Electric Power Systems Research 110 (2014) 180–187

Contents lists available at ScienceDirect

Electric Power Systems Research
journal homepage: www.elsevier.com/locate/epsr

A new passive islanding detection method and its performance
evaluation for multi-DG systems
A.H. Mohammadzadeh Niaki ∗ , S. Afsharnia
School of Electrical and Computer Engineering, College of Engineering, University of Tehran, Tehran 14395-515, Iran

a r t i c l e

i n f o

Article history:
Received 12 September 2013
Received in revised form 14 January 2014
Accepted 20 January 2014
Available online 17 February 2014
Distributed generation
Islanding detection
Empirical mode decomposition
Intrinsic mode function
Multi-DG system

a b s t r a c t
This paper presents a passive islanding detection method for inverter-based distributed generation based
on empirical mode decomposition (EMD) technique. The voltage of point of common coupling (PCC) is
measured and its intrinsic mode functions (IMFs) are obtained using EMD. The first IMF component of
PCC per unit voltage is the parameter used for islanding detection. Performance of the proposed method
is evaluated for single-DG and multi-DG cases. Simulation results performed in MATLAB/SIMULINK environment show that the islanding can be detected in less than two cycles, even for zero power mismatch.
Moreover, the proposed method functions properly for various configurations of multi-DG systems, DGs
switching events, various loadings of DGs and different DG interface controls.
© 2014 Elsevier B.V. All rights reserved.

1. Introduction
Environmental pollution of fossil fuels caused an increased
application and penetration of distributed generation (DG) systems
using renewable energy sources. Integration of these DGs to distribution network has remarkable advantages, including increased
reliability and reduced line losses. On the other hand, some problems and concerns may be generated. One of the most critical
concerns is islanding detection. Islanding is a condition in which a
portion of the distribution network comprising local loads and one
or more DGs remains energized while isolated from the rest of the
system. Islanding detection is one of the mandatory requirements
for DGs, specified in the IEEE Std. 929-2000 and IEEE Std. 15472003 [1,2]. Based on these standards, an unintentional island shall
be detected within 2 s and the related DGs shall be isolated from
the distribution system. Therefore, a fast and accurate islanding
detection method is essential.
Islanding detection techniques are classified in two categories:
remote and local techniques. Remote methods are based on the
communication between utilities and DGs. In contrast, local methods use measured data at the DG site. Remote techniques are more
reliable than local ones, but their implementation is more expensive. So, local methods are widely used for islanding detection. They
can be categorized into passive, active and hybrid methods.

∗ Corresponding author. Tel.: +98 9111009150.
E-mail addresses: a.mohammadzadeh@ece.ut.ac.ir
(A.H. Mohammadzadeh Niaki), safshar@ut.ac.ir (S. Afsharnia).
0378-7796/$ – see front matter © 2014 Elsevier B.V. All rights reserved.

In passive techniques, system parameters like voltage, frequency, etc. are continuously monitored and compared with a
predetermined threshold. Active methods intentionally inject disturbances into the system. Hybrid methods are a combination of
passive and active methods. A comprehensive survey on islanding
detection methods is presented in [3–6].
Active methods have relatively smaller non-detection zone
(NDZ) than passive methods. But they degrade the power quality
due to the perturbations introduced to the system. Since the passive methods are usually simple and easy to implement and do not
introduce any disturbance, applying a passive method with small
NDZ is preferred to an active method.
Time-frequency transform-based passive anti-islanding techniques have been recently proposed. Wavelet transform and
S-transform have been presented for islanding detection in [7–12].
These transforms are applied on PCC voltage and current signals to
get useful information and calculate suitable parameters, e.g. high
frequency components and spectral energy of the signal.
Wavelet transform is basically a time-scale analysis, not a real
time-frequency analysis. One of the problems of the wavelet analysis is its non-adaptive nature. Once the mother wavelet is selected,
it cannot be changed during the analysis and have to be used to
analyze all the data. Moreover, spectral wavelet analysis underlies an uncertainty principle, indicating that a time or frequency
dependent information cannot be classified by the same accuracy,
The S-transform is a combination of the short time Fourier transform (STFT) and the wavelet transform by changing the shape of
the S-transform wavelet. Although the S-transform can perform

