load parameters effect on relays performance for islanding detection

1

Load Parameters Effect on Relays Performance for Islanding Detection

Amin Ghaderi, Mohsen Kalantar

Iran University of Science and Technology Iran

Keywords: Distributed Generation, Islanding Detection, Frequency and Voltage Relay, Anti-Islanding Protection, Non-detection Zone.

Abstract In previous works on islanding detection, it is assumed that the worst case condition for a load to encounter islanding condition is RLC load. As a result, most of islanding detection methods used RLC load as a typical load model. In this paper we used a load model which its active and reactive powers are dependent to frequency and voltage. In this work a constant-power DG designed to work at unity power factor. With different load parameters the efficiency of relays is examined. A general formulation was derived and compared to simulation results. The worst cases for both frequency and voltage detection of islanding detection were presented. Introduction Recently Distributed Generation (DG) gained a lot of momentum, due to the great technological advantages, market regulation, and environmental benefits. Inverters are used to interconnect most of distributed generations to the grid, because the energy that distributed generations produce could not directly supply the network [1]. Despite benefits of DGs, they have brought many technical challenges [2]. The most important

challenge is whether the DG’s control and protection system has the ability to detect islanding condition. Islanding (loss of main) is a state where the main power grid is disconnected, while DG is energizing the load. Such operation condition brings threats to reliability of grid and safety of people and equipments. Therefore the ability of detecting the islanding condition should be one of the main functions of DG-inverter controller [3]. Generally, there are two methods for detecting islanding condition. Firstly, passive methods, which monitor voltage and current signals of DG terminal passively, extract several parameters, and compare them with some predefined thresholds. Practical examples of such methods are Voltage and frequency relays [4]. Although these methods are simple and practical, they have questionable results due to their improper Non-Detection Zone (NDZ). NDZ is a concept that defines as the range of power mismatch values which an islanding detection method would fail. In this paper we used NDZ as a criterion to evaluate voltage and frequency relays performances. Another method for islanding detection is active methods, which gain most of recent attentions

10-E-CAP-1239


25th International Power System Conference

2

on islanding issues. Even though active methods have negligible NDZ, they usually disturb power quality. SMS method and AFD method are some famous active methods [5]. New prevailing methods are the combination of active and passive methods which conceal both methods’ drawbacks. These methods are called hybrid methods [6]. Loading of a DG when encounters to the islanding condition, could affect the islanding detection method, because it is the loading which demonstrates the range of deviation in frequency and voltage after the islanding occurrence. A few publications devoted to evaluation of loading parameters on relaying system. In 2007, Feritas etal [7] examined the load models influences on the performance of a surge relay for detecting islanding. They scrutinized three voltage dependent load models, constant power, constant current, and constant impedance model. The results show different behavior of relays facing different kinds of loads. They also provided a worst case scenario for such load models in islanding detection. The previous publication was completely devoted to synchronous DGs. In 2009 Zeineldin etal [8], analyzed the effects of frequency dependent loads on the performance of a Sandia Frequency Shift relay, in detection of islanding for inverter based DGs. The results show a great influence of frequency dependent load, on the relay performance. Although other researchers worked on the same area for inverter based DGs, they could not come up with a worst case load condition that implies the most sever loading that must be detected with an islanding detection relay. Two major control methods have been widely used for interconnecting DG to the main grid: Constant Current Control, and Constant Power Control. Recently, the effect of DG interface control was studied and turned out that constant power control has a bigger NDZ and so it is the most conservative condition in examining islanding condition [9].

The main objective of this paper is to evaluate two most used relays, frequency and voltage, when different load parameters are used. In this paper we are going to examine the effect of voltage and frequency dependent load, on the relays which should detect the islanding detection. The paper organized as follows: Section II shows the load model used in the study. Section III introduces the NDZ concept which we used to examine the relay’s efficiency. Section IV is dedicated to indicate the system under study. Results of different load parameter simulation are showed in section V. Section VI validates the result with mathematical formulation. Section VII is dedicated to the worst case set of parameters for islanding detection based on NDZ. Conclusion is presented in section VIII. Load Model In this paper we used static load model. Static load is defined as a load in which both active and reactive powers are dependent on voltage and frequency variations. For presenting load variation on voltage and frequency several methods exist. Powers could be function of voltage, both exponentially and polynomial. Frequency variation multiplied by a factor, uses for load frequency dependence. This exponential static load is displayed below.

