21, rue dArtois, F-75008 PARIS http : //www.cigre.

org

(CIGRE-114)

CIGR Canada Conference on Power Systems Vancouver, October 17- 19,

Online Tracking of Voltage-Dependent Load Model Parameters

S. A. Arefifar, and Wilsun Xu, University of Alberta Canada

SUMMARY This paper presents a field-test verified method for continuous monitoring of voltage-dependent load model parameters. The basic idea of the proposed method is to use the movement of the Under Load Tap Changer (ULTC) as the source of voltage disturbance. Load parameters are estimated from the disturbance and for each taping activity. Since a ULTC often has many movements in any given day, the load parameters can be collected in a consistent, frequent and predictable manner and, consequently, online estimation and tracking of load parameters becomes feasible. The paper presents the implementation of the proposed method and its field measurement experiences, especially the characteristics of tap movements. A large number of field measurements presented in this paper have shown that the proposed method is a practical and effective method for determining load parameters online. KEYWORDS Load modeling, Tap changer, ULTC Movement, Power system measurement

arefifar@ece.ualberta.ca, wxu@ualberta.ca

1. INTRODUCTION
The loads in power systems play an important role in power system planning, control and stability analysis [1], [2]. Having reliable models of the loads is essential for the purpose of power system simulation studies, designing automatic control systems and optimizing their configuration [3]. Determining accurate load modeling has been a challenging problem for power system engineers for decades. Over the past 30 years, two types of methods have been proposed. The first type uses the information of load composition such as the NAICS codes and associated computer models to compute the parameters [4], [5]. This approach is theoretically sound but has a number of implementation problems. For example, how do we classify the loads into different types that have similar voltage dependent characteristics? What are the valid models for different load types and at different times and seasons? Such difficulties have led to the development of the second type of methods in recent years direct measurement of the load parameters [6]-[8]. The method, however, has to rely on voltage disturbances to be successful. Since it is not possible to create many intentional voltage disturbances due to power quality concerns and costs, the direct measurement method can only be performed under special arrangements. This constraint has made the method expensive to use and has limited its application to spot-checking the characteristics of certain loads. This lack of representativeness has severely limited the use of such methods. One type of voltage disturbance, which has a high frequency of occurrence is the disturbances caused by the operation of the Under Load Tap Changer (ULTC) transformers. Reference [6] performed extensive load tests on the BC Hydro system and found that the load responses due to the UTLC disturbances are sufficient to capture the voltage-dependent load characteristics. However, the ULTC disturbances were created intentionally in that work so its results also suffer from the problems of lack of representativeness. Based on the above experiences, this paper proposes a method that can measure voltage-dependent load parameters at any given day and on-demand. This method uses the natural tap movements as the source of voltage disturbances. Although the concept is simple, it is important to verify if the natural tap movements can indeed yield disturbances with sufficient magnitude and frequency that can be used to determine the load parameters adequately and continuously. Through extensive field measurements, the above issue has been settled. The results have shown that the disturbances caused by natural tap movement are quite suitable for load parameter estimation.

2. CHARACTERISTICS OF ULTC MOVEMENT

In a power transmission system, load refers to the collective power demand of the various electricity users served by a feeder or by a substation. The proposed measurement scheme for calculation of load model parameters is shown in Fig. 1. In the figure, the load parameter monitors are placed at the terminals of the feeders connected to the substation MV bus. These monitors can also be relays or power measurement units commonly available in substations as long as they can measure the required voltage and current and are capable of performing the parameter estimation algorithms to be disclosed later. The measurement devices have been installed in four different distribution substations to capture the tap movements occurring at each substation transformer during one week.
Substation Bus

Transmission System
I V

Feeder 1

Distribution network
Feeder 2 Feeder 3

Load Parameter Monitor Results

Fig. 1. The proposed load model measurement system

The data has been captured from different substations in Alberta, Canada. These substations are 15 kV substations supplying commercial and residential loads. The loads are classified as mostly commercial, which has roughly 70% commercial loads, mostly residential which has roughly 70% residential loads, and the substations that have almost equal amounts of commercial and residential loads. The list of transformers and their load types are presented in Table I. As shown in Fig. 1, each transformer supplies several feeders and the measurements are taken at the secondary side of the transformers. Therefore, the collection of feeders connected to each transformer is considered as a load.