determined a criterion for the sifting process to stop.21–23]. The system consists of an inverter-based DG. Consequently they have large NDZ and are not suitable for islanding detection in multi-DG systems. the algorithm of EMD can be summarized as follows: 1) Identify local maxima and minima of signal X(t)..22]. one cannot expect the predetermined Gaussian window to fit all signals. 6) Compute the residue. The superiority of this method to wavelet transform. it is applicable to nonlinear and non-stationary data [13]. r1 (t) = X(t) − c1 (t). and (b) at any point. This can be accomplished by limiting the size of the standard deviation. the mean value of the envelope defined by the local maxima and the envelope defined by the local minima is zero.2 and 0. . It functions properly in multi-DG systems. Moreover. some passive methods which perform at small power mismatch have been presented [9–12. c2 (t). Empirical mode decomposition The concept of the empirical mode decomposition method is to identify the intrinsic oscillatory modes by their characteristic time scales in the data and then decompose the data accordingly.cn (t) are the IMF components and rn (t) is the residue extracted by EMD method. straightforward and easy to implement and has a small computation time. . But some of them have not been evaluated in multi-DG systems [9. Otherwise. EMD decomposes the signal with different time scales into its components effectively. the HHT is an adaptive way to produce physically meaningful representation of data. The studied system The sample system shown in Fig.3 [13]. Based on this figure. The paper is organized as follows: Section 2 describes the EMD method and the approach to obtain the IMF components of a signal. non-linear. signal X(t) is considered as follows: X(t) = −20t + 2 sin (40t) + sin (160t) (3) The original signal and its IMFs obtained from EMD method are shown in Fig. 2 has been considered to describe the proposed method and evaluate its performance. the proposed method is very simple. EMD is a key part of the Hilbert–Huang Transform (HHT). An IMF is a function that satisfies two conditions: (a) in the whole data set. 3. respectively. or when the residue r1 (t) becomes a monotonic function from which no more IMF can be extracted. an islanding detection technique based on the empirical mode decomposition (EMD) method is presented. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 multi-resolution analysis and retain the frequency information. Compared with other time-frequency based techniques.A. In this paper. This method is also called the sifting process. accordingly. 2) Perform cubic spline interpolation between the maxima and the minima to obtain the envelopes emax (t) and emin (t). 2. Huang et al. T is the time interval in which EMD method is applied.21. then repeat steps 1–4 on h1 (t) instead of X(t) until the new h1 (t) satisfies the conditions of an IMF. as well. The studied system and its parameters are presented in Section 3.H. the number of extrema and the number of zero crossings must either equal or differ at most by one. The majority of passive islanding detection methods are unable to detect islanding when the power mismatch in islanded system is close to zero [17–20]. Recently. 1. A typical value for SD can be set between 0. The resultant IMF is called c1 (t). Mohammadzadeh Niaki. Since the decomposition is based on the local characteristic time scale of the signal. The signal given in (3) and its IMF components. 7) The sifting process can be stopped either when the component c1 (t) or the residue r1 (t) becomes so small that it is less than the predetermined value of substantial consequence. The details of the proposed method are described in Section 4. The detection time of the method and the required data window is also specified. EMD is used to decompose the signal into a finite and often small number of intrinsic mode functions. 4) Extract h1 (t) = X(t) − m(t). 5) If h1 (t) is not an IMF (based on above definition). it is more time consuming compared with other time-frequency based methods.. Finally. Extensive simulations are performed in single-DG and multi-DG systems with several case studies and the performance of the proposed technique is investigated. If c1 (t). The name “intrinsic mode function” is adopted because it represents the oscillation mode embedded in the data. repeat steps 1–6 on r1 (t) to obtain the next IMF and a new residue. S. SD. 3) Compute mean of the envelopes. which is a powerful tool for analyzing linear. conclusions are given in Section 7. Unlike the above mentioned transforms. Sections 5 and 6 present the simulation results and performance evaluation of the proposed technique in single-DG and multi-DG systems. 181 Fig. computed from the two consecutive sifting results as: SD =    T h1(k−1) (t) − h1k (t)2  t=0 h21(k−1) (t) (1) where. Given a signal X(t). 1. The DG unit comprises a DC . STFT and S-transform has been presented in the literature [13–16]. The proposed passive EMD-based method in this paper can detect islanding even when the generation and load exactly match (zero power mismatch) and thus its NDZ is zero. which all connect to the PCC. a three phase RLC load and the grid. stationary and nonstationary signals. m(t) = (emax (t) + emin (t))/2. respectively. This example simply shows the performance of the EMD method to extract different modes of the signal and the validity of results. the original signal can be reconstructed as: X(t) = n  ci (t) + rn (t) (2) i=1 To see how the EMD works.