(1)

(2)

Where and are load real and reactive power, and and are rated real power, reactive power, and rated voltage, respectively. is the frequency variation.

and are constants used to show the relation between powers and voltage and frequency. Non-Detection Zone The non-detection zone is a graphical tool to evaluate the relays efficiency for islanding



3

detection. The method consists of active power-reactive power mismatch window drawn based on frequency and voltage relay settings. Any possible power mismatch can be shown in this graph. When a mismatch point relies within the NDZ boundaries, the relay cannot detect the islanding. Well-designed islanding detection methods always have negligible NDZ. Fig.1 shows a sample NDZ. Where denotes power mismatch (a positive value shows surplus in power). The NDZ boundary could be calculated based on voltage and frequency setting, load model, and inverter controller model

Fig1. A typical NDZ.

System under Study System under study consists of a distribution grid which is shown by a source behind impedance, a static load, and a DG. DG modeled as a battery for power source, and a three leg constant power inverter. A series reactance used as a filter in the inverter output. The inverter provides 100kW real power constantly, in unity power factor. The system under study is shown in Fig.2.

Fig2. System under Study.

The inverter constant power control system is shown in Fig.3. It has two sets of controllers, one for regulating powers and one for regulating currents. Real and reactive powers are measured and compared with reference values. Because inverter must generate unity power factor, the reactive power reference sets to zero. The current reference for synchronous frame will be produced, and compared with the synchronous frame current, measured from the circuit. The outputs of the control circuit are

(modulation ratio) and (phase displacement). They introduce to Sinusoidal Pulse Width Modulator (SPWM), in order to produce switching signals.

Fig2. DG interface controller.

Load parameters were changed independently, in order to examine their effects on voltage and frequency, and therefore to evaluate relays behavior. We changed four parameters, and

, in different loading condition and monitored the differences. For both real and reactive power, six mismatch cases were examined. Table I shows these cases. In some simulation, reactive and active mismatch occurs simultaneously; otherwise just real or reactive power experiences mismatch. It should be noted that the load assumed to have quality factor equal to 1.8 ( ). In section VII, it is indicated that quality factor has a significant effect on results.



4

TABLE I

Loading mismatch values for different case studies Case P1 P2 P3 P4 P5 P6

-25 -16 -7 8 14 26 Case Q1 Q2 Q3 Q4 Q5 Q6

-5 -3.33 -1.66 1.66 3.33 5

Simulation Results The system designed in the previous section was implemented in PSCAD/EMTDC and the results for different load parameters and mismatch power were extracted. The results show whether in this condition the relays are able to detect islanding or not. The islanding occurs once the breaker, “S”, in Fig.1 opens at t = 0.5 second. Voltage relays setting are 1.12 and 0.88 p.u. for upper and lower thresholds respectively. And settings for frequency relays, are 50.3 and 49.7 for upper and lower threshold respectively [10]-[11]. Based on Table I and load parameter variation, a lot of cases were studied.

Change in concurrently In this section we varied Np and Nq equally, in order to examine some familiar loads. For example constant power load, constant current load, and constant voltage load could be represented with Np=Nq=0, 1, 2 respectively. It should be noted that in these cases, loads are not function of frequency. Figure 3 shows the effect of these loads on. The case (P3) in Table I used for this simulation.

Fig3. Load terminal voltage waveform during the

islanding condition simulated at 0.5second.

It could be noticed that constant power load has a smallest NDZ, and constant impedance load has the biggest NDZ. That is because with a similar power mismatch, the constant impedance’s voltage variation is very smaller than the one with constant current and power. That is why in standards, we are obliged to use the constant impedance load model (RLC), as the most conservative case [10].

