Unplanned islanding detection of renewable energy sources using sequence impedance with zero NDZ

Abstract

Unintentional islanding, defined as the inadvertent separation of distributed generators from the main grid, is a significant issue for wind power and synchronous-based microgrid. Islanding events must be detected to ensure safety and protect utility workers and equipment. The improper detection islanding may lead to power quality degradation, grid protection interference, and out-of-phase reclosing. To overcome these issues, researchers developed the islanding detection technique. This paper proposes a method, for accurately discerning between instances of islanding and those that are not islanding-related events, that employs two new criteria: (1) superimposed positive sequence impedance (SPSI) and (2) superimposed negative sequence impedance (SNSI). The islanding detection indices (IDI₁ and IDI₂) are calculated using the SPSI and SNSI. If the values of IDI₁ and IDI₂ exceed a predefined threshold, islanding is detected. The performance of the proposed approach is tested using the MATLAB/Simulink platform. The simulation results show that the proposed technique detects islanding in 20 ms at nearly zero power mismatch and under balanced conditions. Therefore, it limits unwanted tripping caused by various types of non-islanding events. The proposed technique is compared with existing schemes to demonstrate its superiority.

1 Introduction

1.1 Motivation

In recent years, distributed energy sources like solar, wind, and small hydropower have been heavily integrated into the system for a variety of reasons, including increased power supply reliability and resilience, minimized power outages, avoidance of temporary interruption of critical loads, and reduction of environmental issues. When the penetration of microgrids into the power grid is observed, some technical challenges arise, such as the coordination of multiple generations, the sensitivity of protection schemes, and islanding detection and control. Islanding is one of the most serious risks associated with microgrids. There are two types of islanding: intentional and unintentional islanding. Intentional islanding serves for system maintenance and protection. Unintentional islanding conditions need to be quickly identified to take the necessary steps to maintain the safe operation of the network. The IEEE 1547 standard highlights how important it is to identify an islanding situation and separate the microgrid from the utility grid within 2 s [1].

1.2 Literature review

Several islanding detection algorithms have been presented in the literature over the last two decades by researchers and industry. Select the most appropriate islanding scheme for the converter and synchronous-based microgrid because the dynamic behavior of the synchronous-based microgrid is different from the converter-based microgrid after the islanding occurrence. The islanding techniques are classified as local and remote [2,3,4]. The local technique operated on the microgrid side. There are two types of local methods: active and passive. The passive scheme is affected by changes in parameters such as frequency, voltage, current, and THD. Passive method vulnerability is that choosing a threshold is challenging, and the non-detection zone (NDZ) is large. It includes the ROCONSVAC [5], change in apparent power [6], and the phase angle between negative sequence voltage and current [7]. Furthermore, intelligent-based passive islanding techniques have been proposed, which combine decision trees and data-mining models. In terms of reliability, training time requirements, and handling large data sets, decision trees outperform other data-mining algorithms such as artificial neural networks, support vector machines, and k-nearest neighbors. In a distribution system, decision tree regression is used to locate faults by combining the features of the discrete Fourier transform (DFT) and the discrete wavelet transform. Active islanding algorithms are based on injecting a small perturbance signal into the control network [8]. Numerous active schemes exist, including open-close of the SSSG circuit breaker [9], capacitor switching [10], and injected into the SSSG active/reactive power control loops [11]. It has a smaller NDZ than passive techniques. Active techniques degrade power quality. The drawbacks of passive and active strategies will be overcome by the use of hybrid techniques. Normally, the hybrid approaches are a combination of passive and active [12]. A few examples of hybrid techniques are fuzzy classifier and reactive power [13], rate of change of frequency and reactive power insert, and ROCOV and ROCOF [14, 15]. The remote technique is operated when information sends between the utility grid and distributed generation through the communication channel. The remote schemes operated on the utility grid side [16]. Remote techniques do not have a non-detection zone. These techniques are more accurate and reliable than local islanding methods, However, their utilization is extremely expensive due to the requirement of costly communications signals between the power system network.

