Sunteți pe pagina 1din 9

3038 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO.

3, AUGUST 2013

International Industry Practice


on Power System Load Modeling
Jovica V. Milanović, Fellow, IEEE, Koji Yamashita, Member, IEEE, Sergio Martínez Villanueva,
Sasa Ž. Djokić, Senior Member, IEEE, and Lidija M. Korunović, Member, IEEE

Abstract—Power system load modeling is a mature and gen- load models was performed in the mid 1990s [1]. Additionally,
erally well researched area which, as many other in electrical load models and their parameters currently used by utilities and
power engineering at the present time, is going through a period system operators for power system analysis are generally not in
of renewed interest in both industry and academia. This interest is
fueled by the appearance of new non-conventional types of loads public domain [2], and there is a level of uncertainty regarding
(power electronic-based, or interfaced through power electronics) industry acceptance of research efforts in this area.
and requirements to operate modern electric power systems with Although the majority of current power system research
increased penetration of non-conventional and mostly intermittent and industry acknowledges the importance of accurate load
types of generation in a safe and secure manner. As a response modeling, they still use typical representation of static loads
to this renewed interest, in February 2010 CIGRE established
working group C4.605: “Modelling and aggregation of loads in by the constant impedance/current/power load types, while
flexible power networks”. One of the first tasks of the working dynamic loads are usually represented with the induction
group was to identify current international industry practice motor (IM) model. General inadequacy of currently used
on load modeling for static and dynamic power system studies. load models was highlighted in several unsuccessful at-
For that purpose, a questionnaire was developed and distributed tempts to reconstruct recent blackouts in the corresponding
during the summer/autumn of 2010 to more than 160 utilities
and system operators in over 50 countries on five continents. “post-mortem” simulations and analysis (e.g., [3]). During
This paper summarizes some of the key findings from about 100 the power system stability analysis, the emphasis is mainly
responses to the questionnaire received by September 2011 and placed on modeling power generating units, while load
identifies prevalent types of load models used as well as typical models are regarded as of secondary importance, although
values of their parameters. the influence of load representation on the stability was
Index Terms—Dynamic and composite load models, load mod- recognized a long time ago [4]–[7]. Power system engineers
eling, power system study, questionnaire, static, survey. began to pay more attention to the load modeling since the
Swedish blackout of 1983 [8], as inappropriate representation
of system loads has usually led to the discrepancies between
I. INTRODUCTION
the recorded and simulated system responses [8]–[11].
Recently, there is a renewed interest in both industry and

A N important requirement for the correct analysis of academia for load modeling due to appearance of new types
planning, operation and control of both distribution and of loads, offering increased efficiency and controllability. For
transmission networks is accurate representation of steady state example, different types of modern non-linear power electronic
and dynamic characteristics of system loads. Most of the load loads are now responsible for more than 30% of the total de-
models used today were developed several decades ago, and mand in the U.K. residential load sector [12]. Similarly, there
are not adequately updated after the subsequent changes in load are currently no appropriate load models available for the cor-
structure and load characteristics-the last systematic update of rect representation of various directly connected and inverter-
interfaced micro and small-scale distributed generation tech-
nologies, which, in some of the future network scenarios, will
Manuscript received July 01, 2012; revised October 14, 2012; accepted De-
cember 01, 2012. Date of publication January 04, 2013; date of current version strongly impact real and reactive power demands, as they will
July 18, 2013. Paper no. TPWRS-00755-2012. be installed in large numbers [13].
J. V. Milanović is with the School of Electrical and Electronic Engineering,
In a response to this renewed interest in load modeling,
The University of Manchester, Manchester M13 9PL, U.K. (e-mail: mi-
lanovic@manchester.ac.uk). CIGRE established a new working group C4.605: “Modelling
K. Yamashita is with the Central Research Institute of Electric Power In- and aggregation of loads in flexible power networks” in Feb-
dustry, Komae-shi, Tokyo 201-8511, Japan (e-mail: yama@criepi.denken.or.
ruary 2010 with a remit to: 1) provide a critical and updated
jp).
S. Martínez Villanueva is with Red Eléctrica de España, 28109 Alcobendas overview of existing load models and their parameters for
(Madrid), Spain (e-mail: smartinez@ree.es). power system studies at all voltage levels, and identify types
S. Ž. Djokić is with the Institute for Energy Systems, School of Engi-
of loads and load classes for which adequate load models
neering, the University of Edinburgh, Edinburgh EH9 3JL, U.K. (e-mail:
sasa.djokic@ed.ac.uk). are presently missing; 2) provide a comprehensive overview
L. M. Korunović is with the Faculty of Electronic Engineering, University of of existing methodologies for load modeling, with a critical
Niš, 18000 Niš, Republic of Serbia (e-mail: lidija.korunovic@elfak.ni.ac.rs).
overview of component-based and measurement-based ap-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. proaches, clearly identifying their advantages and disadvan-
Digital Object Identifier 10.1109/TPWRS.2012.2231969 tages; 3) develop a set of recommendations and step-by-step