has not been shown in these figures. etc.e. respectively. is used in the proposed method to be applicable to all systems with different voltage levels. System under study. the PCC voltage is maintained by the network. The quality factor. The studied system has been modeled in MATLAB/SIMULINK software environment.1. the concept. are presented in this section. including non-linear and non-stationary signals. The islanding is simulated by grid breaker disconnection at t = 0. The islanding is simulated with opening the grid breaker. Figs. 4. Concept of the proposed method When DG operates in grid-connected mode. voltage source (Vdc ). Fig.1 ␮F 2. Any signal.H. based on network rated voltage. i. After the grid disconnects and islanding occurs. the PCC voltage will be affected and its components will be changed considerably. On the other hand. the value of first IMF. 3 and 4 show the PCC per unit voltage and its IMFs in non-islanding and islanding conditions for system described in Section 3. 4. Proposed islanding detection method The details of the proposed technique. the high frequency components of voltage and consequently. 2. It can be seen from Figs. can be decomposed into its components by EMD technique. Therefore. the support of the network will be lost and thus. It should be noted that only the first IMF is required for islanding Table 1 Parameters of the studied system. IMF components of PCC per unit voltage for islanding condition.5 mH . 4. The system parameters and their values are given in Table 1. of the RLC load are equal to 2. c1 .89 mH 1439. detection time. f0 . PCC per unit voltage. the first IMF of PCC per unit voltage is called IMF1 hereinafter. For this purpose. Qf .07 s.015  0. 3 and 4 that once the islanding occurs. first IMF component of PCC per unit voltage is the parameter used for islanding detection in the proposed method.5 and 60 Hz. it was shown in Section 2 that EMD can extract different modes of the signal with different time scales. The residue. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 Fig. Mohammadzadeh Niaki. respectively. time domain voltage is measured at the PCC and its IMF components are obtained using EMD method. IMF components of PCC per unit voltage for non-islanding condition. because it represents the trend of the signal and does not contain any useful information.182 A. for brevity. a current controlled voltage source inverter (VSI) with Hysteresis control and a series filter (Lf ). increases significantly. the proposed islanding detection method is based on extracting the components of the PCC voltage and detection of changes in these components. Grid parameters Value Load parameters Value DG parameters Value Line–line voltage Nominal frequency Rg Lg 480 V 60 Hz 0. S.2 mH R L C Qf f0 4. So the value of the Fig. selecting appropriate window.5 60 Hz Rated power Terminal voltage Vdc Lf 50 kW 480 V 900 V 2. rn . 3.608  4. and resonant frequency. For simplicity.

