ISLANDING DETECTION FOR UTILITY INTERCONNECTION OF MULTIPLE DISTRIBUTED GENERATORS

Dung Tran Khanh Viet(1), Kodjo Agbossou(2) and Mamadou Lamine Doumbia(3)

Institut de Recherche sur l’Hydrogène

Department of Electrical and Computer Engineering Université du Québec à Trois-Rivières, C.P. 500

3351 boul. des Forges Trois-Rivières (Québec) Canada G9A 5H7 (1)Dung.Trankhanhviet@uqtr.ca, (2)Kodjo.Agbossou@uqtr.ca and (3)Mamadou.Doumbia@uqtr.ca

ABSTRACT

At present, islanding detection is one of the important research orientations in the area of distributed generation (DG). There exist currently in the literature several methods for islanding detection in the case of single distributed generator connected to a utility grid. These methods are classified as active, passive and methods installed on the utility. However, these methods have limitations or become ineffective in the case of multiple distributed generators connected to the same point of the utility grid because of the interferences between sources. In this paper, we present a robust islanding detection method based on the correlation technique. Analytical and simulation studies were performed in order to validate the accuracy of the method. The results showed that this method behaves well in the case of multiple distributed generators connected to the utility grid even in the critical situations i.e. when the power of the load is provided solely by the distributed generator.

Index Terms— Islanding detection, multiple distributed generators, correlation technique, anti-islanding protection.

1. INTRODUCTION

Islanding is a condition when a portion of the utility network, which contains both load and generation, is isolated from the remainder of the utility system and continued to operate [1], [2]. Islanding can be intentional or non-intentional. During maintenance service on the utility grid, the shut down of the utility may cause islanding of generators. As the loss of the grid is voluntary, the islanding is known. Non-intentional islanding, caused by accidental shut down of the grid, is of more interest. Various risks resulting from this include the degradation of the electric components as a consequence of voltage and frequency drifts, problems with phase shift between the utility grid and the distributed generator when the grid resumes after shut down, safety problems for the public and maintenance personnel. It is a primary

importance to detect any situation of islanding and to reduce the operating time of the islanding system.

2. ISLANDING DETECTION For islanding detection of single distributed generator connected to the utility, there exist in the literature more than fifteen methods which can be classified into three main categories [3]: • Passive methods: over/under voltage, over/under

frequency, voltage phase jump detection, detection of voltage harmonics, etc. [4], [5].

• Active methods: impedance measurement, detection of impedance at specific frequency, slip mode frequency shift, frequency bias, Sandia frequency shift, Sandia voltage shift, frequency jump, etc. [6], [7].

• Methods installed on the utility: impedance insertion, Power Line Carrier Communications (PLCC), SCADA, etc.

In the case of multiple distributed generators connected to the same point in the utility grid, several methods of detection of islanding are ineffective or create other problems on the electrical supply network [8], [9]. In this paper, we present a robust method to effectively detect the islanding in the case of several distributed generators connected to the utility grid. The method is based on the correlation technique.

3. CORRELATION TECHNIQUE 3.1. Principle of the technique Basically, correlation is a measure we use to tell how much two waveforms or two images are like each other. We also use autocorrelation to tell how much a waveform at time t is like itself at some other time, t +τ , or how much an image at location (x,y) is like itself at some other location,

),( τσ ++ yx [10]. Suppose we have two waveforms, x(t)

000557

CCECE/CCGEI May 5-7 2008 Niagara Falls. Canada978-1-4244-1643-1/08/$25.00 � 2008 IEEE

and y(t), defined in the range t = [0, ∞ ). Then, the correlation function of x and y is:

� +=+=∞→

L

Lxy dttytx

LtytxE

0

)()(1

lim)]()([)( τττϕ (1)

Where E[*] stand for the expected value. The independent variable τ has the effect of shifting y forward in time relative to x, thus making the correlation function a continuous function. For sampled waveforms, the correlation function is the discrete form of (1), that is:

�−

=

+∞→

+ ==1L

0nmnn

Lmnnxy yx

L

1lim]yx[E)m(ϕ (2)

We can observe in either (1) or (2) that the correlation function, as defined here, is the average product of two signals. The product is averaged without an upper bound so that, as long as the time shift is positive, we do not need to be concerned with end effects. If the waveforms are defined only in the finite interval [0, NT), then we usually make two assumptions:

- The waveform values are zero outside the interval - The waveform is periodic with period NT.

3.2. Correlation technique for single DG system An equivalent circuit of a single distributed generator connected to the utility is presented in Figure 1. The switch is closed in normal operation mode and opened in the islanding mode.