0885-8950/$31.00 © 2013 IEEE


MILANOVIĆ et al.: INTERNATIONAL INDUSTRY PRACTICE ON POWER SYSTEM LOAD MODELING 3039

procedures for load model development and validation using TABLE I


either component-based or measurement-based approaches, or SURVEY QUESTIONS AND CATEGORIES
their combination; 4) develop load models for all typical de-
vices and classes of customers for which there are no existing
models and recommend their typical parameter values and
ranges; 5) provide recommendations on developing equiv-
alent static and dynamic models of networks with signifi-
cant amount of distributed generation, including equivalent
models of micro-grids and active distribution network cells
[14]–[16].
One of the first tasks of the CIGRE WG C4.605 was to
survey and identify current industry practice for power system
load modeling. For that purpose, a comprehensive question-
naire on load modeling practices was developed and distributed
during the summer/autumn of 2010 to more than 160 utilities
and system operators from over 50 countries on all continents.
This paper summarizes some of the key findings of that ques- TABLE II
tionnaire, based on about 100 responses received by September PARTICIPANTS AND RESPONSE RATES OF THE SURVEY
2011.
The initial, partial, results of the CIGRE WG C4.605 ques-
tionnaire originally reported in [14] are updated in this paper by
incorporating the responses received after publishing [14] and,
more importantly, by full statistical analysis of the results of the
survey in order to identify prevalent types of load models used
in industry and typical values of their parameters.
The previous work reported in [1] and [2] provided impor-
tant information on load modeling and also recommended dif-
ferent load models for specific power system studies, targeted Note: “Americas” denotes both North and South America.
mainly for the utilities and system operators in the U.S. The aim
of this paper is not to recommend any specific load model for respondents who expressed willingness in completing it after
any specific power system study, but, rather, to identify what returning the short questionnaire. This paper considers only re-
load models are currently used and to provide some basic in- sponses to the questions from the short questionnaire (Table I),
formation on their use. Furthermore, this paper presents infor- which are classified into four categories: 1) types of load models
mation on currently used load models and their parameters by for static and dynamic power system studies, 2) approaches for
the utilities and system operators not only in the U.S., but also identification of load models and model parameters, 3) ade-
in the rest of the world. Although that makes this paper signif- quacy of load models and used load simulation tools, 4) ap-
icantly different from [1] and [2], the presented results can be proaches for including small distributed generation (DG) in load
used to check whether some of the load models recommended models. It should be pointed out that multiple answers were al-
in [1] and [2] are actually used in power system studies. Sim- lowed for questions Q1, Q2, Q4, Q5, and Q6, as respondents
ilarly compared to another milestone reference, [11], which is might use more than one load model, or more than one approach
a bibliographical list of papers related to load modeling and for load model data collection. Multiple answers however, were
load model parameter identification published in the literature not allowed for questions Q3, Q7, Q8, and Q9.
by 1993, this paper presents and important update as it reports
the results of a survey conducted almost twenty years after [11]
B. Survey Participants
was published.
Potential survey participants were identified by WG members
and contacted by e-mail. The questionnaire was sent to 160 con-
II. SURVEY tacts around the world (different utilities and system operators)
between June and December 2010. By September 2011, the re-
A. Organization of the Questionnaire sponses to short survey were received from 97 contacts from dif-
ferent continents, resulting in the overall response rate of about
A two-part questionnaire was developed for the purpose of 60.6% (see 2nd column, 3rd column and last row in Table II).
conducting the survey on load modeling practices. The first part Per-continent overall response rate was higher than 50% for 4
was shorter and contained only nine general questions, which out of 5 continents (see 4th column in Table II). If only the re-
were sent to all participants in order to facilitate high response ceived responses are considered, the relative per-continent re-
rate and provide a basic, but comprehensive overview of the in- sponse rates are shown in the last column in Table II. In case of
ternational load modeling approaches and practices. The second a detailed questionnaire, the response rates were much lower, as
part was a longer questionnaire, which was sent only to those these responses were received from only 19 initial participants.
3040 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

Fig. 1. Q1: Load models used for steady state power system studies.

