A Novel Island Detection Threshold Setting Using Phasor Measurement Unit Voltage Angle in a Distribution Network

Abstract

Islanding detection needs are becoming a pivotal constituent of the power system, since the penetration of distributed generators in the utility power system is continually increasing. Accurate threshold setting is an integral part of the island detection scheme since an inappropriate threshold might cause a hazardous situation. This study looked at the islanding conditions as well as two transient faults, such as a single line to ground fault and a three-phase balance fault, to assess the event distinguishing ability of the proposed method. Therefore, the goal of this research was to determine the threshold of the island if the distributed generator (DG) capacity is greater than the connected feeder load, which is the over-frequency island condition, and if the DG capacity is less than the connected feeder load, which is the under-frequency island condition. The significance of this research work is to propose a new island detection threshold setting method using the slip angle and acceleration angle that comes from phasor measurement unit (PMU) voltage angle data. The proposed threshold setting method was simulated in the PowerWorld simulator on a modified IEEE 30 bus system equipped with DG. There are three different interconnection scenarios in the test system and the performance of the proposed method shows that getting the island threshold for all the scenarios requires a single time step or 20 mile seconds after incepting an island into the network. In addition, it can distinguish between the real islanding threshold and the transient faults threshold.

1. Introduction

The transition from a traditional grid to a microgrid with the combination of DG units is happening at a rapid pace around the world [1]. Despite the great economic and environmental well-being provided by renewable DG, it has its own drawbacks, such as difficulties in power system operation, control, and security [2]. Voltage fluctuations, power fluctuations, and power quality problems are all caused by non-dispatchable renewable energy sources’ intermittent behavior [3]. The microgrid’s DG units are still energized to satisfy local demand in an island state, but the microgrid is electrically disconnected from the utility grid [4]. As a result, island identification in a microgrid is unavoidable in order to fulfill its role. Furthermore, in the case of unscheduled islanding, the system’s behavior may be unpredictable [5]. IEEE 1547-2003 and IEEE 1547a-2014 guidelines recommend that DG be disconnected within 2 s [4]. The microgrid may be disconnected from the rest of the network for an isolated service in the event of a recurrent fault in the utility [6]. If a fault occurs inside the microgrid and the circuit breaker is opened, the islanding detection algorithm can work to isolate the smallest possible faulted region of the microgrid [7]. Furthermore, the islanding detection algorithm should have the ability to distinguish between true islanding and transient events such as the initiation and clearing of a fault in a nearby feeder [8]. Local and remote islanding detection techniques (IDTs) have been described in the literature [9]. IDTs are divided into two categories based on local measurements: passive and active techniques. When the power difference between the DG and the load is very low or zero, the passive methods fail [10]. The advantage of the active technique is the minimum non-detecting zone (NDZ), but it decreases the power quality [11,12,13,14,15,16,17,18]. The combination of active and passive techniques is able to decrease the NDZ and also maintain the desired power quality, but the algorithm is unable to differentiate between the events [7,19,20,21]. An effective islanding detection algorithm should be able to detect all possible islanding scenarios, but most schemes currently focus only on grid-side faults [22].

An effective and fast monitoring system using systemic principal component analysis was proposed in [19]. However, their scheme would have necessitated longer computational time. In [23], they proposed a multi-functional fault detection algorithm by using PMU data of voltage angle with dynamic monitoring and supervision. Pattern recognition techniques for island events based on transient signals have been discussed [24,25], but the implementation of this proposed system is extremely difficult. Utilizing PMU data for detecting islanding events has been proposed [26,27]. In [26], they used frequency difference to detect islanding events, whereas [27] used the change in phase angles, but those algorithms are not effective for the match frequency condition due to the minimal voltage and frequency excursion in the match frequency condition.