First IMF component of the signal is extracted using EMD technique. In summary. i. Consequently. Determining the suitable window EMD needs a set of data to obtain the IMF1.2.A. There is no complexity in the algorithm and consequently. it will be detected at the end of that cycle. Variations of IMF1 after islanding occurrence: (a) one-cycle window and (b) two-cycle window. for 20 dB signal-to-noise ratio). when the islanding occurs near the end of a cycle. UL1741 Std. IMF1 for different active power mismatches (Q = 0). Islanding takes place at t = 0. load imbalance. detection. straightforward and easy to implement. An islanding will be detected. islanding will be detected in one cycle. the islanding condition is simulated by opening the grid breaker at t = 0. 5. one-cycle data window is sufficient for the proposed method to detect islanding. the appropriate window for EMD technique shall be determined. 4. the last N cycle data is utilized for EMD. Simulation result for this case was previously shown in Fig. Power mismatch is defined as the difference between the power supplied by DG unit and the power consumed by load. Fig. But in some cases. Similarly. The results for all windows are similar. Active and reactive power mismatch The sensitivity of the proposed method to the values of active and reactive power mismatch has been investigated. Since the proposed method shall be suitable for on-line islanding detection. only phase “A” of PCC per unit voltage is used for islanding detection. the data used to find the IMF1 is updated each cycle. Simulations have been carried out for a wide range of active and reactive power mismatches.1.g. Islanding As stated in Section 4. Moreover.07 s.2. After finding the IMF1. if the DG is islanded at any instant of a cycle.3. 5. considering the instance of the islanding occurrence in a cycle and the required computation time. 6 shows the effect of active power mismatches on the proposed method.07 s. This leads to increasing the fastness of the method even more. 2. Fig. 5. 6. different active and reactive power mismatches.1. Therefore. 4. The proposed method functions properly in all mentioned cases and is completely robust against noise. switching of non-linear load. induction motor starting. the islanding may be detected in the next cycle. power mismatch is the power supplied/consumed by the grid (positive/negative power mismatch).H. 5(a) and (b) for brevity. The proposed method is very simple.07 s). the detection time of the proposed method is less than two cycles. 5.g. it is not required to continue the sifting process and calculate the higher order IMF components. Only the results for one-cycle and two-cycle window cases are presented in Fig. For most cases. The Simulation results presented in the paper confirm this claim. For this purpose. with the active power mismatch fixed at 0%. no problem in its implementation on an experimental set-up. It is worth noting that due to the advent of powerful and high speed digital signal processors (DSPs). Consequently. the sensitivity of the proposed method to noise has been investigated (e. transient voltage dip. 3–5 are related to phase “A” of PCC voltage. test conditions [24]. Detection time The proposed method works with one-cycle data. the proposed islanding detection method works as follows: One cycle data of phase “A” of PCC per unit voltage is collected at the end of each cycle. only the results for islanding condition and the effect of active and reactive power mismatches on the proposed method in case of single-DG system will be presented. S. Fig. The DG output power can be set by adjusting the magnitude and phase of reference currents in hysteresis control to obtain the required power mismatch. It should be mentioned that the IMFs shown in Figs. Extensive simulations have been done and various cases have been considered: islanding condition. numerous simulations have been performed for a variety of windows. Performance of the proposed method for single-DG system Single-DG system shown in Fig. The threshold value is set to 0. Since the value of IMF1 (c1 ) is greater than the threshold. if the absolute value of this IMF is greater than a predetermined threshold. Mohammadzadeh Niaki. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 183 Fig. These figures show the variations of IMF1 after islanding occurrence (t = 0. For system shown in Fig. 7 represents the effect of reactive power mismatches. The results for phases “B” and “C” are similar to phase “A” and are not shown for brevity. For brevity. the computation time required to obtain the IMF1 is very small. e. The values of active . 2 is tested for a variety of conditions to prove the effectiveness of the proposed method. while the reactive power mismatch is set to 0%.e. load switching event and the effect of load quality factor and load resonant frequency. 4. It is worth noting that the meaning of an “N-cycle window” is that at the end of each cycle.01 based on simulation results.

6. 5.2). RLC load and grid are given in Table 1. Detection time of proposed method for various power mismatches. i. and reactive power mismatches are specified in percent with respect to the load rated active power and the load rated inductive reactive power (balanced condition). It should be mentioned that for all following scenarios. Both DG units are equipped with the proposed islanding detection method. It is assumed that the DG maximum output power is equal to the load rated active power and consequently. respectively. the effect of DG interface control is investigated. For this purpose. the sum of power Fig. S. the proposed method has zero NDZ. even for zero active and reactive power mismatches. Then. The line resistance and inductance (Rline and Lline ) is 0. The parameters of DG unit I. Mohammadzadeh Niaki.02  and 0. Based on this figure. performance of the proposed method is evaluated for multi-DG systems.H. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 Fig. which both are considerably below the required detection time specified in standards [1. 7.3 mH. which can be generalized to multi-DG systems. 9.e. This is the main advantage of the proposed passive islanding detection method compared to other passive methods. two different configurations of DGs are considered: connecting to the same PCC (two parallel-DG system) or connecting to separated PCCs (double-DG system) [25]. The result is shown in Fig. The proposed method is then tested under different loadings of DGs. the output power of DG units are controlled in such a way that zero active and reactive power mismatch is obtained in the islanded system. Performance of the proposed method for multi-DG system Fig. end of cycle and different instances within a cycle.3. In this section. IMF1 for different reactive power mismatches (P = 0). However. both positive and negative mismatches are considered for reactive power. Without loss of generality. sample system shown in Fig. the islanding is detected for most cases in one cycle (16. 8. islanding is applied at different instances. The threshold value is set to 0.67 ms for 60 Hz system) and for other cases in two cycles (33. . NDZ of the proposed method Detection time of the proposed method is investigated for various active and reactive power mismatches (refer to Section 5.33 ms for 60 Hz system).01 for each DG unit. Rating of DG unit II is different from that of DG unit I and is considered as 100 kW.2]. respectively [24]. start of cycle. a VSI unit and a filter. The simulation results show that the proposed method can effectively detect islanding in all cases. Finally. Accordingly. In other words. Maximum detection time is obtained for each active and reactive power mismatch case. in which each DG consists of a DC voltage source. 6. 8. Two-DG sample system. the results will be presented in case of two-DG system. First.184 A. islanding detection performance of one DG is verified in case of switching the other DG. 9 is considered. negative values of active power mismatches are not shown in Fig. Various scenarios are implemented. For this purpose.