III. RESULTS OF THE SURVEY

A. Q1: Load Models Used in Steady State Power System


Studies
Several sample answers were offered to the respondents to
this question: constant power load model (PQ); simple poly-
nomial load model, i.e., combination of constant impedance,
constant current and constant power (ZIP) load model; combi-
nation of ZIP and equivalent induction motor model; detailed
composite load model; and exponential load model, [1].
Exponential and polynomial (ZIP) load models (only for real
power demand) are given by (1) and (2):

Fig. 2. Q2: Load models used for dynamic power system studies. a) Real power
(1) load model. b) Reactive power load model.

(2)
are very different. According to the received responses to ques-
where ; tion Q2, it was not possible to derive clear and confident conclu-
; ; sions about the types of load models used in different dynamic
and : Actual and initial real power of modeled load; studies, as only 23% of the respondents indicated which load
and : Actual and initial load bus voltage magnitudes; , models are used in specific dynamic studies (most of them did
: corresponding shares of constant P, I or Z load, and not differentiate between transient stability studies and voltage
. stability studies, while only a small portion mentioned small
The widely accepted practice for power flow analysis of signal stability studies or frequency stability studies). In total,
electric networks is to assume that the distribution system tap 8 different types of load models are used, featuring additional
changing transformers and voltage regulators have brought bus differences between load models used for representing real and
voltages close to nominal values (i.e., close to 1 p.u. value). In reactive power demands. The responses to this question (Q2)
this case, loads may be treated as a constant real and reactive suggest that there is no dominant dynamic load modeling prac-
power demands, and constant PQ load model can be used. tice, which is summarized in Fig. 2 separately for real and reac-
Therefore, as expected, the responses to this question (Q1, tive power load models.
Fig. 1) identified constant power PQ load model as by far the It can be seen from Fig. 2 that although there is a relatively
most dominant type of load model used in steady state power even spread of different load models used in dynamic system
system studies (84% of all responses). studies for modeling real power demand, static load models
are again dominant. Constant power and constant current load
models account for about 42% of all used models. Similar dom-
B. Q2: Load Models Used in Dynamic Power System Studies
inance of static load models is observed in case of modeling
Load models used in dynamic power system studies (for, e.g., reactive power demand. In this case, however, constant power
transient stability and frequency stability, or short-term voltage and constant impedance load models account for 45% of all
stability analysis) by different network operators and utilities used load models. For modeling both real and reactive power
MILANOVIĆ et al.: INTERNATIONAL INDUSTRY PRACTICE ON POWER SYSTEM LOAD MODELING 3041

Fig. 3. Q3: Use of different load models for different load classes in dynamic
power system studies. Fig. 4. Q4: Approaches for load model parameter identification.

demand, about 30% of the used models represent dynamic load D. Q4: Load Model Collection and Parameter Identification
by some form of induction motor model (IM with ZIP or expo-
nential, or composite load model). During the past few decades, capabilities of various measure-
In case of load models for dynamic system studies, there was ment devices have been significantly improved and their num-
no uniformity between different utilities, countries and conti- bers in power systems have substantially increased. This opened
nents, as in case of load models for static power system studies. a possibility for the measurement and capture of data required
More than a half of utilities and system operators surveyed in for load modeling and, particularly, load model parameter iden-
Americas use load models which include different types of in- tification and acquisition. According to received responses to
duction motor models, i.e., dynamic load models for dynamic this question (Q4, Fig. 4), it is clear that power industry is taking
system studies. In contrast, 100% of utilities in Africa use static advantage of available measuring and monitoring systems, as
load models for dynamic power system studies. the current practice for load model parameter identification is
In general, the constant power PQ load model is still the most based on the measurements in over 50% of the cases (on av-
widely used in a majority of power system stability studies (as erage), while in Americas and Asia this figure is 60% and 65%,
it is the most conservative approach), together with the con- respectively.
stant current for real power and constant impedance for reac- It is interesting to note that in 19% of the cases (on average),
tive power model. Arguably, this might be a consequence of a the utilities rely solely on model parameters available in the lit-
recommendation given almost twenty years ago in [2], that in erature. This highlights the importance of publishing the results
the absence of a detailed information on the load structure/com- of load model parameter identification studies based on mea-
position, real power demand can be represented using constant surements by utilities around the world. These results are likely
current and reactive power demand using constant impedance to be used by other utilities, e.g., those that might not have re-
load models. sources to conduct their own load modeling studies.

