Sie sind auf Seite 1von 8

Load modeling for fault location in distribution

systems with distributed generation.


A. Herrera-Orozco, S. Prez-Londoo and J. Mora-Flrez, Member, IEEE

therefore this improvement has become an important target for


Abstract-- This paper shows the influence of the load modeling utilities, since this will avoid penalties and help to improving
in the behavior of voltage (V), current (I) and impedance (Z) the quality of electric supply. Currently there are several
measurements during pre-fault and fault steady states of power applications that estimate both, the distance to the fault and the
distribution systems. The test system is the 24.9 kV IEEE 34
faulted zone, through techniques such as methods based on the
nodes power distribution system, modeled by ATP software,
considering load models as constant impedance, constant current, model (MBM) and knowledge-based methods (MBC) [4].
constant power and hybrid load composed using the previous However, an important aspect for the adequate performance of
three models. All loads were coded using Models of ATP. As the locators is the modeling of the power system.
result of the test, a comparative analysis of the signals of V and I When performing simulation of a power distribution
measurements and also Z as seen from the power substation is system using a specific software, it is noticed that the circuit
performed. The result shows that the load model affects the fault
behavior can be appreciably different from the system
steady state of the power distribution system, which directly will
influence the fault locator performance, since this uses the operated by the utility. These differences affect considerably
measurements of V and I at the power substation at pre-fault and the performance of locators, since all the methods are based on
fault steady states. the system model. A common case and caused by the absence
of information in the distribution companies, leads to the
Index Terms--Distributed generation, fault location, power existence of circuit modeling parameters which are assumed,
distribution, static load models, ZIP load model. and therefore may be different to those in the real power
system, as the value and load type, among the most important
I. INTRODUCTION ones. The load models used in most of the applications of fault

T HE power distribution system is an important part of the


power system and is directly related to the end user, then
the requirements of high quality of power are mandatory. This
location are constant impedance [5], [6] and [7]. Considering
the actual behavior of the electrical power system, the set of
protections and loads, this paper is devoted to analyze their
is the reason of why the research activity in areas associated to possible influence on the fault location.
power quality has experienced a strong progress in recent Thus, this paper proposes a comparative analysis between
decades, specifically in relation to the waveform, service different load models, using the values of the voltage (V) and
continuity and customer support [1]. The current importance current (I) measurements, and values of the impedances (Z) as
of these aspects is associated with the active participation of seen from the power substation considering a power
private resources in the electricity sector which requires of the distribution system with distributed generation (DG), and also
establishment of remuneration guidelines and quality indexes performing different types of shunt faults (single-phase, phase
[2] [3]. Product quality means satisfying conditions of wave to phase, double phase to ground and three-phase), with
quality (or power) and supply continuity [4]. The problem different fault resistances (Rf) and varying the load type from
discussed here is associated with supply continuity. a programmed model using the tool Models in ATP (Constant
The faults in distribution systems affect negatively the impedance, constant current, constant power, and hybrid
indices proposed to quantify the continuity of supply, and models [5], [8]), simulated at different distances from the
substation. Additionally, these models were used by fault
locators in power distribution systems with distributed
This work was developed in ICE3 (Col) Research Group on Power
Quality and System Stability. It is was supported by the Young Researcher generation [9].
Program Virginia Gutierrez de Pineda of the Colombian Institute for the As contents, section two presents the basic theoretical
Science and Technology Development (COLCIENCIAS) and the Universidad aspects for understanding the paper; the third section presents
Tecnolgica de Pereira.
A. Herrera-Orozco is student of the Faculty of Engineering, in the
the proposed methodology. Section four is devoted to present
Masters program in Electrical Engineering, scholarship holder of the the test and the analysis of the results. Finally, conclusions of
Universidad Tecnolgica de Pereira, Pereira, Colombia. (e-mail: this research are highlighted.
arherrera@utp.edu.co).
S. Prez-Londoo is assistant Professor in the Faculty of Electrical
Engineering, Universidad Tecnolgica de Pereira, Pereira, Colombia. (e-mail: II. BASIC THEORETICAL ASPECTS
saperez@utp.edu.co).
In this section the basic aspects of theory required for
J. Mora-Flrez is assistant Professor in the Faculty of Electrical
Engineering, Universidad Tecnolgica de Pereira, Pereira, Colombia. (e-mail: performing of this work are presented. More detailed
jjmora@utp.edu.co).