1.3 Contributions

The literature survey above indicates that active techniques degrade power quality while remote schemes are more expensive. It is found that passive techniques fail to detect islanding at zero power mismatch. This article proposes a novel mixed islanding detection scheme based on an integrated superimposed sequence impedance. The mixed islanding technique provides a better response in all islanding and non-islanding environments. The proposed technique has been tested on synchronous and wind turbine-based DG microgrids. The following are the benefits of the suggested mixed detection method:

• This method is more effective than the other techniques.

• The suggested scheme quickly and accurately detected islanding at zero power mismatch.

• It performs well in NIEs, such as nonlinear loads and low fault resistance values.

• The scheme detects islanding in 20 ms.

• It is easy to implement at a low cost.

• The proposed method is tested for a converter and synchronous machine-based distributed generators.

• No issue with power quality.

The layout of this article is organized into sections: Sect. 2 describes the proposed scheme. Section 3 studies the test system and threshold setting. Section 4 is the test simulation and outcomes. Section 5 contained a comparison analysis and discussion. The conclusion of this paper is presented in Sect. 6.

2 The proposed methodology based on SPSI and SNSI

2.1 Mathematical calculation of sequence component

Figure 1 demonstrates step by step flow diagram of the proposed methodology. At first, the three-phase voltage and current signal are acquired at the PCC point. The voltage and current samples are collected over a one-cycle period at a sampling frequency of 1.2 kHz, and the system operates at a fundamental frequency of 60 Hz. The samples of voltage and current signals contain large harmonics. It is removed when passing through recursive DFT.

Fig. 1

Flow diagram of the islanding detection method

Let y(t) be the input signal that operates at the system frequency $f_{0} ,$, whereas $Nf_{0}$ is a sampling frequency and $\phi$ be phase angle in radians. To compute the phasor and magnitude of (1) using recursive DFT [17, 18].

$$y\left( t \right) = y_{m} Cos\left( {2\pi f_{0} t + \phi } \right)$$

(1)

where N is the input data sample $y_{n} :\left\{ {n = 0,1,2,....\left. {N - 1} \right\}} \right.$ given in (2). And sampling angle $\theta = 2\pi /N$.

$$y_{n} = Y_{m} Cos\left( {n\theta + \phi } \right)$$

(2)

$$y_{c}^{N - 1} = \frac{\sqrt 2 }{N}\sum\limits_{n = 0}^{N - 1} {y_{n} \cos \left( {n\theta } \right)}$$

(3)

$$y_{c}^{N - 1} = \frac{{X_{m} }}{\sqrt 2 }\cos \phi$$

(4)

Only the cosine $\cos \left( {2n\theta } \right)$ term is present in the above Eq. (4). The sine sum is estimated in the same way:

$$Y_{s}^{N - 1} = \frac{\sqrt 2 }{N}\sum\limits_{n = 0}^{N - 1} {y_{n} \sin \left( {n\theta } \right)}$$

(5)

$$Y_{s}^{N - 1} = - \frac{{y_{m} }}{\sqrt 2 }\sin \phi$$

(6)

The phasor $Y^{N - 1}$ is represented by

$$Y^{N - 1} = Y_{c}^{N - 1} - jY_{s}^{N - 1}$$

where $a = Y_{c}^{N - 1}$ and $b = Y_{s}^{N - 1}$ is the real and imaginary coefficients of the input signal, respectively. The above equation calculates the input signal's magnitude and phase angle using recursive DFT.

$$\left| {Y^{N - 1} } \right| = \sqrt {a^{2} + b^{2} } \,{\text{and}}\,\theta = \tan^{ - 1} \left( \frac{b}{a} \right)$$