TABLE I. LIST OF TRANSFORMERS AND THEIR LOADS Transformer Transformer load Substation No. types 1 Commercial and residential One 2 Commercial and residential 3 Commercial and residential 4 Commercial and residential 5 Mostly Residential Two 6 Mostly Residential 7 Mostly Residential 8 Mostly Commercial 9 Mostly Commercial Three 10 Commercial and residential 11 Mostly Commercial 12 Mostly Commercial 13 Mostly Residential Four 14 Commercial and residential

2.1 Voltage Disturbance Characteristics According to the manufacturers of the ULTC, the total range of regulation and the size of individual steps for tap movements are most often specified as 10% voltage in 33 steps of 0.625% voltage per step [9]. The exact information about the transformers in the substations is not always available to detect the operation of ULTC and adjust the recordings. Thus, to remain in a safe margin, the step change could be assumed as 0.5% of the initial voltage. In a 15kV line to line voltage, or an 8.4 kV line to ground voltage, the tap movement step will be around 42 volts. It should be noted that this value has been used only to detect the operation of ULTC; however, for load parameter estimations, the actual step change has been calculated from the real recorded waveforms. Fig. 2 shows the minimum, maximum and the average values of the voltage disturbance, caused by the tap movements, during the period of measurement. It is evident that the average magnitude of step changes in voltage is close to one percent and is almost the same for all the transformers in different substations.
2 1.8 Voltage Step Change (%) 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 5 Transformer No. 10 15 Average Tap change Values

Fig. 2. The minimum, maximum and average tap movement effect on voltage for different transformers

2.2 Frequency and Time of Tap Movement Fig. 3 shows the average number of tap movements per day for different transformers that could be used for load monitoring. The average for weekdays and weekends are plotted separately. The figure also shows the distribution of the tap movement occurrence time for transformer one.
20 18 Average Number of Tap Movements per Day 16 14 12 10 8 6 4 2 0
Number of Tap Movements 5

WeekDays Weekends

4.5 4 3.5 3 2.5 2 1.5 1 0.5 0

5 Transformer Number

10

15

10 Hour of the Day

15

20

(a)

(b)

Fig. 3. (a). The average number of tap movements for transformers, (b). The distribution of tap movement time for transformer one

As can be seen there is usually enough tap movements per day for online tracking of load parameters. Fig. 3 also shows that the tap movement can occur anytime during 24 hours. This could be very helpful in having the load model parameters in an online and continuous manner. For the hours of the day which we do not have any tap movements, the load model parameters could be estimated through interpolation methods. 3. ALGORITHMS FOR LOAD PARAMETERS ESTIMATION In this study the step changes in voltage caused by the operation of ULTCs has been used for load parameter estimations. Based on field experiences, the required time for tap movement is usually a few cycles. Fig. 4 shows an example of the variation of voltage, active power and reactive power due to a tap movement.
1.01 V(pu) 1 0.99 0 1.02 P(pu) 1 0.98 0 1.05 Q(pu) 1 0.95 0

50

100

150

200

250

300

50

100

150

200

250

300

50

100

150 Cycle s

200

250

300

Fig.4. A step change in voltage and its effect on the power

It can be seen that the step change in the voltage creates disturbances in the corresponding active and reactive power of the transformer. The voltage-dependent load characteristics, also called (voltagedependent) load models, denote the real and reactive power demands of the loads as affected by their

supply voltage. The main focus of this paper is to investigate the voltage dependency characteristics of the loads in steady state conditions; therefore, the static load models have been used for load estimations. Having the load model parameters for one specific model (exponential) we can estimate the parameters for different static load models such as the ZIP model [4]. Traditionally, the voltage dependency of load characteristics has been presented by the exponential model:
P V a =( ) P0 V0

Q V b =( ) , Q0 V0

(1)

where P and Q are the active and reactive components of the load when the bus voltage magnitude is V. The subscript 0 identifies the values of the respective variables at the initial operating conditions. Assuming P0 and V0 as the initial or pre-tap movement active power and voltage, and assuming P and V as post-tap movement values, for a set of voltage and power values measured from a disturbance, we can calculate a using the following formula (similar for b):
a = log( Ppost Ppre ) log( V post V pre ),