978-1-4673-2673-5/12/$31.00 2012 IEEE


2

explanations are out of the scope of this document, but could tool for describing the dynamic behavior of complex physical
be found in the references provided. systems. In ATP, it can be used for describing the operation of
circuit components and of control components. It can also be
A. Static Load Modeling
used for generating signals or for analyzing measurements
Static load models are those which generally can be from the circuit. Finally, it can be used as the interface
represented as polynomial functions of the applied voltage [5] between ATP and outside programs [12].
[6] [7]. In general, load could be represented as (1).
III. PROPOSED METHODOLOGY
np nq
The proposed methodology is divided into three stages. The
k =0
(
S = Pn a pk V p .u . )b pk
(
+ jQn aqk V p.u .
k =0
)b qk
first is oriented to the power distribution circuit modeling
s.a (1) using ATP, which was modeled as a base case (Loads
modeled as Zcte using the elements of ATP). Additional
np nq
models were proposed, using modeling tools provided by the
a pk = aqk = 1 Models tool of ATP.
k =0 k =0

Where: The second stage is concerning to the fault simulation of


n Subscript that denotes the nominal value of the the circuit nodes considering different fault resistances (RF
variable. between 0.5 and 40), and different fault types [13]. Four
fault types were simulated: a) Single-phase to ground fault
a p aq Power participation coefficient.
(phase A), b) Phase to phase fault (phases A-B), c) Phase-
n p nq Maximum number of power participation phase to ground fault (phases A-B-ground) and d) Three phase
fault (phases A-B-C).
coefficients, (usually 3).
As a result of the simulation, a database is obtained for V, I
b p bq Characteristic power exponent. and Z at pre-fault and fault steady states at the generating
This model is commonly known as the ZIP load model substation (DG1 and DG2).
(constant impedance (Zcte), constant current (Icte) and constant As a last point, the third stage involves the comparative
power (Scte)) [7]. analysis of the load models proposed, through comparison of
On the other hand a special case is presented in (2), which the results of the simulation. Thus, a comparative analysis of
represents the exponential load model [5] [7]. different load models through graphs for observing the
differences between the base case (Zcte) and the loads models
proposed is performed. These graphs show the distance at
(
S = Pn V p.u . )b p
(
+ jQn V p.u . )b q (2)
which the fault occurs, which is measured with respect to the
main power substation (DG1), the fault resistance considered
In this case bp and bq are real values. If bp = 0 , the load in the simulation (RF) and the difference which is presented
between the models (error).
consumes the same power despite fluctuations in voltage The error considered for magnitude values comparison is
magnitude (Scte). If bp = 1 , the load consumes the same the relative error as presented in (3).
current despite fluctuations in voltage magnitude (Icte).
bp = 2 , the load maintains the same impedance despite X Base X Cal (3)
fluctuations in voltage magnitude (Zcte). The same ER =
X Base
consideration applies for the reactive power with bq .
B. Simulation Software Where X Base , is the value of the measured signal (V, I or
The software used for modeling and simulation of the test Z) of the base circuit modeled with constant impedance using
circuits is Alternative Transient Program (ATP) [10]. the modules of ATP. X Cal , is the value of the measured
ATP clearly offers a way to adapt the load model according signal (V, I or Z) considering different load models.
to the voltage variation on the power circuit, and is therefore a The comparison for angle values was made using
useful tool for modeling the power distribution system, where difference in degrees between the base case and the result
is crucial to have this type of models. using the respective load model.
Until now, the implemented power systems for fault
location were modeled as constant impedance load [11]. Using A. Load Modeling Proposal
ATPDraw, the values for the load parameters are introduced The proposed load model was coded using Models in ATP
as impedance and are normally estimated from the average and allows modeling the load according to ZIP polynomial
values and registers from customers, and also taking into model, or if desired also from exponential form depending on
account the nominal values of the power transformers. the need. This model uses one of the predefined Models of
Models is a general technical description language ATP (Type 94), Norton non-transmission version [12], [14].
supported by a simulation solver. It is a programming The data used for configuration of the load model in ATP
language targeted to time-domain simulation, and is used as a tool Models are:
3