Voltage dependence of load active power (effects of Np) To evaluate the effect of Np, we used mismatch cases (P2) and (Q3) simultaneously. Because we want to examine the frequency relay performance, we should set one of the factors, , or in the load model. So we set Fig.4 shows frequency and voltage waveform for different amount of Np, before and after the islanding occurs.

Fig4(a). Load terminal voltage waveform during the

islanding condition simulated at 0.5second, with different Np.

Fig4(b). Load terminal voltage frequency waveform

during the islanding condition simulated at 0.5second, with different Np.

The results evidently demonstrate that, as the Np decreases the voltage and frequency tend to change wider. As a result it could be seen that when Np increases in a load model, the



5

islanding will be harder to detect. Table II illustrates this idea, by comparing the amount of active power mismatch needed for different values of Np, to detect islanding.

TABLE II

Real power mismatch needed to exceed voltage threshold as a function of Np

Np 1 1.5 2 3 4.5 Over

Voltage(%)

14 21 29 37 80

Under Voltage(%)

-9.2 -13 -17 -25 -34

It is obvious that when real power becomes more dependent on voltage, the voltage could harder exceed the defined thresholds, and so the islanding might not be detected. Voltage dependence of load reactive power (effects of Nq) The parameter Nq, indicates how much reactive power depends on voltage. To examine the Nq’s level of effect on voltage and frequency variation, we used power mismatch cases (P4) and (Q2). Because we want to examine the frequency, we should also set one the factors, , or in the load model. So we set . Fig.5 shows frequency and voltage waveform for different amount of Nq, before and after the islanding occurs. Due to the power mismatch, these parameters change to satisfy the generation- load equilibrium. During the simulations, Nq is changed while the other parameters remained fixed.


islanding condition simulated at 0.5second, with different Nq.


during the islanding condition simulated at 0.5second, with different Nq.

Voltage variation mostly depends on the real power mismatch, and that is why there is negligible change in voltage signal. On the other hand frequency will change, although slightly, with variation of Nq. As you can see, the bigger the Nq, the smaller the change in frequency. Therefore, like the case of Np, when Nq increases islanding becomes harder to detect. Frequency dependence of load active power (effects of Kpf) Kpf shows the level of active power’s dependence on frequency variations. In order to examine the effects of this parameter on voltage and frequency relays for detecting islanding, we used mismatch cases (P3) and (Q6) concurrently. Np and Nq were fixed on constant impedance values (Np=Nq=2). Kqf is set to 1. Islanding condition occurs at t=0.5 , and results are shown in Fig.6. In order to satisfy the balance necessity of inverter generated power and load utilization power, voltage and frequency change.


islanding condition simulated at 0.5second, with different Kpf.



6


during the islanding condition simulated at 0.5second, with different Kpf.

These graphs indicate that Kpf has a considerable effect on both voltage and frequency waveforms. As you can see in Fig.6(a), once Kpf increases, the voltage change more, and as a result NDZ becomes smaller. That is because with an equal amount of power mismatch, voltage of a load with bigger Kpf, more likely to exceed the relay voltage thresholds. Similarly on frequency, as it is shown in Fig.6(b), when Kpf increases, the load frequency is more probable to exceed the thresholds. The voltage variation, in this case, depends on frequency variation, since active power depends on frequency. It can be concluded that, voltage deviation is dependent on both active and reactive power mismatches. Frequency dependence of load reactive power(effects of Kqf) This parameter shows the relation between load reactive power and frequency. Like previous parameters, we change Kqf as the primary factor, while other parameters remain constant (Np=Nq=2, Kpf=0). The voltage and frequency performance was examined and results have been shown here. Loading is chosen on a way that both real power and reactive power experience the mismatch. Both reactive and active powers were taken in to account by using mismatch case (P6) and (Q2). When islanding occurs, frequency and voltage should both change according to mismatches to satisfy load-generation equations. This changes could be seen in Fig.7.


islanding condition simulated at 0.5second, with different Kqf.


during the islanding condition simulated at 0.5second, with different Kpf.