(2)

where the subscript "pre" ("post") refers to the pre- (post-) disturbance values. In order to calculate the load model parameters, we need the steady state active and reactive power and the corresponding voltage. Since the whole tap change occurs in around 5 cycles, and the magnitude of the step change, based on field tests, is at least 0.5%, a moving window on voltage of phase A, can be used for starting point detections. The field tests show that the transients in active and reactive power caused by a tap movement is very small and dies out in, at most, one second. Thus, the pre- and post-tap values are considered as the average of 5 cycles before and the average of 5 cycles after one second following the starting of the tap movement, respectively. The averaging is done to reduce the effect of noise or transients in the measurements. 4. FIELD MEASUREMENT RESULTS The method has been applied to the field data taken from the substations mentioned in Table I. The data recording procedure, effect of voltage step change and basic load parameters are presented here. 4.1 Data Recording Procedure Two different recording procedures have been followed simultaneously to record the data. These procedures are continuous recording and tap movement recording. The continuous recording captures 12 cycles of the three phase voltages and currents waveforms every 12 seconds. The other data recording procedure captures the tap movement instant waveforms based on the variation of voltage in phase "A". Checking the voltage of phase A every second and comparing it to the voltage at the previous second, can help us to detect the tap movement. 4.2 Grouping Similar Loads All the measured transformer loads in substations have been grouped in three as shown in Table II.
TABLE II. CONSTRUCTED GROUPS BASED ON LOAD TYPES
Group I II III Members F.1, F.2, F.3, F.4, F.10, F.14 F.5, F.6, F.7, F.13 F.8, F.9, F.11, F.12 Description Commercial and residential Mostly residential Mostly commercial

The load profiles of the third group and the calculated load parameters are plotted in Fig. 5. This group of loads consists of mostly commercial load types. As can be seen in Fig. 5, the load profiles of transformers in group III are very similar, which makes grouping them together reasonable. Fig. 5 also, shows the calculated load model parameters for the members of group III. Each plotted value is the average of the calculated parameters during the corresponding hour of the day. Since the load

types and load profiles of the transformers are similar, we expect to have similar model parameters for the loads in each group, which can be seen from Fig. 5.
60 50 P(MW) 40 30 20 10 0 5 10 15 20 25

3
F. F. F. F. 8 9 11 12

F. 8 F. 9 2 "a" F. 11 F. 12

10

15

20

25

25 20 Q(MVar)

12 10 8 "b"
0 5 10 Time (24H) 15 20 25

15 10 5

6 4 2 0 0 5 10 Time (24H) 15 20 25

(a) (b) Fig. 5. (a). Load profiles, (b). Estimated load parameters of different transformers in group III.

The load profiles and calculated load parameters for all the transformers have been studied and it is seen that the load profiles of each group are similar with the calculated load parameters being almost the same for feeders of the same group. This study shows that the tap movement can happen at any time during the day and that the load parameters vary during 24 hours. It is also seen that the load parameters vary based on the type of loads, with similar loads having similar load profiles and load parameters. 4.3 The Load Parameters as affected by Voltage Step Change The load parameters are calculated using the step change caused by the ULTC. The effect of voltage step changes magnitude on the calculated load parameters is investigated in this part. For this purpose, the voltage step changes versus time of tap movement occurrence and versus the calculated load model parameters are studied for all the transformers. As an example, the parameters related to one of the commercial transformers are shown in Fig. 6.
3 "a" 2 1 0 -2 10 "b" 5 0 -2 2 DV(% ) 0 -1.5 -1 -0.5 0 0.5 1 1.5 2

-1.5

-1

-0.5

0 DV (%)

0.5

1.5

-2 0

10

15 Time (24H)

20

25

Fig. 6. Load parameters versus voltage step change of a commercial transformer

It can be seen that most of the step changes are around one percent of the initial voltage and the stepup and down tap changes do not have significant effect on the calculated load model parameters for all the measured transformers. Furthermore, the magnitude of voltage step change does not affect the calculated load parameters and the step change in voltage does not depend on the time of tap movement occurrence. This finding is very important and is in agreement with the findings of [6]. Through intentionally created large voltage changes, that work has shown the load parameters estimated are very close no matter how large or small the voltage disturbances are.

4.4 Typical Load Model Parameters of the Groups As explained in previous sections, the measured transformers' loads are grouped into three different types. To have a comparison between the load parameters of different groups, in this section, the average of load parameters for each group is presented in chart format. For this purpose, the 24 hours of each day have been split into 6 equal sections and the average for each part is shown in Fig. 7.
1.6 Commercial and Residential 1.4 1.2 5 1 Mostly Residential Mostly Commercial 7 Commercial and Residential 6 Mostly Residential Mostly Commercial