DG active power loading is defined as the ratio of active power supplied/absorbed (positive/negative loading) by DG to the load rated active power. 6. 11(a) and (b). Figs. Similarly.07 s and is switched on at t = 0. Double-DG system The DGs are connected to separated PCCs in this case. IMF1 in DG switching event for two parallel-DG configuration. Effect of DG switching events Islanding detection methods may malfunction in multi-DG systems. supplied by two DGs is equal to the power consumed by local load and there is no power exchange with the grid. The results are represented in Fig. IMF1 for positive active power loadings with zero reactive power loadings.3. Islanding occurs at t = 0.1–6. i. 12. 6. 11(b) shows that for DG II. In Sections 6. Positive and negative reactive power loadings are considered too. 11.07 s. The effect of one DG switching on islanding detection of the other DG unit is evaluated in this section. respectively. So. negative values of active power loading are applicable. the islanding detection method of each DG uses its related PCC voltage to obtain IMF1. 13 and 14 show the results for positive and negative active power loadings. in case of one or more DGs switching in the system. The IMF1 variation of PCC per unit voltage is shown in Fig.3.e. while zero reactive power loading is considered. These figures indicate that the value of IMF1 calculated for both DG units is increased above the threshold level after islanding. 10 clearly shows that once the islanding happens. Fig. In this regard. Fig.4. it is supposed that the DC link of DG unit I can store energy and thus. Fig. 185 Fig. 15 .e. 6. Since the rating of DG unit II is greater than load rated active power. Therefore loading of each DG should be considered for two cases: active power loading and reactive power loading. the proposed technique does not detect islanding during DG switching events and is completely secure. Therefore the proposed method successfully detects islanding in double-DG system. The values of Rline and Lline in Fig.A.2. the effectiveness of the proposed method is tested under various loading conditions of two DGs. Effect of DG loading DGs are usually designed to operate at unity power factor [26]. Loading of both DGs are equal. 11(a) shows the variation of IMF1 for DG I and Fig.07 s. S. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 Fig. 6. Only two parallel-DG configuration is considered. the line resistance and inductance in Fig. Islanding occurs at t = 0. Fig. 9 is not zero. IMF1 for DG unit I is shown in Fig. the value of IMF1 becomes greater than the threshold and the proposed method effectively detects islanding in one cycle. In this section. But in some cases. the DG units are connected to one PCC. Mohammadzadeh Niaki. Only the results for two parallel-DG configuration are presented. DG unit II is switched off at t = 0. equal loadings are considered for both DG units. 10. the value of IMF1 is below the selected threshold and consequently. IMF1 in double-DG system for: (a) DG unit I and (b) DG unit II. both islanding detection algorithms have the same input data and thus. reactive power delivery/consumption by DG may be required. Fig. the DG unit I can absorb power. 10. the active power loading of each DG is 50% with zero reactive power loading. DG reactive power loading is defined as the ratio of reactive power supplied/absorbed (positive/negative loading) by DG to the load rated inductive reactive power. Since both DG units connect to the same PCC. DG II switching events do not affect the islanding detection method of DG I. islanding detection method of both DG units operate similarly.07 s. Based on this figure.1 s. IMF1 for two parallel-DG system. Therefore. Two parallel-DG system In this scenario. The system is islanded at t = 0.H. for brevity. Results for double-DG case are similar to two parallel-DG case and are not presented for brevity.1. i. 12. 9 are set to zero to obtain a two parallel-DG configuration. 13.