The three-phase voltage $\left( {V_{\alpha } ,V_{\beta } \,{\text{and}}\,V_{\gamma } } \right)$ and current $\left( {I_{\alpha } ,I_{\beta } \,{\text{and}}\,I_{\gamma } } \right)$ are estimated from recursive DFT as explained above. The voltage $\left( {V_{0} ,V_{1} \,{\text{and}}\,V_{2} } \right)$ and current $\left( {I_{0} ,I_{1} \,{\text{and}}\,I_{2} } \right)$ sequence components are acquired from the fundamental waveforms of three-phase voltage and current. Equations (7) and (8) express the connection between the phase and sequence of voltage and current.

$$\left[ \begin{gathered} V_{0} \hfill \\ V_{1} \hfill \\ V_{2} \hfill \\ \end{gathered} \right] = \frac{1}{3} \times \left[ {\begin{array}{*{20}c} 1 & 1 & 1 \\ 1 & k & {k^{2} } \\ 1 & {k^{2} } & k \\ \end{array} } \right] \times \left[ \begin{gathered} V_{\alpha } \hfill \\ V_{\beta } \hfill \\ V_{\gamma } \hfill \\ \end{gathered} \right]$$

(7)

$$\left[ \begin{gathered} I_{0} \hfill \\ I_{1} \hfill \\ I_{2} \hfill \\ \end{gathered} \right] = \frac{1}{3} \times \left[ {\begin{array}{*{20}c} 1 & 1 & 1 \\ 1 & k & {k^{2} } \\ 1 & {k^{2} } & k \\ \end{array} } \right] \times \left[ \begin{gathered} I_{\alpha } \hfill \\ I_{\beta } \hfill \\ I_{\gamma } \hfill \\ \end{gathered} \right]$$

(8)

where $k = 1\angle 120^\circ$ is the complex operator.

2.2 Computation of positive and negative sequence impedances

If the microgrid network is operated under abnormal conditions, the voltage and current sequence components will change. Therefore, the impedances of the positive and negative sequences vary. From Eqs. (7) and (8), positive and negative sequences of voltage and current are expressed in Eqs. (9), (10), (11), and (12):

Where $V_{\alpha }$, $V_{\beta }$, and $V_{\gamma }$ are the three-phase voltage signal and can be written as phasor form $V_{\alpha } \angle \theta_{\alpha }$, $V_{\beta } \angle \theta_{\beta }$ and $V_{\gamma } \angle \theta_{\gamma }$, respectively. Similarly $I_{\alpha }$, $I_{\beta }$ and $I_{\gamma }$ are a three-phase current signal that can be written as phasor form $I_{\alpha } \angle \theta_{\alpha }$, $I_{\beta } \angle \theta_{\beta }$ and $I_{\gamma } \angle \theta_{\gamma }$, respectively.

$$V_{1} = \frac{1}{3}\left( {V_{\alpha } \angle \theta_{\alpha } + V_{\beta } \angle \left( {\theta_{\beta } + 120} \right) + V_{\gamma } \angle \left( {\theta_{\gamma } + 240} \right)} \right)$$

(9)

$$I_{1} = \frac{1}{3}\left( {I_{\alpha } \angle \theta_{\alpha } + I_{\beta } \angle \left( {\theta_{\beta } + 120} \right) + I_{\gamma } \angle \left( {\theta_{\gamma } + 240} \right)} \right)$$

(10)

$$V_{2} = \frac{1}{3}\left( {V_{\alpha } \angle \theta_{\alpha } + V_{\beta } \angle \left( {\theta_{\beta } + 240} \right) + V_{\gamma } \angle \left( {\theta_{\gamma } + 120} \right)} \right)$$

(11)

$$I_{2} = \frac{1}{3}\left( {I_{\alpha } \angle \theta_{\alpha } + I_{\beta } \angle \left( {\theta_{\beta } + 240} \right) + I_{\gamma } \angle \left( {\theta_{\gamma } + 120} \right)} \right)$$

(12)

Positive and negative sequence impedance can be calculated using Eqs. (9), (10), (11), and (12).

