modeling and simulation of photovoltaic arrays,” IEEE Transactions
on Power Electronics, vol. 24, no. 5, pp. 1198–1208, May 2009.
[20] National Renewable Energy Laboratory. (2014, Sep.). Data Resources.
[Online]. Available: http://www.nrel.gov/gis/data.html
[21] Electric Reliability Council of Texas, Inc. (ERCOT). (2014). Annual
load profiles. [Online]. Available: http://www.ercot.com/mktinfo/
loadprofile
[22] P. Xiao, D. C. Yu, and W. Yan, “A unified three-phase transformer model
for distribution load flow calculations,” IEEE Transactions on Power
Systems, vol. 21, no. 1, pp. 153–159, Feb. 2006.
[23] American National Standard for Electric Power Systems and Equipment
– Voltage Ratings (60 Hertz), ANSI C84.1-2011, 2011.
[24] J. Fuller, Y. Xu, B. Kersing, R. Dugan, and S. Carneiro, “Distribution
Test Feeders - IEEE PES Distribution System Analysis Subcommittee’s
Distribution Test Feeder Working Group,” 2000. [Online]. Available:
https://ewh.ieee.org/soc/pes/dsacom/testfeeders/

You might also like