C. Q3: Load Models for Specific Load Classes E. Q5: Most Recent Update of Load Model Parameters

Although the same type of the load model can be used for Another interesting observation from the survey is how
representing loads at different buses, its parameters could vary frequently load model parameters are updated. This has been
depending on the modeled load class (e.g., residential, indus- traditionally one of the major difficulties associated with load
trial, commercial, etc.) [17]–[20]. However, according to the re- modeling, as the appearance of new types of loads/devices
sponses to this question (Q3, Fig. 3), the current practice mostly with different characteristics may invalidate load models and
does not discriminate between different load classes. In large parameters established in the past. Regular update of load
majority of the cases, the same load model and model parame- model parameters is, therefore, essential to ensure accurate
ters are used throughout the modeled system. Therefore, in gen- simulation results. According to the responses to this question
eral, there is no discrimination neither between load classes nor (Q5, Fig. 5), the utilities and system operators are very well
loads used at different voltage levels (i.e., high, medium and low aware of this issue, and they relatively frequently update load
voltage networks) based on received responses. This could be model parameters used in simulations. In 41% of the cases,
explained by the difficulties that transmission system operators the load model parameters were updated within the last five
encounter in obtaining accurate information on load classes at years. Significance that utilities in general, and those in Europe,
different network buses, because these data are typically owned Africa and Americas in particular, are currently placing on load
by distribution system companies. modeling is demonstrated through the fact that in 20% of the
3042 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

Fig. 5. Q5: Frequency of updating load model parameters. Fig. 7. Q7: Adequacy of available load models for system stability studies.

Fig. 6. Q6: Power system analysis and load simulation tools [21]–[27]. Fig. 8. Q8: Extent of use of user-defined load models.

cases, the load model parameters have been updated during the In spite of wide reliance on models available in commercial
past year. software packages, it should be noted that a relatively large por-
tion of surveyed utilities and system operators in Europe and
F. Q6: Load Simulation Tools Used for System Studies Americas (about 35%) are basically not satisfied with provided
load models and they either modify them, or would like them to
Whatever mathematical model of power system load is devel- be modified.
oped or used, it has to be incorporated with the models of other
system components into the overall model of power system, H. Q9: Modeling of Small Distributed Generation
which is then applied in computer simulations by in-house de- Increasing penetration levels of both medium/large dis-
veloped or commercially available software. According to the tributed generation (DG) systems in medium and high voltage
responses to this question (Q6, Fig. 6), currently most widely networks, as well as highly dispersed micro/small DG in low
used simulation software by industry for power system analysis voltage networks, require development of new load models
are PSS/E (41% of users) and DIgSilent (13% of users), indi- for correct representation of aggregate system demands and
cating that standard load models (i.e., “load model libraries”) power flows. According to responses to this question (Q9,
from these software packages are also the most widely used. Fig. 9), however, 40% of the surveyed network operators and
utilities do not currently consider at all these technologies when
G. Q7 and Q8: Adequacy of Available Load Models
modeling demand at bulk load supply points (BLSP) in system
According to the received responses, most engineers are sat- studies.
isfied with the load models available in existing software pack- In further 28% of the cases, DG is simply modeled as a nega-
ages (85% on average, Q7 in Fig. 7) and do not develop their tive load. Some utilities and system operators, however, model
own load models for power system studies (78% on average, DG as an independent component from the system load, espe-
Q8 in Fig. 8). This places significant pressure on power system cially when DG penetration in the distribution system is rela-
software developers to keep load models and model parameters tively high. In most of cases, current penetration levels of var-
regularly updated in their simulation tools. ious DG technologies are reasonably low, which may justify ne-
MILANOVIĆ et al.: INTERNATIONAL INDUSTRY PRACTICE ON POWER SYSTEM LOAD MODELING 3043