It is obvious that increasing in Kqf leads to harder detection of islanding. In other words, when Kqf decreases, the frequency variation increases, and it is more possible to exceed the lower or upper settings of frequency relay. On the contrary voltage waveform was not affected from voltage waveform. It is because voltage only change by real power mismatch (because Kpf=0). Worst Case Scenario As you observed in last section, variation in load parameters has major effects on relays efficiency. Power system protection engineers always design a worst case scenario to examine the performance of relays. So far it believed that a simple RLC load leads to worst conditions. In this section we compare a simple RLC load with our worst case load. As mentioned before, when Np, Nq, and Kqf increase, and Kpf decreases, NDZ area increases in size. It means that the variation in voltage and frequency become smaller, and it is less probable for them to transgress relays setting values. For this reason, we set load



7

parameters to the maximum values we studied in this paper (Np=Nq=4, Kpf= -5, Kqf=5) As you can see in Fig. 8, RLC load has a big NDZ. However our worst case load parameters load, has a much bigger one. In other words, in same power mismatches before islanding, the probability of exceeding relays setting in an RLC load is bigger than our worst case scenario. It should be noted that the RLC load used for simulation, has the same characteristic of static load, which are the inductive, capacitive, and resistive load power.

Fig8(a). Comparison between voltage waveforms of worst case scenario load parameters and traditional

RLC load model after islanding.

Fig8(b). Comparison between frequency waveforms of

worst case scenario load parameters and traditional RLC load model after islanding.

Formulation of Islanding with Static Load Once islanding occurs, the load just could be supplied by the DG. Therefore for calculation of voltage, frequency, and other parameter after islanding, we should replace the load powers with inverter output powers in (1) and (2). The general formulation could be seen in (3) and (4).

(3)

(4)

Where and are reactive and active power mismatches, relatively. For better calculation, we should divide the reactive load power to an inductive and capacitive part. As a result we could also take quality factor of the load in to account. We assumed that reactive load is an simple LC load, which helped us to find a simpler formulation. Before islanding occurs the reactive equations are:

=

Where and are capacitive and inductive load’s reactive power, and is utility frequency. When islanding occurs, the frequency changes until it reaches the resonance frequency ( ). So

from (5) and (6), the ultimate frequency after islanding could be derived from (7).

(7)

=

(8)

Where is the final frequency after the islanding, is the load factor, and is the reactive mismatch power, per unit. Using (8), we can also show the effect of quality factor on



8

voltage performance. The NDZ for reactive power can be calculated, assuming 49.7 Hz and 50.3 Hz as the lower and upper setting of frequency relay, and equation (8). Results are

and for upper and lower setting of frequency relays, relatively. The NDZ for real power is influenced by both frequency and voltage, as you can see in (9).

(9)

(10)

By using (9) and (10), the frequency and voltage variation after islanding could be calculated. These equations show that active power NDZ also depends on reactive power mismatch. The upper and lower settings of voltage relay in distribution are 1.1 and 0.88 p.u. , respectively. Based on different amount of Np and is shown in Fig. (9) and (10). In Fig. 9(a), the results of Np on NDZ are shown. When Np increases the NDZ also increases significantly. For example the NDZ for Np=3 is about three times bigger than Np=1. The results completely correspond with the simulation results in section IV. Similarly in Fig. 9(b), the results of Kpf are shown. The NDZ boundary slope depends on Kpf. In other words, when Kpf decreases, NDZ includes bigger amounts of active power mismatches. Therefore the results obtained in section IV are similar to what is shown in Fig. 9(b). However, the total area of NDZ does not change.

Fig9(a). NDZ for different values of Np and Kpf=2.5.