)(tVd

dZ

rZ

0Z

)(0 tV)(tVr

Fig. 1. Equivalent circuit of single distributed generator system

From the equivalent circuit, the voltage equations can be deduced for the distributed generator and the Point of Common Coupling (PCC):

)2cos()()( ddd fttetV θπ += (3)

)2cos()()( rrr fttetV θπ += (4)

)()( teete ddod Δ+= (5)

)()( teete rror Δ+= (6)

Where:

( )tVd : Output voltage of the DG

( )tVr : Voltage at the PCC

( )ted : Amplitude of the DG output voltage

( )ter : Amplitude of the PCC voltage

The steady state terms are denoted by ‘o’ and the disturbance terms are represented by Δ . In this method, the frequency f is fixed by the grid and

( )tedΔ is randomly disturbed. The correlation function

between ( )tedΔ and ( )terΔ is written:

� +ΔΔ=

T

dr dtteteT 0

)().(1

)( ττϕ (7)

Where T is the period of calculation of the correlation. In

the normal mode, ( )ter is dominated by ( )teo , so ( )tedΔ

and ( )terΔ have a small correlation, )(τϕ is not

significant. However, when islanding occurs, ( )tedΔ and

( )terΔ have a great correlation, )(τϕ is significant. Thus,

the islanding can be detected by the calculation of the value amplitude of )(τϕ . To obtain the random disturbance of the

voltage ( )tedΔ , a random M sequence signal is added to

the amplitude of the DG output voltage. The waveforms of the M sequence and the DG output voltage are shown in Figure 2 and Figure 3 respectively.

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8-5

-4

-3

-2

-1

0

1

2

3

4

5

Time (s)

Vol

tage

(V

)

Fig. 2. Waveform of the M sequence

0 0.05 0.1 0.15 0.2 0.25-200

-150

-100

-50

0

50

100

150

200

Time (s)

Vol

tage

(V

)

Fig. 3. Distributed generator output voltage

4. COMPUTER SIMULATION

4.1. Model of simulation

To simulate the correlation technique for the normal operation and islanding cases, we developed a simulation model using Matlab/Simulink. This model has two blocks: the electric block and the control block. The electric parameters corresponding to the critical case are given in

000558

Table I. For critical case, the nominal load is powered by the DG.

TABLE I. ELECTRIC PARAMETERS OF CRITICAL CASE DG impedance ][ΩdZ 3

0 10*6.0j4.0 −+ ω Nominal load

impedance ][ΩrZ 3

0 10*36.14j58.9 −+ ω

Grid impedance ][0 ΩZ 30 10*049.0j044.0 −+ ω

Grid voltage ])[(0 VtV )t(cos2120 0ω DG voltage ])[( VtVd

)tcos(2125 0 θω + Frequency sec]/rad[0ω π120

4.2. Single distributed generation system

The technique of correlation is initially applied to the single distributed generator connected to the utility. Figure 4 and Figure 5 show the correlation function plots for normal and islanding operations of a single DG connected with utility grid. The system initial parameters are set such as 100% of the load’s powers are supplied by the DG. This is the most difficult case for islanding methods. Figure 4 shows the simulation results which indicate that for normal operation, there is no remarkable peak value on the correlation function

( maxϕ <0.5). When the utility disconnect switch is opened

at time t=0.5 s, i.e the DG is operated in islanding mode, the correlation function presents remarkable peaks. Its

maximum value maxϕ is higher as 2.5 (Figure 5).

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.5

1

1.5

2

2.5

3

Time (s)

Coe

ffic

ient

of

corr

elat

ion

Fig. 4. Correlation coefficient during normal operation of a single connected DG

The other cases of the loads (higher and smaller than nominal value) are also simulated. Correlation coefficient as a function of the load in the case of normal operation and the case of islanding is presented in Figure 6. The curve in bottom corresponds to the case of normal operation and the curve in top corresponds to the islanding case. The most significant correlation coefficient (8.38) is obtained in the case when the load connected at the PCC is 4 times greater than the DG nominal power. The correlation coefficient is least significant (2.5) in the critical case when the load is almost powered only by the DG. However, even

in this case, the correlation coefficient is still higher than that of the normal case of operation (0.8), so that the islanding can be detected by setting a threshold in the protection system. From this study, we can conclude that the correlation technique performs well in the single distributed generation system.

0.1 0.2 0.3 0.4 0.5 0.60

0.5

1

1.5

2

2.5

3

Time (s)

Coe

ffic

ient

of

corr

elat

ion

Fig. 5. Correlation coefficient for single connected DG during islanding operation

0123456789

1/4*

Nomina

l

1/3*

Nomina

l

1/2*

Nomina

l

Nominal

2*Nom

inal

3*Nom

inal

4*Nom

inal

Loads

Co

effi

cien

t o

f co

rrel

atio

n

Normal

Islanding

Fig. 6. Correlation coefficient as a function of the load 4.3. Theoretical validation

To validate the simulation results provided by Simulink software, an analytical study was carried out by programming equation 2. The correlation coefficients obtained from both approaches are identical (Figure 7). Therefore, the correlation technique can be used to investigate islanding detection for the multiple distributed generation system.