TABLE III
PARAMETERS OF ZIP MODEL AND EQUIVALENT EXPONENTIAL
MODEL: FOR ALL RESPONDENTS FROM NORTH AND SOUTH AMERICA

Presented as the constant current (I) load type in Fig. 2


Presented as the constant impedance (Z) load type in Fig. 2

TABLE IV
Fig. 9. Q9: Inclusion of small DG in load models. EXAMPLES OF EXPONENTIAL LOAD MODEL PARAMETERS
PROVIDED BY ONE RESPONDENT FROM EUROPE

glecting their influence on bulk-supplied aggregate loads. Antic-


ipated increase of various DG technologies (particularly those
based on power electronic-interfaced intermittent/variable re-
newable energy resources), however, will effectively change the
However, for the conversion of the ZIP model into the exponen-
nature of BLSP responses to network disturbances. Generally,
tial model, a separate procedure is required.
it is not feasible to model large number of small and highly dis-
A simple conversion between ZIP load models reported in
persed individual DG units as individual generators due to sig-
the survey (2) and equivalent exponential load model (1) can be
nificant increases in model complexity and computational time.
done by (3) and (4):
Therefore, their representation will require some form of ag-
gregation and equivalenting, either as an “aggregate generation
model”, or with the DG correctly incorporated as a part of the (3)
overall aggregate (BLSP) load model.

IV. PREVALENT LOAD MODELS


Based on the results of the survey, some additional analysis
is performed in this section, in order to identify prevalent (i.e., (4)
dominant or overall average) load models used by utilities and
network operators around the World. It is difficult to specify a where and are active and reactive power coefficients of
single prevalent or typical dynamic load model, as load models the equivalent exponential load model, and all other quantities
currently used in dynamic system studies vary in relatively are defined in (1) and (2).
wide ranges (Fig. 2). Regardless of the type of power system The conversion by (4) is based on Taylor expansion, which
study (static or dynamic), the most common types of static load is more accurate if the voltage changes are smaller. Table III
models by far are individual constant power, constant current shows an example of the conversion and equivalent and
and constant impedance loads, or their combination expressed coefficients for five network operators and utilities from North
in a form of a ZIP load model (Figs. 1 and 2). and South America. (Note: Some respondents from Americas
Based on the individual constant power/current/impedance did not provide any specific information for the selected ZIP
and ZIP load models reported in the survey, parameters of equiv- model as the answer to Q2.) On the other hand, Table IV shows
alent exponential load model, i.e., voltage exponents and four examples of the coefficients of exponential load models
, are calculated next. The conversion from polynomial to used by one respondent from Europe. (Note: In this case, the
exponential load model consists of two steps: 1) conversion conversion is not performed, as this respondent indicated four
of different polynomial load models into the exponential load exponential load models with the corresponding coefficients as
model, and 2) calculation of equivalent values of exponential the actually used load models.)
load model coefficients for real and reactive power, and . Following the above conversion procedure, currently used
The main point of this analysis is to uniformly convert all static load models for dynamic stability studies, shown in Fig. 2,
static load models used for power system studies into a single are converted into equivalent exponential load models shown in
exponential load model. For example, Fig. 2 shows that used Fig. 10.
static load models include “constant Z”, “constant I”, “constant The conversion of static load models used by the survey
P”, ZIP, and exponential load models. As shown in (2), in case respondents in dynamic system stability studies into equivalent
of the constant Z, I, and P load models, the corresponding ex- exponential load models is further statistically analyzed in
ponential load model coefficients are 2, 1, and 0, respectively. Tables V and VII, as well as in Figs. 11 and 12. These results
3044 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

Fig. 11. Histogram of coefficients of equivalent exponential models for


static load models used worldwide in dynamic power system studies.

Fig. 12. Histogram of coefficients of equivalent exponential models for


static load models used worldwide in dynamic power system studies.

TABLE VI
ANALYSIS OF VALUES OF COEFFICIENT WITHOUT CONVERSION

Fig. 10. Identified prevalent static load models used for dynamic power system
studies, represented as equivalent exponential load model. a) Real power load
model. b) Reactive power load model.