Fig9(b). NDZ for different values of Kpf and Np=1. Conclusion In this paper we scrutinized different load parameters to examine their effects on frequency and voltage relays. The software PSCAD/UMTDC is used to simulate different set of load parameters. The static load model has four different parameters to show the load active and reactive power dependence on voltage and frequency. The results indicate that, relays performance highly relies on load parameters. A worst case scenario set of load parameters is extracted from simulation results, and its NDZ compared with traditional RLC load’s NDZ. The comparison shows the severity of our worst case parameters. A general formulation for voltage and frequency variation after islanding is derived and substantiates the results obtained from simulation.



9

References 1- J.A.Pecas Lopes, N.Hatziargyriou, J.Mutale, P.Djapic, and N. Jenkins, “Integrating distributed generation into electric power systems: A review of drivers, challenges and opportunities,” Elsevier Electric Power Research, vol. 77, pp. 1189-1203, 2007. 2- S.conti, “ Analysis of distribution network protection issues in presence of dispersed generation,” Elsevier Electric Power Research, vol. 79, pp. 49-56, 2009. 3- W .Xu, K.Mauch and S.martel “An Assessment of Distributed Generation Islanding Detection Methods and Issues for Canada,” CANMET Energy Technology Centre, Natural Resources Canada, Report CETC-Varennes 2004-074 (TR), July 2004. 4- X.Zhu, C.Du, G.Shen, M.Chen, and Dehong Xu, “Analysis of the Non-detection Zone with Passive Islanding Detection Methods for Current Control DG System,” IEEE Conf. on Applied Power Electronics, Feb. 2009. 5- M.E.Ropp, M.Begovic, and A.Rohatgi, “Analysis and Performance Assessment of the Active Frequency Drift Method of Islanding Prevention,” IEEE Transaction on Energy Conversion, vol.14, no.3, pp. 810-815, September 1999. 6- P.Mahat, Z.Chen, , and B. Bak-Jensen, “A Hybrid Islanding Detection Technique Using Average Rate of Voltage Change and Real Power Shift,” IEEE Transactions on Power delivery, vol. 24, no. 2, pp. 764-771, April 2009. 7- W. Feritas,W. Xu, J. Viera, “Characteristics of Vector Surge Relays for Distributed Synchronous Generator Protection,” Elsevier Electric Power Research, vol. 77, no. 2, pp. 170-180, 2007. 8- H. Zeineldin, and A. Salma, “Impact of Load Frequency Dependence on the NDZ and Performance on the SFS Islanding Detection Method,” IEEE Transaction on Industrial Electronic, Oct 2009. 9- H. Zeineldin, E. El-Saadany, M.Salama, “Islanding detection of inverter-based distributed generation,” IEE Proceeding

Generation, Transmission and Distribution, vol.153, no.6, pp. 644-652, November 2006. 10- IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) Systems, IEEE Std. 929-2000. 11-M.Hamzeh, Sh.Farhangi, M.Saradarzadeh, “Anti-Islanding Protection of Distribution Networks with Photovoltaic Sources by Active Frequency Drift Method,” PSPC 2008. 12- P.Mahat, Z.Chen, and B.B.Jensen, “Review of Islanding Detection Methods for Distributed Generation,” IEEE DRPT 2008, Nanjing China, vol.4, No.2, Pp.2743-2748, 6-9. April. 2008. 13- A. Pigazo, M .Liserre, “Wavelet-Based

Islanding Detection in Grid-Connected PV Systems,” IEEE Trans. on Industrial Electronics, vol. 56, no. 11, pp. 4445-4454, November 2009. 14- P.Mahat, Z.Chen, and B.B.Jensen, "Review of Islanding Detection Methods for Distributed Generation," IEEE DRPT2008, Nanjing China, vol.4, No.2, Pp.2743-2748, 6-9. April. 2008. 15- M. A. Redfern, O.Usta,” Protection against Loss of Utility Grid Supply or a Dispersed Storage and Generation Unit’, IEEE Trans. Power Delivery, vol.8, no.3, pp. 948–954,1993

load parameters effect on relays performance for islanding detection

Documents