For detecting islanding events, decision tree (DT) approaches show higher accuracy performance [28,29]. The supervisory control and data acquisition (SCADA) system is a very popular approach for detecting islanding events but is unable to feedback accurately due to the conveying delay and also the higher investment required for the installment [30]. An islanding detection scheme using the voltage angle difference between two PMUs was discussed in [31], where detection time showed better performance but the under-frequency and over-frequency islanding condition performance was absent from their algorithm. Differentiating the islanding operation with the normal operation using PMU voltage angle was discussed in [32], where a simulation was conducted in the IEEE 30 bus system, but they did not consider the transient fault event. In [33,34], they proposed a probabilistic component analysis-based islanding detection method. The proposed method showed faster detection without false triggers but the scheme was collapsed detect when phase angle and frequency are well matched during islanding events. A voltage angle and current angle-based islanding detection algorithm was proposed in [35,36], where they lessened the false triggers but did not show any performance analysis on the match frequency islanding condition. A voltage magnitude-based islanding detection method was proposed in [37,38], but the proposed method was unable to detect islanding of the minimal power exchange condition due to the local detection scheme. In [39], they calculated the current flow through the breaker, and in that case, an island was detected before opening the breaker. However, in that case transmission loss was considered as the DG islanding conditions. In [40], they used a ridgilet probabilistic neural network to detect the islanding and were able to detect islanding but did not consider the transient faults.

The goal of this research was to figure out the island thresholds with a minimum time and distinguish islands from transient faults. The challenges of this research work were (1) the phase angle value of the different DG islanding conditions being located in dissimilar regions, (2) setting a single threshold that would be applicable for the under-frequency and over-frequency islanding conditions, and (3) distinguishing between the real islanding event and the transient fault. This research work proposes a new slip angle and acceleration angle-based island detection threshold setting method.

The proposed island threshold setting method is different from other schemes because it is able to distinguish between the real islanding threshold and the transient fault threshold by a single time step. According to other literature, most islanding detection schemes do not distinguish transient faults from the island event, which means that detecting a transient fault as an island event is possible. However, in reality, a transient fault is not island event. However, the studies that attempted to address this issue discovered a longer island detection time as a result of the inappropriate threshold setting. Therefore, this research work identified that specific research gap and worked to set the appropriate island detection thresholds, which ensure event differentiation with faster island detection. The remainder of the paper is organized as follows: The methodology is explained in Section 2. The simulation of the system is in Section 3, and the results discussion is in Section 4.

2. Methodology

2.1. Slip Angle and Acceleration Angle

The proposed islanding detection threshold determining methodology focuses on the slip angle and acceleration angle. This scheme calculates the slip angle based on the rate of change of the angle difference of the DG and the utility bus with respect to the time, and the acceleration angle is the rate of change of the slip angle with respect to the time. This proposed novel slip angle and acceleration angle theory was motivated by [41,42].

$$\mathrm{Slip} \mathrm{Angle}=\frac{d}{dt}\Delta v$$

(1)

$$\mathrm{Acceleration} \mathrm{Angle}=\frac{d}{dt}\left(\frac{d}{dt}\Delta v\right)$$

(2)

where Δv is the voltage angle difference between the DG and the utility bus. However, the slip angle and acceleration angle come from the PMU voltage angle data.

2.2. Proposed Island Detection Threshold Setting Method

To detect island events in a timely manner, a substantial part involves setting the threshold precisely to detect islands as well as differentiate the transient faults. Generally, if the threshold is not set properly the scheme cannot detect islanding events or may require more time. Figure 1 shows the proposed threshold setting method of the islanding event and transient fault. There are five units in this proposed method: bus, determining load threshold, island inception, threshold value calculation, and threshold setting.

Details of the proposed islanding detection threshold setting methodology are as follows:

(a) 

Bus

It is critical to know which buses we can use for our calculation when determining the island threshold setting. Figure 1 depicts two buses, one of which is a DG bus and the other a utility bus. We used a modified IEEE 30 bus system in this research work, and there were two DG buses: Bus 11 (33 KV) and Bus 13 (33 KV). In this research work, we considered Bus 13 (33 KV), and for the utility buses, Bus 3 (KV).

(b) 

Determining Load Threshold

Different load values result in different network conditions. From Figure 1, it is evident that the load could be increased until the generator loses its synchronism [43]. However, the under-frequency condition allows the generator to trip 0.2 s at a frequency less than 46.5~48 Hz, and the over-frequency condition allows the generator to trip 4 s at a frequency greater than 51~52.5 Hz [42]. Only a permissible load can be determined for under-frequency and over-frequency conditions from there.

(c) 

Island Inception

Islanding occurs when generators and loads are disconnected from the main system but remain energized [4]. Voltage and frequency fluctuate, resulting in power quality degradation and synchronization issues on the utility grid.