$$Z_{1} = \frac{{V_{1} }}{{I_{1} }}$$

(13)

$$Z_{2} = \frac{{V_{2} }}{{I_{2} }}$$

(14)

2.3 Estimate of superimposed positive and negative sequence impedances (SPSI and SNSI)

When an islanding or non-islanding event occurs, the superimposed positive and negative impedance is affected. It is referred to as the difference between the post-event and pre-event elements of positive and negative sequence impedance. SPSI and SNSI are computed using Eqs. (15) and (16)

$$\Delta Z_{1} = \left| {Z_{1pos} } \right| - \left| {Z_{1pre} } \right|$$

(15)

$$\Delta Z_{2} = \left| {Z_{2pos} } \right| - \left| {Z_{2pre} } \right|$$

(16)

2.4 Islanding detection index (IDI)

The absolute outputs of the SPSI and SNSI are used to calculate IDI₁ and IDI₂ (absolute means of ∆Z₁ and ∆Z₂) as shown in (17) and (18), respectively.

$$IDI1 = \frac{1}{T}\int_{t - T}^{t} {\left| {\Delta Z_{1} (t)} \right|} dt$$

(17)

$$IDI2 = \frac{1}{T}\int_{t - T}^{t} {\left| {\Delta Z_{2} (t)} \right|} dt$$

(18)

T in (17) and (18) shows the fundamental period of the system; t in the second is the instantaneous time. The IDI₁ and IDI₂ are compared with threshold th1 and th2 using a comparator (COMP) as shown in Fig. 2. When IDI₁ is greater than th1, logic is '1'; otherwise, logic is '0.' Similarly, if IDI₂ is greater than th2, the logic returns '1' otherwise '0.' Whenever at least one of them is ‘1’ otherwise, its value is ‘0’. If logic '1' appears in the trip signal, it indicates islanding events and if logic '0' appears, it indicates non-islanding events.

Fig. 2

Logic diagram of the proposed algorithm

3 Test system

Figure 3 represents the multi-bus validation test network operating at (25 kV, 60 Hz). Table 1 lists all the parameters of the distributed test network. There are two distributed generators in the plant: a 9 MW wind farm and another 3.5 MW synchronous distributed generator integrated with the main grid of 1000MVA, 120 kV, and 60 Hz. The distribution network has evaluated various tests of islanding and non-islanding events with different sizes of DG at multiple buses.

Fig. 3

Distributed test network under study

Table 1 The test Simulink model parameters

3.1 Selection of threshold

Several factors affect threshold setting, including sampling interval, rating of distribution generators, system parameters, and network topology. Therefore, an optimal threshold value has been selected. In both islanding and non-islanding scenarios, the maximum magnitude of the IDI1 and IDI1 is calculated for selecting the threshold value, as illustrated in Table 2. The maximum magnitudes of IDI1 and IDI2 for a non-islanding event are 0.80 and 0.80, respectively. The minimum values of IDI₁ and IDI₂ for an islanding event are 1.3 and 2.8, respectively. The safety margin considered in this case is 0.1. To discriminate between IEs and NIEs perfectly, the value of the threshold (th1 and th2) is set to 0.9.

Table 2 The maximum value of the islanding detection index (IDI) for various IEs and NIEs

Threshold value = (The maximum value of the islanding detection index (IDI) in a non-islanding event + safety margin).

4 Simulation test results

Figure 2 illustrates a single-line diagram of the distributed network study. The various cases tested on the multi-bus network are classified as follows: The IEs at different active and reactive power (high, low, and zero) mismatch. NIEs include load, capacitor, fault, motor switching, voltage sag, and swell.

4.1 Islanding condition

When the microgrid is completely disconnected from the power grid and operates in stand-alone mode, it continuously supplies power to a local load. This event has an impact on the DG's voltage and frequency.