TABLE V
FURTHER RESULTS OF THE ANALYSIS OF EQUIVALENT COEFFICIENT

coefficient for all continents and worldwide. Using these stan-


dard deviations and corresponding mean values, somewhat
narrower ranges of are obtained according to the formulae:

(5)

where is mean value and is standard deviation. These


ranges are shown in the last column in Table V, and the same
approach is repeated for calculated equivalent coefficient.
For comparison, Table VI presents the results of statistical anal-
ysis of processing of excluding those values that are obtained
show minimum, maximum, median and mean values of co- by conversion. Most of the differences between Table V and
efficients and , their standard deviations and ranges for Table VI are small, but generally greater standard deviations and
different continents and for the World as a whole. therefore wider ranges are obtained in Table VI.
It can be seen that in case of Asia, Americas, Africa and Eu- Table VII shows that the values of are spread over wider
rope, varies from 0 to 2, while in Oceania it ranges from 0 to range than the values of . Mean values of vary from 0.79
1.7. Mean values of are less than 1 for all continents except for Africa, to 2.25 for Oceania, while the mean worldwide value
for Oceania, where it is slightly greater than 1. Consequently, of is 1.3 (worldwide median value is 1.8, Fig. 10(b)). Stan-
mean value of for the World is approximately 0.7, while its dard deviations of values are again of the same order as the
median value is 1.0 (Fig. 10(a)). corresponding mean values, confirming a rather widely spread
Standard deviations are of the same order as the corre- of values. Table VIII presents the results of the analysis of
sponding mean values, indicating a large dispersion of values that are not obtained by conversion. As expected, the
MILANOVIĆ et al.: INTERNATIONAL INDUSTRY PRACTICE ON POWER SYSTEM LOAD MODELING 3045

TABLE VII The following main conclusions can be drawn from the re-
FURTHER RESULTS OF THE ANALYSIS OF EQUIVALENT COEFFICIENT sults of the survey:
• About 70% of utilities and system operators around the
World use only static load model for power system stability
studies.
• About 30% of utilities and system operators use some form
of induction motor model to represent dynamic loads in
power system stability studies.
• Dominant practice in the USA is to use a combination of
static (typically ZIP) and dynamic load model (typically
induction motor), while use of static load models is preva-
lent in the rest of the world.
TABLE VIII
• In about 40% of the cases, currently used load model pa-
ANALYSIS OF VALUES OF COEFFICIENT WITHOUT CONVERSION rameters have been updated within the last five years.
• Typical static load model used in steady state studies is
constant power (PQ) load for both real (active) and reactive
power, i.e., .
• Typical values of and coefficients of the equiva-
lent exponential static load models used in dynamic studies
worldwide are 0.7 and 1.3, respectively
• Most of the utilities and system operators represent dis-
tributed generation as a negative load in system studies,
without modeling them explicitly. Some of them, however,
recognize the importance of appropriate DG models, par-
ticularly for future power system studies.
greater standard deviations and therefore wider ranges are ob- The results of the survey presented in the paper can be used to
tained in Table VIII in comparison with corresponding standard indicate in which areas some additional work on load modeling
deviations in Table VII. Exception is in the case of Oceania (or on updating of currently used load models) may be required
where none of values is obtained by conversion. by some utilities and system operators and such help them to
Since the dispersion of parameters and used in dy- identify where to put more effort. The results of the survey,
namic studies is significant, both mean values and median however, even if used for simple benchmarking purposes by in-
values can be treated as prevalent, i.e., typical. For example, dividual utilities and system operators (e.g., to compare their
prevalent value of for Europe is approximately 0.6, while load modeling practices with those indicated as the prevalent
for Americas and the World it is approximately 0.7/1.0 (Note: ones in the survey), should not be considered as the guidelines
mean value/median value). Prevalent or typical value of for load modeling. Each utility/system operator should perform
is almost 1/0.4 for Europe, but it is approximately 2.3/2.8 for their own analyses and carry out studies to determine most ap-
Oceania and 1.3/1.8 globally. propriate load models (and identify their parameters) for their
network. For utilities and system operators that cannot perform
V. CONCLUSIONS their own load modeling studies (for whatever reason), or have
This paper presents summary of the results from a worldwide not identified appropriate load models and their parameters yet,
survey of industry practice on power system load modeling. The the presented results could be used for initial studies, or as an
survey was initiated by CIGRE Working Group C4.605, and indication for selecting provisional load models and parameter
was sent to 160 system operators and utilities around the World. values.
Main questions from the survey covered following areas and
topics: 1) types of load models for static and dynamic power ACKNOWLEDGMENT
system studies, 2) approaches for identification of load models The authors appreciate the support of their colleagues from
and model parameters, 3) adequacy of load models and used CIGRE WG C4.605, who contributed in many different, but
load simulation tools, 4) approaches for including small dis- always useful ways to this paper: providing industry contacts,
tributed generation (DG) in load models. In total, 97 responses helping with the development and distribution of the survey, etc.
to the survey were received. Members of the WG C4.605 are J. V. Milanovic (Convener,
The survey indicated that there are large differences between U.K.), O. Auer (AR), A. Borghetti (IT), S. Djokic (U.K.), Z.
the load models used for static and dynamic power system Dong (AU), A. Gaikwad (US), A. Halley (AU), K. Karoui (BE),
studies by different system operators and utilities. Therefore, L. Korunovic (RS), D. Kosterev (US), J. Leung (AU), S. M.
additional analysis of the received responses is performed, Villanueva (ES), J. Ma (CN), J. Matevosyan (US), D. Mtolo
including the calculation of equivalent exponential load model, (ZA), P. Pourbeik (US), F. Rosende (PT), S. Sterpu (FR), F.
in order to identify prevalent (i.e., dominant or overall average) Villella (BE), and K. Yamashita (JP).
load models used by system operators and utilities in different Equally important, the authors would like to acknowledge the
continents and worldwide. support of the colleagues from utilities and system operators
3046 IEEE TRANSACTIONS ON POWER SYSTEMS, VOL. 28, NO. 3, AUGUST 2013