- Snom: Load power [VA] with the load. The test power system was modeled using
- FP: Power factor of the load elements and the load Zcte from the ATP (Base Case). Next,
- Vnom: Nominal voltage [V] new load models at the same circuit, but considering those
- Vopi: Initial operating voltage [V] coded using the tool Models of ATP, to obtain the proposed
- Vbreak: Breakdown voltage of the load model in p.u. Its different load models (Zcte, Icte, Scte, 80%Icte plus 20%Zcte [5]
default value is 0.6 p.u as suggested in [15]. and the hybrid model circuit IEEE34 [8]).
- bp y bq: Parameters that define the type of model used, 0 Fig. 2(a) and 2(b), shows the test circuit modeled in
for Scte, 1 for Icte and 2 for Zcte [5] [7]. ATPDraw with loads modeled with elements of the ATP (base
- ap0, ap1, ap2: Are the values of the coefficients ZIP load case) and loads modeled with the tool of Models.
model between 0 and 1 which indicates the percentage
distribution of each type of load for the active power; i.e. that,
if one has a load placed with an active power distribution
DG2
Scte=10%, Icte=30% and Zcte=60%, these coefficients will be
ap0=0.1, ap1=0.3 and ap2=0.6. Ap0 is the coefficient for Scte,
ap1 for Icte and ap2 for Zcte.
- aq0, aq1, aq2: Are the values of the coefficients ZIP load
model between 0 and 1 which indicates the percentage DG1
distribution of each type of load for the reactive power. Aq0 is
the coefficient for Scte, aq1 for Icte and aq2 for Zcte.
If a hybrid model is required, exponential bp and bq values Fig. 2(a). Test circuit with loads modeled with elements of ATP.
can change; on the other hand, if the hybrid model is desired
ZIP bp and bq values are equal to 1 and change the values of
the model coefficients ZIP (ap0, ap1, ap2 y aq0, aq1, aq2).

Fig. 1(a), shows the data window load model coded using
the tool Models of ATP (Type 94), and Fig. 1(b), shows the
model in ATPDraw.

DG1
DG2

Fig. 2(b). Test circuit with loads coded using Models of ATP.

B. Automatic Fault Simulation


An extensive fault simulation was performed using an
automatic fault simulation tool developed by the research
Fig. 1(a). Fig. 1(b). group ICE3 [16]. This tool, through the interaction between
Fig. 1(a). Data window load model programmed using the tool Models of
ATP and Matlab software, can automatically simulate
ATP (Type 94) in ATPDraw. (b). Schematic model programmed into different faults, with different fault resistances and at nodes of
ATPDraw. a specific system, and thus obtain a set database of pre-fault
and fault steady state measurements of V and I in Matlab in
IV. TEST AND RESULTS both power substation (DG1 and DG2). In total, 1012 faults
were simulated for power system, for a cumulative total of
A. Test System 6072 faults that provide rms values of V and I in DG1 and
As test system, the 24.9 kV IEEE of 34 nodes power DG2.
distribution feeder is selected [8]. It corresponds to a circuit The simulation was performed on a HP Compaq 6000 pro
currently located in Arizona U.S.A. This circuit was modified SFF computer, with a processor Intel(R) core(TM)2 Quad
to include a distributed generation source (DG2) at node 840. CPU Q8400 @ 2.66 GHz and 4 GB RAM installed, with 32-
This last source has the same characteristics as the main bit operating system. The average time simulation of the
source (DG1), but has a lower power output. circuit with load programmed with Models tool for obtaining
It is also important to note that in this case, voltage descriptors was 2.976 hours for each circuit with a time step of
regulators are not included, and the transformer connected to a 0.1ms in ATP.
three phase load located at node 832, was reduced together
4

C. Analysis of Results lower than 0.35 degrees from the base case, presented errors
The most relevant results are shown in this analysis section. more pronounced with the load model corresponding to Scte,
followed by the Icte model, then the cases of the hybrid models
1) Analysis of Pre-fault Steady State and finally with errors approximately 0% in the case Zcte
The voltages (V), currents (I) and impedances (Z) have to model (As was expected for this last case).
be equal for all types of loads considered in pre-fault steady
state. From fig. 3 to 6, show the percentage error obtained
respect to the base circuit for DG1 and DG2.
For all graphs are taking the following symbolic
convention:
Circuit with load model for Zcte.
Circuit with load model for Icte.
Circuit with load model for Scte.
Circuit with hybrid load model with 80% Icte plus 20%
Zcte [5].
Circuit with hybrid load model that was taken from the
database IEEE34 circuit [8].