4.1.1 Islanding at different power mismatches

In most cases, hybrid methods are unable to detect islanding at low or zero power mismatches. This section evaluates the performance of the proposed scheme under various active and reactive power mismatches (0%, 25%, 75%) between distributed generation units and load demands. From Fig. 4a–b, the maximum magnitude of the islanding detection index (IDI₁ and IDI₂) at various power mismatches exceeds the predefined threshold (0.9). Islanding began at t = 2 s in this event. However, the proposed algorithm detects islanding at t = 2.02 s, as shown in Fig. 4c. The islanding detection time is 20 ms. According to the above discussion, the proposed mixed technique is more effective at zero power mismatch.

Fig. 4

Power mismatch Islanding detection condition a IDI₁ versus time b IDI₂ versus time c trip signal versus time

4.2 Non-islanding condition

The collaboration of the microgrid grid with the utility grid is known as grid-connected mode. The following are examples of non-islanding events (NIEs): load, capacitor, motor, and nonlinear load switching.

4.2.1 Impact of load switching

In this event, a load with a capacity of 4.5 MW and 1.2 MVAR is switched in the distribution test network at t = 2 s. To evaluate the performance of the mixed detection methodology for load switching. According to Fig. 5a–b, the response of the islanding detection index (IDI₁ and IDI₂) in both cases is less than the threshold value (0.9). Hence, the suggested technique remains stable when encountering load switching, without generating a trip signal, as illustrated in Fig. 5c.

Fig. 5

Load switching response a IDI₁ versus time b IDI₂ c trip signal versus time

4.2.2 Effect of capacitor switching

Parallel-connected capacitors are employed within the power system to enhance power factor and address voltage sags. In order to assess the efficiency of the suggested approach, an abrupt alteration in capacitive load, rated at 4.5 MVAR, is introduced at t = 2 s. From Fig. 6a–b, the magnitude of the islanding detection index (IDI₁ and IDI₂) in both cases is less than the predefined threshold value (0.9). Therefore, it does not initiate the trip signal as depicted in Fig. 6c. Hence, the proposed algorithm remains stable during abrupt changes in capacitive load.

Fig. 6

Capacitor switching response a IDI₁ versus time b IDI₂ versus time c trip signal versus time

4.2.3 Effect of motor switching

The electrical characteristics of the power system undergo modifications while switching the motor, during both connection and disconnection phases. These changes could potentially lead to disruptions in the anti-islanding protection of the DG system. To evaluate the performance of the proposed scheme, a 230HP motor load is switched in the test network at t = 2 s. In both cases, the values of the islanding detection index (IDI₁ and IDI₂) are less than the threshold value (0.9), as shown in Fig. 7a–b. Consequently, the proposed technique fails to detect islanding, and the initiated trip signal is zero, as shown in Fig. 7c. According to the above discussion, the proposed method remains stable during abrupt changes in motor load.

Fig. 7

Motor switching response a IDI₁ versus time b IDI₂ versus time c trip signal versus time

4.2.4 Effect of fault switching

To investigate the effectiveness of the mixed technique. However, various faults occurred in the test network. The different types of faults are given below:

• Symmetrical fault (LLLG, LLL)

• Asymmetrical fault (LG, LL, LLG)

A fault resistance of 0.01Ω is switched in the distribution test network at t = 2 s. From Fig. 8a–b, the magnitude of the islanding detection index (IDI₁ and IDI₂) for symmetrical and asymmetrical fault is less than the threshold value of 0.9. In this case, the proposed approach does not detect islanding. Therefore, the algorithm does not generate the trip signal as shown in Fig. 8c.

Fig. 8

Fault switching response a IDI₁ b IDI₂ c trip signal

4.2.5 Effect of nonlinear switching

In this section, a three-phase nonlinear diode rectifier load with a rating of 0.9 MW is switched in the multi-bus network at t = 2 s. From Fig. 9a–b, the response of the islanding detection index (IDI₁ and IDI₂) values in both cases is below a predefined threshold (0.9). Therefore, the proposed methodology does not detect it as islanding, and the system remains stable. The proposed method does not generate a trip signal as shown in Fig. 9c.