around the world, who took the time to fill in the questionnaire/ [19] EPRI Report of Project RP849-7, Load Modelling for Power Flow
survey and kindly shared their knowledge and experience. and Transient Stability Studies, EPRI EL-5003, Jan. 1987, prepared
by General Electric Company.
[20] D. Kosterev, A. Meklin, J. Undrill, B. Lesieutre, W. Price, D. Chassin,
R. Bravo, and S. Yang, “Load modeling in power system studies:
REFERENCES WECC progress update,” in Proc. 2008 IEEE Power and Energy
Society General Meeting—Conversion and Delivery of Electrical
Energy in the 21st Century, 2008, pp. 1–8.
[1] IEEE Task Force on Load Representation for Dynamic Performance,
[21] PSS/E, Program Application Guide, SIEMENS Energy Inc., Oct. 2010.
“Standard load models for power flow and dynamic performance sim-
[22] DIgSilent, Technical Documentation: General Load Model, DIgSI-
ulation,” IEEE Trans. Power Syst., vol. 10, pp. 1302–1313, 1995.
LENT GmbH., Oct. 2008.
[2] IEEE Task Force on Load Representation for Dynamic Performance,
[23] Integrated Analysis Software for Bulk Power System Stability, Power
“Load representation for dynamic performance analysis (of power sys-
System Study Group, CRIEPI Report, ET90002, Jul. 1991.
tems),” IEEE Trans. Power Syst., vol. 8, no. 2, pp. 472–482, May 1993.
[24] GE Energy. [Online]. Available: http://www.ge-energy.com/prod-
[3] D. N. Kosterev, C. W. Taylor, and W. A. Mittelstadt, “Model validation
ucts_and_services/products/concorda_software_suite/.
for the August 10, 1996 WSCC system outage,” IEEE Trans. Power
[25] W. Zhongxi and Z. Xiaoxin, “Power system analyais software package
Syst., vol. 14, no. 3, pp. 967–979, Aug. 1999.
(PSASP)—An integrated power system analysis tool,” in Proc. 1998
[4] W. Mauricio and A. Semlyen, “Effect of load characteristics on the
Int. Conf. POWERCON, 1998, vol. 1, pp. 7–11.
dynamic stability of power systems,” IEEE Trans. Power App. Syst.,
[26] TRACTEBEL Engineering GDFSUEZ. [Online]. Available:
vol. PAS-91, pp. 2295–2304, 1972.
http://www.eurostag.be/en/products/eurostag/the-reference-power-
[5] P. Kundur et al., in 294 T 622 Investigation of Low Frequency-area
system-dynamic-simulation/.
Oscillation Problems in Large Interconnected Power Systems, Ontario
[27] Manitoba HVDC Research Center. [Online]. Available: https://pscad.
Hydro, Toronto, ON, Canada, 1993.
com/products/pscad.
[6] C. Concordia and S. Ihara, “Load representation in power system sta-
bility studies,” IEEE Trans. Power App. Syst., vol. PAS-101, no. 4, pp.
969–977, Apr. 1982. Jovica V. Milanović (M’95–SM’98–F’10) received the Dipl.Ing. and M.Sc. de-
[7] M. H. Kent, W. Schmus, and L. M. Wheeler, “Dynamic modelling of grees from the University of Belgrade, Belgrade, Yugoslavia, the Ph.D. degree
loads in stability studies,” IEEE Trans. Power App. Syst., vol. PAS-88, from the University of Newcastle, Newcastle, Australia, and Higher Doctorate
no. 5, pp. 756–763, May 1969. (D.Sc. degree) from The University of Manchester, Manchester, U.K., all in
[8] D. S. Brereton, D. G. Lewis, and C. C. Young, “Representation of electrical engineering.
induction motor loads during power system stability studies,” AIEE Currently he is a Professor of Electrical Power Engineering and Director of
Trans., vol. PAS-76, pp. 451–460, 1957. External Affairs in the School of Electrical and Electronic Engineering of the