Fig. 5. Comparative graph for current magnitude of pre-fault in DG1.

Fig. 3. Comparative graph for voltage magnitude of pre-fault in DG1.

Fig. 6. Comparative graph for angle of current to pre-fault in DG1.

As presented in figures, the model with the lowest error


from the base case is Zcte, because the basis circuit is also Zcte,
but considering loads modeled with elements of ATP
predefined and non-programmed by tools Models.
The comparative graphs for the pre-fault currents (Fig. 5
and 6), both in magnitude and phase angle, show a behavior
similar to the tensions, but the difference is more remarkable
from the base case for the magnitude of currents, appearing
Fig. 4. Comparative graph for angle of voltage to pre-fault in DG1.
the same order as presented previously for the models,
According to Fig. 3, the differences shown by the five showing greater difference in the model of Scte.
models have errors less than 0.7% from the base case; i.e., Errors in estimation of the impedance as seen from the
have values slightly less than the value based on the value of power substation (Z) are a reflection of the reason for the
the magnitude of voltage and are constant in all settings. graphs of V and I, presenting the same or more pronounced
From to fig. 4, the differences presented in the angle are errors in both magnitude as phase angle, and also a similar
5

behavior to presented at the above figures.


This behavior is presented only of a single phase (phase A),
and similar behaviors are expected at the other phases.
Voltages and currents at DG2 present a similar behavior as
in DG1. On the values of Z, the angle has an error slightly
larger, but with the same behavior. Therefore, the graphs
belonging to pre-fault DG2 are not shown.

2) Analysis of Fault Steady State


This analysis shows the most representative graphs
considering significant differences in ATP models (Models)
respect to the proposed load as Zcte, using resistors and
inductors from the ATPDraw.

a) Single phase to ground fault (Phase A):


In fig. 7 to 10, the most important results for this fault type
are shown.

Fig. 8. Comparative graph for angle of current at fault steady state in DG2.

In fig. 8, it is observed that the differences between the


angles of the current of phase C in DG2 are especially
pronounced for the model of the load Scte, followed by the case
of Icte, after the hybrid model IEEE34 and the hybrid model
0.8Icte plus 0.2Zcte. The behavior is similar to the current
magnitude. This occurs for small values of RF, at
intermediate zones from the nodes of generation center DG1.
In case of DG1 considering the angle of the phase C
current, the error is less than DG2, and in this case the model
which presented the greatest difference is Scte, having a range
of percentages from -6 degrees from to +10 degrees to the
base case. Other models have lower errors range.

Fig. 7. Comparative graph for current magnitude at fault steady state in DG2.

The comparison of the magnitude of current in phase C


measured at DG2 is shown in Fig. 7, where is a considerable
error at the Scte load model, having a percentage of 125% from
the base case, for RF small with distances of about 25
kilometers; followed by the Icte load model, then the hybrid
model as proposed for the IEEE34 power system (known as
IEEE 34 model), next the hybrid model 0.8Icte plus 0.2Zcte
and finally Zcte load model. The above behavior is also
presented for DG1, but with a difference less than 35% for the
more critical case (Scte).

Fig. 9. Comparative graph for impedance magnitude at fault steady state in


DG2.
6

Fig. 11. Comparative graph for current magnitude at fault steady state in
DG1.
Fig. 10. Comparative graph for impedance angle at fault steady state in DG2.

Considering the magnitude of Z (Fig. 9), it is observed that


the model of the load with greater differentiation from to the
base case is Scte as noted earlier in the case of V and I. Some
models that also become relevant are those of Icte and hybrids.
The model that still has the lowest errors is Zcte.
In fig. 10, the difference in the angle of Z of phase C in
DG2 are shown, where the model Scte presents a pronounced
difference in the angle and it has a similar behavior to the
impedance magnitude.
These differences in measurements of V and I in DG1 and
DG2, due to the load model will directly influence the fault
locators, as these locators use the measurements at the power
substation (DG1 and DG2), for the fault estimation.

b) Phase to Phase Fault (Phases A-B):