Fig. 9

Nonlinear load switching response a IDI₁ b IDI₂ c trip signal

4.2.6 Effect of power quality disturbance

To evaluate the performance of the proposed scheme for power quality issues. A heavy load with capacity (22.5 MW and 3.4 MVAR) is injected into the test network at t = 2 s. In this case, the voltage drop is less than the normal voltage and is referred to as voltage sag. The response of this case is shown in Fig. 10a. When a heavy load with capacity (22.5 MW and 3.4 MVAR) is ejected from the test system at t = 2 s. In this event, the voltage is higher than the rated voltage and is known as a voltage swell. The response to this event is represented in Fig. 10b. From Fig. 10c–d, the magnitude of IDI₁ and IDI₂ in both cases (injected and ejection) is less than the threshold value (0.9). Therefore, the proposed logic-initiated trip signal is zero as shown in Fig. 10e. Therefore, the power quality (sag and swell) of the distribution network does not affect the proposed mixed scheme.

Fig. 10

a Voltage sag at PCC versus time b voltage swell at PCC versus time c IDI₁ d IDI₂ e trip signal

5 Comparative analysis and discussion

The proposed approach is tested on a power network with synchronous and wind power DG units in various scenarios of islanding and non-islanding cases. According to the above discussion, the proposed methodology easily distinguishes between islanding and non-islanding conditions for every kind of DG. In this section, a comparative analysis of the proposed technique with other existing methods is presented in Table 3 which reflects the betterment of the proposed scheme over the existing schemes—be it NDZ, complexity, threshold setting, power quality issue, etc. A comparison of the proposed scheme with a [22] (PAOPSV) is represented in Fig. 11a–b. When the proposed method operated at zero power mismatch, islanding is detected within 20 ms, which is faster than the existing method [22]. However, the previous method [22] detects islanding in 100 ms. The rate of change of power (ROCOP) based on the terminal voltage (TV) (ROCOP-TV) [28] scheme accurately detects the islanding for the photovoltaic (PV) inverter-based microgrids. The ROCOP-TV method is fast and distinguishes between islanding and regular grid faults. However, nonlinear load switching was not considered during the evaluation and NDZ value is not provided. Further, another rate of change of superimposed impedance (ROCSI)-based scheme has a non-detection zone of ± 0.1 to ± 0.5% [29]. Also, this method finds limitations in distinguishing the islanding event (IE) from the non-islanding event (NIE) for LLLG faults with fault resistance Rf = 0.01Ω created in the test network at t = 2 s, as shown in Fig. 12. Moreover, the technique of [29] was only tested for inverter-based microgrids. However, the proposed scheme successfully distinguishes IEs from NIEs with zero NDZ and it works well for synchronous and wind power-based microgrids.

Table 3 Comparison of several islanding detection schemes

Fig. 11

Zero power mismatch islanding condition a PAOPSV b Trip signal

Fig. 12

The magnitude of ROCSI for the LLLG fault switching with fault resistance R_f = 0.01Ω

6 Conclusion

In this article, a new mixed islanding technique is proposed to overcome the issue of unintentional islanding. This scheme successfully distinguishes between islanding and non-islanding conditions at zero power mismatch and hence avoids unwanted tripping. The operation of the proposed scheme works accurately with various islanding and non-islanding events such as load switching, capacitor switching, nonlinear switching, asynchronous motor switching, and fault switching. According to the simulation results, this technique can detect islanding quickly and correctly in 20 ms. The proposed method does not degrade power quality and is simple to implement with low operating costs. This algorithm detects islanding in less than 2 s and adheres to the IEEE 1547 standard. A comparative analysis is also conducted, and the results are believed to be significantly superior. The advice for future work is observed. In this study, only wind power and synchronous-based DG are considered, so an islanding scheme must be developed to extend the results of this paper and work effectively for solar and fuel cell DG systems.