0

0,5

1

1,5

2

2,5

3

Normal Islanding

Mode of operation

Co

effi

cien

t o

f co

rrel

atio

n

Simulation

Analytical study

Fig. 7. Comparison of simulation and analytical studies 4.4. Multiple distributed generations system

Based on the model of single generator, we developed a model of multiple distributed generators (DGs) (2 and 4

Disconnection time

000559

generators) connected with the utility. The equivalent circuit of the model is presented in Figure 8. The DGs connected in parallel are identical and the electrical parameters are provided in Table I. The two DGs system with the variation of the load, was first simulated. The normal operation mode and the islanding mode were simulated for various cases: synchronized M sequences, unsynchronized M sequences, and M sequences with different amplitudes. The simulation results are presented in Figure 9.

)(1 tVd

1dZ

1rZ

0Z

)(0 tV)(1 tVr

)(2 tVd

2dZ

2rZ

)(tVdn

dnZ

rnZ

Fig. 8. Equivalent circuit of multiple generators system

0

1

2

3

4

5

6

7

8

9

1/4*

Nomina

l

1/3*

Nomina

l

1/2*

Nomina

l

Nomina

l

2*Nom

inal

3*Nom

inal

4*Nom

inal

Loads

Co

effi

cien

t o

f co

rrel

atio

n

Normal

Synchronized Msequences

Coefficient 1, Msequence 1=2V, Msequence 2=4V

Coefficient 2, Msequence 1=2V, Msequence 2=4V

Coefficient 1, Msequence 2 out ofphase 90 degree

Coefficient 2, Msequence 2 out ofphase 90 degree

Fig. 9. Correlation coefficient as a function of the load for 2 DG sources

The four DGs interconnected system were investigated for different load conditions. The normal operation mode and the islanding mode with synchronized M sequences and unsynchronized M sequences (90 degrees out of phase) were also simulated. The results of simulation in these cases are presented in Figure 10. It can be observed that the correlation technique still works effectively with the multiple distributed generators system (the curve of the normal mode is the lowest) even the correlation coefficient is decreased slightly in the case of 4 sources as compared to the other cases.

This coefficient is influenced tremendously by the change of the amplitude of the M sequence. On the other hand, it is less dependent on the variation of the phase of M sequence. The correlation coefficient is always the most remarkable in the case of maximum loading and the least significant in the critical case.

00,5

11,5

2

2,53

3,54

1/4*

Nomina

l

1/3*

Nomina

l

1/2*

Nomina

l

Nominal

2*Nom

inal

3*Nom

inal

4*Nom

inal

Loads

Co

effi

cien

t o

f co

rrel

atio

n

Normal,synchronized Msequences

Islanding,synchronized Msequences

Normal, Msequences out ofphase 90 degree

Islanding, Msequences out ofphase 90 degree

Fig. 10. Correlation coefficient as a function of the load for 4 DG sources

5. CONCLUSION

In this paper, a robust method based on the correlation technique for islanding detected is proposed. The method was validated by the analytical study followed by simulation. The method effectively detects the islanding in the case of single and multiple distributed generators connected with the utility and even in the critical situation.

ACKNOWLEDGMENT

This work has been supported by the LTE Hydro-Québec and the Natural Sciences and Engineering Research Council of Canada.

REFERENCES

[1] F.S. Pai, S.J. Huang, “A Detection Algorithm for Islanding-Prevention of Dispersed Consumer-Owned Storage and Generating Units”, IEEE Transactions on Energy Conversion, Vol. 16, No. 4, pp. 346-351, December 2001.

[2] W. Bower and M. Ropp, Evaluation of Islanding Detection Methods for Utility-Interactive Inverters in Photovoltaic Systems, Sandia Report SAND2002-2591, 2002.

[3] M. Robitaille, K. Agbossou, M. Doumbia, “Modeling of an islanding protection method for a hybrid renewable distributed generator”, CCECE/CCGEI, Saskatoon, Canada, pages 1477-1481, May 2005.

[4] H. Kobayashi, K. Takigawa, E. Hashimoto, “Method for preventing phenomenon of utility grid with a number of small scale PV systems”, 22nd IEEE Photovoltaic Specialists Conf, pp 695-700, 1991.

[5] A. Kitamura, M. Okamoto, F. Yamamoto, K. Nakaji, H. Matsuda, K. Hotta, “Islanding phenomenon elimination study at Rokko test center”, 1st IEEE World Conf Photovoltaic Energy Conversion, pp 759-762, 1994.

[6] G. A. Kern, “Sunsine 300 utility interactive AC module anti-islanding test results” 26th IEEE Photovoltaic System Conf, pp 1265-1268, 1997.

[7] M. E. Ropp, M. Begovic, A. Rohatgi, “Analysis and performance assessment of the active frequency drift method of islanding prevention”, IEEE Transaction of Energy conversion, vol 14, pp 810-816, Sept 1999.

[8] O. Tsukamoto et al, “Study on islanding of dispersed photovoltaic power systems connected to a utility power grid”, Solar Energy, Vol 70, No 6, pages 505-511, 2001.

[9] C. Jeraputra, P.N. Enjeti, “An improved anti-islanding algorithm for utility interconnection of multiple distributed fuel cell powered generations”, APEC 2005, Vol 1, pages 103-108, 6-10 March 2005.

[10] S. D. Stearns, Digital Signal Processing, CRC Press, 2003.

000560