In fig. 11 and 12, some of the most important results in case
of phase-phase faults are presented.
The fig. 11 shows the differences in magnitude of current
in DG1, having considerable errors in the cases of Scte, and Icte.
For DG2 was noted the same behavior with similar errors.
In fig. 12, some differences with high values of RF, short
distances, especially in Scte, Icte and hybrids models are
observed.
For values of Z were observed similar behavior to that Fig. 12. Comparative graph for angle of current at fault steady state in DG2.
already described, where the load model of Scte has the highest
differentiation. c) Phase to Phase-Ground Fault (Phase A-B-Ground):
Load models such as the Icte and the 0.8Icte plus 0.2Zcte Some of the most important results for phase to phase-
present in some cases greater deference in the magnitude and ground fault are presented in fig. 13 and 14.
angle of V that those presented by the Scte model.
7

Fig. 13. Comparative graph for impedance magnitude at fault steady state in
DG1.
Fig. 15. Comparative graph for voltage magnitude at fault steady state in
DG2.

Fig. 14. Comparative graph for impedance angle at fault steady state in DG1.

In this type of fault, the behavior is similar to past faults for


different models, where the most relevant model or greatest Fig. 16. Comparative graph for impedance angle at fault steady state in DG2.
difference from the base model is the Scte. This is because this
model has a linear behavior of power, while the Zcte model is a The errors in V magnitude (fig. 15), is are pronounced,
quadratic function of power. when there is a low RF at a distance far from the center DG1,
In some of the cases, Icte models and hybrids (0.8Icte - but in turn nearest to DG2, since DG2 is connected to node
0.2Zcte and IEEE34 model) show marked differences from 840 which is at a distance of 57,677 km of DG1. Models as
base case (errors greater than 20%). the Icte and hybrids show differences, though not as noticeable
in nearby nodes, and in other cases away from DG1 with high
d) Three-phase Fault (Phase A-B-C): RF.
Some of the most remarkable results for this type of fault The angle of Z shown in Fig. 16, has a similar behavior to
are shown in fig. 15 and 16. the value of the magnitude of V (fig. 15), but with a less
8