[9] K. Walve, Modelling of Power System Components at Severe Distur- University of Manchester (formerly UMIST), Manchester, U.K., Visiting Pro-
bance CIGRE Report 38-18, 1986. fessor at the University of Novi Sad, Novi Sad, Serbia, and Conjoint Professor
[10] CIGRE Committee No. 13 Report, Load Representation in Stability at the University of Newcastle, Newcastle, Australia.
Studies and the Effect of the Load-voltage Characteristics on Tran-
sient Stability, Submitted by Swedish and Norwegian Power Compa-
nies, 1964.
[11] IEEE Task Force on Load Representation for Dynamic Performance, Koji Yamashita (M’04) received the B.Sc. and M.Sc. degrees from Waseda
“Bibliography on load models for power flow and dynamic perfor- University, Tokyo, Japan, in 1993 and 1995, respectively.
mance simulation,” IEEE Trans. Power Syst., vol. 10, no. 1, pp. Since 1995, he has been with the Department of Power Systems at the Central
523–538, Feb. 1995. Research Institute of Electric Power Industry in Tokyo, Japan. He was a visiting
[12] Department of Energy and Climate Change (DECC), “Energy Trends, researcher at Iowa State University, Ames, during 2006–2007
Section 5: Electricity” and “Electricity Consumption in the UK: Do-
mestic,” National Statistics Publications, 2008, 2009, and 2010.
[13] Department of Energy and Climate Change (DECC), “Energy Act
2008” and “Microgeneration Strategy Consultation” 2008 and 2010. Sergio Martínez Villanueva received the degree of Electrical Engineer from
[14] K. Yamashita, S. M. Villanueva, and J. V. Milanovic, “Initial results the Universidad Politécnica de Madrid, Madrid, Spain, in 2002.
of international survey on industrial practice on power system load He joined Red Electrica de España S.A. in 2005. Presently, he works in the
modelling conducted by CIGRE WG C4.605,” in Proc. CIGRE Symp., Network Studies Department.
Bologna, Italy, 2011, vol. C4-333.
[15] J. Matevosyan, S. M. Villanueva, S. Z. Djokic, J. L. Acosta, S.
M-Zali, F. O. Resende, and J. V. Milanovic, “Aggregated models of
wind-based generation and active distribution network cells for power
system studies—Literature overview,” in Proc. IEEE Power Tech Sasa Ž. Djokić (M’05–SM’11) received the Dipl.Ing. and M.Sc. degrees in elec-
Conf., Trondheim, Norway, 2012. trical engineering from the University of Nis, Nis, Serbia, and the Ph.D. degree
[16] K. Yamashita, S. Z. Djokic, J. Matevosyan, Z. Y. Dong, F. O. Re- in the same area from the University of Manchester Institute of Science and
sende, and J. V. Milanovic, “Modelling and aggregation of loads in Technology (UMIST), Manchester, U.K.
flexible power networks—Scope and status of the work of CIGRE WG Currently, he is a Senior Lecturer in the School of Engineering at the Univer-
C4.605,” in Proc. IFAC Symp., Toulouse, France, 2012. sity of Edinburgh, Edinburgh, U.K.
[17] EPRI Report of Project RP849-3, Determining Load Characteristics for
Transient Performance, EPRI EL-840, May 1979, vol. 1–3, prepared by
the University of Texas at Arlington.
[18] EPRI Report of Project RP849-1, Determining Load Characteristics for Lidija M. Korunović (M’04) received the Dipl.Ing., M.Sc., and Ph.D. degrees
Transient Performance, EPRI EL-850, Mar. 1981, prepared by General from the Faculty of Electronic Engineering, The University of Nis, Nis, Serbia.
Electric Company. Currently she is a Lecturer at the same university.

S-ar putea să vă placă și