4 0.8

3 2 1 0

0.6 0.4 0.2 0 0 to 4 4 to 8 8 to 12 12 to 16 16 to 20 20 to 23

0 to 4

4 to 8

8 to 12

12 to 16

16 to 20

20 to 23

Hour of the Day

Hour of the Day

(a) (b) Fig. 7. The typical calculated a and b for different groups of loads

The graphs show that the residential feeders have the maximum and the commercials have the minimum values for the model parameters. The average model parameters of the loads, for the group which has both commercial and residential types of loads, resides between the other two groups. It is seen that the load parameters for different types of loads are not similar. 5. CONCLUSION This paper has presented a practical and effective method for online monitoring of voltage-dependent load model parameters. The method has been verified through extensive field test results. The data collected from the tests have been analyzed to show the various characteristics of tap movements. The main advantages of using natural tap movements for load parameter estimation are summarized as follows: The tap movements occur on a daily basis, so load parameters can be collected in a consistent, frequent, and predictable manner, They cause a step voltage change. If there are no other disturbances occurring at the same instant which is commonly the case, load parameters can be estimated using simple and reliable algorithms, Pre-established disturbance signatures uniquely suitable to detect them can be used to detect and capture these disturbances, Since there are many of such disturbances, one can afford to miss some of them because of various uncertainties. The selected disturbances, thus, have a higher quality. This rigorous selection process improves the reliability of the estimated results. System operators can create these disturbances intentionally with a desired disturbance severity (i.e. the amount of DV produced by the tap changer). As a result, the proposed technique can estimate load parameters on-demand. Finally, there is no need to verify the results as the load parameters are actually measured from step-change disturbances. They are not estimated with sophisticated algorithms using complex disturbance waveforms. Since the load parameters can be obtained many times a day, online tracking of load parameters becomes possible and real-time power system applications such as dynamic security assessment can use the measured load parameters as input.

6. REFERENCES
[1] [2] [3] [4] [5] [6] [7] IEEE Task Force on Load Representation for Dynamic Performance, Bibliography on load models for power flow and dynamic performance simulation, IEEE Trans. Power Syst., vol. 10, no. 1, pp. 523538, Feb. 1995. Renmu, H.; Ma Jin; Hill, D.J.; , "Composite load modeling via measurement approach," Power Systems, IEEE Transactions on , vol.21, no.2, pp. 663- 672, May 2006. Knyazkin, V.; Canizares, C.A.; Soder, L.H.; , "On the parameter estimation and modeling of aggregate power system loads," Power Systems, IEEE Transactions on , vol.19, no.2, pp. 1023- 1031, May 2004. Vaahedi, E.; Fl-Kady, M. A.; Libaque-Esaine, J. A.; Carvalho, V. F.; , "Load Models for Large-Scale Stability Studies from End-User Consumption," Power Systems, IEEE Transactions on , vol.2, no.4, pp.864-870, Nov. 1987. Dong Han; Jin Ma; Ren-mu He; Zhao-yang Dong; , "A Real Application of Measurement-Based Load Modeling in Large-Scale Power Grids and its Validation," Power Systems, IEEE Transactions on , vol.24, no.4, pp.1756-1764, Nov. 2009. Xu, W.; Vaahedi, E.; Mansour, Y.; Tamby, J.; , "Voltage stability load parameter determination from field tests on BC Hydro's system," Power Systems, IEEE Transactions on , vol.12, no.3, pp.1290-1297, Aug 1997. Byoung-Kon Choi; Hsiao-Dong Chiang; Yinhong Li; Hua Li; Yung-Tien Chen; Der-Hua Huang; Lauby, M.G.; , "Measurement-based dynamic load models: derivation, comparison, and validation," Power Systems, IEEE Transactions on , vol.21, no.3, pp.1276-1283, Aug. 2006. Ma Jin; Renmu, H.; Hill, D.J.; , "Load modeling by finding support vectors of load data from field measurements," Power Systems, IEEE Transactions on , vol.21, no.2, pp. 726- 735, May 2006. Holladay, D.W.; Grigg, C.H.; Dallman, B.D.; , "Equipment Performance Study on Voltage Regulators," Power Delivery, IEEE Transactions on , vol.23, no.4, pp.2496-2512, Oct. 2008

[8] [9]

BIOGRAPHY
Seyed Ali Arefifar (S06) received his B.Sc. and M.Sc. degrees in Electrical Engineering from Isfahan University of Technology, Isfahan, Iran in 2001 and 2004, respectively. He is currently pursuing his Ph.D. program in Electrical and Computer Engineering at the University of Alberta, Canada. Wilsun Xu (M90, SM95, F05) received Ph.D. from the University of British Columbia, Vancouver, Canada in 1989. He worked in BC Hydro from 1990 to 1996 as an engineer. Dr. Xu is presently a professor at the University of Alberta. His main research interests are power quality and harmonics.