differences, also for the magnitude of Z has a similar behavior. [8] Distribution System Analysis Subcommittee. IEEE 34 Node Test
Feeder. [Online]. Available: http://www.ewh.ieee.org/soc/pes/dsacom
On the values of I, it was noted that with high RF and near /testfeeders/index.html
distances of DG1, Icte load models and hybrids show [9] C. Orozco-Henao. J. Mora-Florez, and S. Prez. A robust method for
differences, although they are less than 6% below the base single phase fault location considering distributed generation and current
compensation. Transmission and Distribution Latin America
case.
Conference. September, 2012 / Montevideo Uruguay.
[10] User Group. Rule Book Alternative transient program (ATP).
V. CONCLUSIONS [Online]. Available: http://www.emtp.org
[11] J. U. Nunes. A. S. Bretas. (2010). Impedance-Based Fault Location
The model coded using the tool Models in ATP, adequately Formulation for Unbalanced Primary Distribution Systems with
reproduces the behavior of the load for different fault types. Distributed Generation. Presented at the Int. Conf. on Power System
As demonstrates by simulations, the model that has the lowest Technology. 2010.
[12] Laurent Dub. Users Guide to Models in ATP. April. 1996.
error from to the base case is Zcte. This is because the base [13] J. Dagenhart. The 40-ohms ground-fault phenomenon. IEEE
circuit is also Zcte, but modeled using predefined elements of Transactions on Industry Applications. 2000; 36 (1): 30-32.
ATPDraw. Moreover, the proposed models reproduce [14] N. Watson and J. Arrillaga, Power Systems Electromagnetic Transients
Simulation. IET Power and Energy Series 39. Published by Institution of
adequately the circuit behavior and have relatively small Engineering and Technology, London, United Kingdom. 2007. ISBN: 0
errors in pre-fault conditions. 85296 106 5. Page(s): 68 73.
Additionally, for the different types of simulated faults it is [15] Siemens Energy, Inc. Program Operation Manual PSSE 32.0.5.
October 2010. Section: 5 10.
observed that the difference between the models is similar [16] Mora-Florez, J., Bedoya, J., Melendez, J. Extensive Events Database
from one fault to another. The biggest difference, in most of Development Using ATP and Matlab to Fault Location in Power
the cases, occurs in the Scte model, followed by Icte model, then Distribution Systems. Transmission & Distribution Conference And
Exposition: LA, 2006.
the hybrid model (0.8Icte - 0.2Zcte), next the hybrid IEEE34
and finally the smallest difference that present (0%
VII. BIOGRAPHIES
approximately) is the model of Zcte.
The impedance errors are a reflection of the reason of A. Herrera-Orozco was born in Santa Rosa de
errors in V and I, and therefore have more pronounced Cabal, Risaralda, Colombia in 1987. He received the
differences in both, magnitude and angle. B.Sc. degree in Electrical Engineering from the
Universidad Tecnolgica de Pereira (UTP), Pereira,
It is important to note that in DG2 is where the greatest Colombia, in 2010. He is currently a M.Sc. student
difference is presented, considering the measures of V and I in Electrical Engineering from the UTP and is
with different load models analyzed. It is because it is a researcher at the research group Calidad de la
Energa Elctrica y Estabilidad ICE3 (Col). His
generator with a lower power output, and also at its location
areas of interest are power quality, distribution
(N840) several loads are connected. system modeling and efficient use of energy.
Finally, differences in measures of V and I in DG1 and
DG2, considering the proposed load modeling, will directly S. Perez-Londoo received the B.Sc. degree in
influence the fault locators, since these use the measurements electrical engineering from the Universidad
Tecnolgica de Pereira (UTP), Pereira, Colombia, in
of V and I at the power substations to estimate the fault 2001, the M.Sc. degree in electrical engineering
distance. As a good estimation of the fault location is from UTP in 2005. Currently, she is a associated
extremely important, it must take into account the load professor of the Electrical Engineering School, at the
Universidad Tecnolgica de Pereira, Pereira,
modeling at the power system under study. Colombia. Her areas of interest are electric
machines, power system stability, fault analysis and
VI. REFERENCES soft computing techniques. S. Perez-Londoo is a
member of Research Group Calidad de la Energa Elctrica y estabilidad
[1] M. Bollen. Understanding Power Quality Problems: Voltage Sags and ICE3 (Col).
Interruptions. IEEE Press. 2000.
[2] Comisin de Regulacin de Energa y Gas CREG, Resoluciones CREG
070 de 1998, CREG 096 de 2000, CREG 084 de 2002, CREG 097 de J. Mora-Flrez received the B.Sc. degree in
2008. Electrical Engineering from the Universidad
[3] C. Gelling, Power delivery system of the future Power Engineering Industrial de Santander (UIS), Bucaramanga,
Review, IEEE Volume 22, Issue 12, Dec. 2002 Page(s): 712 Colombia, in 1996. The title M.Sc. in electrical
[4] Mora J. Localizacin de Faltas en Sistemas de Distribucin de Energa potency of the UIS in 2001, the M.Sc. degree in
Elctrica usando Mtodos basados en el Modelo y Mtodos de information technology at the Universidad de Girona
Clasificacin Basados en el Conocimiento. Espaa. 2006. Tesis (UdG), Girona, Spain, in 2003, and the Ph.D. degree
Doctoral. Universidad de Girona. ISBN 978-84-690-4513-8 in Information Technology and Electrical
[5] J. L. Aguero, M.B. Barbieri, and M.C. Beroqui,"Voltage depending load Engineering from the UdG in 2006.
models. Validation by voltage step tests", IEEE Power Engineering He currently is associated professor of the Electrical
Society General Meeting, 2006. Engineering School, at the Universidad Tecnolgica
[6] I. D. Serna, C. D. Ferreira, S. A. Martnez, M. F. Surez and G. Carrillo. de Pereira, Pereira, Colombia. His areas of interest are power quality, transient
Impact of static load models on the power distribution fault location analysis, protection, and soft computing techniques.
problem. Transmission and Distribution Conference and Exposition: J. Mora Flrez is a member of Research Group Calidad de la Energa
Latin America. IEEE/PES 2010. Page(s): 706 711. Elctrica y estabilidad ICE3 (Col).
[7] Taskforce, "Load representation for dynamic performance analysis of
power systems", IEEE Transactions on Power Systems, vol. 8, pp. 472-
482, 1993.

Das könnte Ihnen auch gefallen