International Energy Agency Implementing Agreement Photovoltaic Power Systems. The results show that the proposed method functions properly for all active and reactive power loading conditions. which is used for islanding detection. 2000. Moreover. 2003. Effect of DG interface control Fig. Garcera. In this section. IMF1 for various reactive power loadings with equal active power loadings (50%).H. represents the effect of reactive power loadings on the proposed method. IMF1 for negative active power loadings with zero reactive power loadings. Finney. Variation of IMF1 is shown in Fig.186 A. respectively. IEEE Standard 1547TM . The instance of islanding occurrence is t = 0. E. straightforward. [7] C. A review of the islanding detection methods in grid-connected PV inverters.L.N. active and reactive power loadings are specified as “LP” and “LQ”. Lin. Renewable and Sustainable Energy Reviews 21 (2013) 756–766. Simulation results show that the method can effectively detect islanding in less than two cycles. even for zero active and reactive power mismatch conditions. Bower. Conclusion In this paper. Huang. [4] A. 7. DGs switching events. Simulation results show that various DG interface controls do not affect the proposed technique and the method effectively detects islanding. constant current controller with Hysteresis control is considered for DG unit I.H. [2] IEEE Standard for Interconnecting Distributed Resources with Electric Power Systems. Figueres. . Paris. Williams.T. with the active power loading set to 50% for each DG. S. the effectiveness and fastness of the proposed method is proved in multi-DG systems for various configurations (two parallel-DG and double-DG). S. a new passive islanding detection method with zero NDZ is presented for inverter-based DGs. Ahmad. K.4.M.J. N. S. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 Fig. [6] K. B. Masoud. The EMD process is applied to find the first IMF component of PCC per unit voltage. In these figures. J. J. [3] W. The interface control of DG unit II is implemented in d–q synchronous reference frame with two sets of PI controllers and the active and reactive power references are set to fixed values [27]. For this purpose.5.1–6. IMF1 for the case of different DG interface controls. Ropp. different loadings of DGs and various DG control techniques.A.M. G. Current controlled voltage source inverters with Hysteresis control have been considered in Sections 6. Trujillo. Ahmed. the effect of DG interface control on the proposed method is investigated. 2002.J.K. while the DG unit II is equipped with a constant power controller. Technical Report IEA-PVPS T5-09: 2002. Evaluation of Islanding Detection Methods for Photovoltaic Utility-interactive Power Systems. Rahim. Harmonic distortionbased island detection technique for inverter-based distributed generation. 6. M. C. IEEE Standard 929. Mohammadzadeh Niaki. respectively. 16. Renewable and Sustainable Energy Reviews 14 (6) (2010) 1608–1614.E. Selvaraj. robust against noise and therefore suitable for on-line implementation in real networks. Velasco.07 s. France. The active and reactive power loading of each DG is set to 50% and 0%. 14. [5] D. One-cycle data window is used for this purpose. [1] IEEE Recommended Practice for Utility Interface of Photovoltaic (PV) Systems. Review of anti-islanding techniques in distributed generators. Measurement of one phase of PCC voltage is sufficient for the proposed technique. Enhancement of islanding-detection of distributed generation systems via wavelet transform-based approaches. 16. IET Renewable Power Generation 3 (4) (2009) 493–507.W. Hsieh. Two parallel-DG configuration is considered. References Fig. 15. The proposed method is very simple.

M. A new approach to islanding detection of dispersed generators with self-commutated static power converters. F. Iravani. P. [24] Underwriters Laboratories Inc. Alaboudy. H. 903–995. A hybrid time-frequency approach based fuzzy logic system for power island detection in grid connected distributed generation. Bright. IEEE Transactions on Power Delivery 15 (2) (2000) 500–507. M. K. P. Converters. The empirical mode decomposition and the Hilbert spectrum for nonlinear and non-stationary time series analysis. A. S. IEEE Transactions on Smart Grid 3 (3) (2012) 1082–1094. A. Ray. Samui. Dinavahi. Nondetection zone assessment of an active islanding detection method and its experimental evaluation. M. S. International Journal of Electrical Power and Energy Systems 42 (1) (2012) 453–464. B. IEEE Transactions on Energy Conversion 22 (3) (2007) 792–794. Samantaray. IEEE Transactions on Power Delivery 26 (4) (2011) 2687–2696.H. International Journal of Computers. Kishor. Padhee.W.H. A comparison study of improved Hilbert–Huang transform and wavelet transform: Application to fault diagnosis for rolling bearing. Vehicle System Dynamics: International Journal of Vehicle Mechanics and Mobility 47 (4) (2009) 437–456. Wavelet singular entropy-based islanding detection in distributed generation. Dell’Aquila. [18] S. H. IEEE Transactions on Smart Grid 2 (2) (2011) 391–398. and Controllers for Use in Independent Power Systems.C. Comparative study of Hilbert-Huang transform.O. N. S. M. Islanding and power quality disturbance detection in grid-connected hybrid power system using wavelet and Stransform. IET Renewable Power Generation 6 (4) (2012) 289–301. Xu. Donnelly. O.R. Fielding. Time-frequency transform-based islanding detection in distributed generation. Panigrahi. Afsharnia / Electric Power Systems Research 110 (2014) 180–187 [8] [9] [10] [11] [12] [13] [14] [15] [16] International Journal of Electrical Power and Energy Systems 30 (10) (2008) 575–580. Freitas. IET Renewable Power Generation 5 (6) (2011) 431–438. Tse. Protection against loss of utility grid supply for a dispersed storage and generation unit. Freitas.K. V. 454. Z. Mechanical Systems and Signal Processing 19 (5) (2005) 974–988. pp. Basu. IEEE Transactions on Industrial Electronics 56 (11) (2009) 4445–4455. R. . A Bayesian passive islanding detection method for inverter-based distributed generation using ESPRIT. Liu. Hanif.C. Development of EN50438 compliant wavelet-based islanding detection technique for three-phase static distributed generation systems. Huang.H. Pai.A. Huang.R. G. V. 70–73.A.A. W. F. W. N. 2001. M. A. Bahrani. W. Yen. Assessment of ROCPAD relay for islanding detection in distributed generation. [27] X. [25] B. S. [21] A.G.M. in: Proceedings of the Royal Society of London. IEEE Transactions on Power Delivery 26 (2) (2011) 517–525. Najy. 1998.Y. 2001.-S.-J. Waveletbased islanding detection in grid-connected PV systems. Mohammadzadeh Niaki. Wu.K. Affonso. A. IEEE Transactions on Power Delivery 8 (3) (1993) 948–954.K. 187 [17] M. Wang. in: Proceedings of the 7th IEE International Conference on Developments in Power System Protection. Zheng. IEEE Transactions on Power Electronics 24 (3) (2009) 665–673.L. Karimi. Q. H. S. [19] C. UL Standard 1741. Xu. Gaughan. Mohanty.H. Attoh-Okine. [26] H. Moreno. H. Tung. Shen. Samui. Refern.K. Woon.K. Z. Dash. Comparative analysis between ROCOF and vector surge relays for distributed generation applications. P. Shih.R. Inverters. D. Usta. Chu. [20] W. Samantaray. IEEE Transactions on Power Delivery 20 (2) (2005) 1315-1324. T. COROCOF: comparison of rate of change of frequency protection: a solution to the detection of loss of mains. [23] A. C.K. N. Z. A Q–f droop curve for facilitating islanding detection of inverterbased distributed generation. C.H. The fast Fourier and Hilbert-Huang transforms: a comparison.R. Communications & Control (4) (2006) 45–52. N. R. Samui.R.E. Pigazo. Liserre. Fourier transform and wavelet transform in pavement profile analysis.H.L. Peng. Impact of DG interface controls on the Sandia frequency shift antiislanding method. Zeineldin. W. Samantaray. Huang.M. pp.A.C. Chitti Babu. IEEE Transactions on Power Delivery 28 (1) (2013) 411–418. S. Long. Ayenu-Prah. A. Mastromauro. [22] W. Zeineldin.