(d) 

Threshold Value Calculation

This section consists of three important calculation segments: the angle difference, slip angle, and acceleration angle. From the PMU voltage angle data, the voltage angle difference between DG bus and utility bus is calculated, then using (1) and (2) the slip angle and acceleration angle value are calculated.

(e) 

Island Threshold Setting

Once all of the PMU slip angle and acceleration angle data from (d) have been calculated, a common threshold value can be created that will be applicable to all cases of under-frequency and over-frequency island conditions.

(f) 

Transient Faults

A substantial segment of this proposed methodology is to distinguish transient faults from the island. To figure out the transient faults threshold, the following steps are carried out:

Step 1. Perform fault current analysis at the utility buses.

Step 2. Find out the highest fault current bus.

Step 3. Create transient faults at the highest fault current bus by allowing 100 ms as fault clearing time according to [44].

Step 4. Get the voltage angle data of the transient faults.

Step 5. Calculate the slip angle from the voltage angle data using Equation (1).

Step 6. Take the maximum slip angle value from that 100 ms region as a transient fault threshold.

3. Simulation

This section discusses details of the test power system model. In Figure 2 there are three scenarios in this test system network. Details of the test system scenarios are as follows.

Scenario 1

The Bus 4 (132 KV) to Bus 12 (32 KV) interconnection point is connected, where Scenario 2 and Scenario 3 will both remain open. In this scenario Bus 4 (132 KV) imports power from Bus 12 (33 KV) to feed the connected load.

Scenario 2

In this case Scenario 1 is opened and the DG side is connected with the Bus 6 (132 KV). Here the interconnection point between Bus 6 (132 V) and Bus 9 (33 KV) is connected and Bus 9 (132 KV) exports power to Bus 6 (132 KV) to feed the Bus 4 (132 KV) connected load.

Scenario 3

In this scenario interconnections of Scenario 1 and Scenario 2 are both opened. Here the interconnection point between Bus 6 (132 KV) and Bus 10 (33 KV) is connected and Bus 10 (132 KV) exports power to Bus 6 (132 KV) to feed the Bus 4 (132 KV) connected load.

Table 1. Parameters of the test system.

Cases	Interconnection Points		Scenarios	Threshold Loads of Bus 4 (132 KV)	Conditions	DG1		DG2
	To Bus	From Bus				MW	MVAR	MW	MVAR
A1	4	12	1	0.5 MW	Over frequency	72	27	48	21
A2	6	9	2	0.80 MW			67		29
A3	6	10	3	1.10 MW			38		38
B1	4	12	1	18.600 MW	Under frequency		25		22
B2	6	9	2	18.85 MW			67		28
B3	6	10	3	18.150 MW			19		37
F1	4	12	1	0.5 MW	Single line to ground (SLG) fault		27		21
F2	4	12	1	0.5 MW	Three-phase balance (3PB) fault		27		21

and

Table 2. Data preoperational parameters of the test system.

Name of the Parameters	Settings
System frequency	50 Hz
DG bus	Bus 13 (Hancock)
Utility bus	Bus 3 (Kumis)
Island event	Open the line between Bus 3 and 4
Island incepted	2 s
Time step	0.02 s

present the parameters of the simulated power system model and data preparation. PMU data was prepared using a PowerWorld simulator.

4. Results and Discussion

This section discusses the results of the cases. The axes in Figure 3, Figure 4, Figure 5, Figure 6, Figure 7 and Figure 8 follow the

Table 3. Axis setting for the figures.

Figures	Legends
	x-axis	y-axis
3, 5	Timestamp	Frequency
4, 6, 7, 8	Timestamp	Voltage angle

.

4.1. Case A1, A2, and A3

Cases A1, A2, and A3 represent the bus frequency of the islanding conditions when the interconnection points went from Bus 4 to Bus 12, from Bus 9 to Bus 10, and from Bus 6 to Bus 9, and the DG was greater than the connected feeder load or the over-frequency condition.

Figure 3 and Figure 4 present the case A1 bus frequency and voltage angle of the DG and utility bus. Figure 3 is after incepting the island at a 2 s frequency, going up over the time and creating an over-frequency condition. The voltage angle profile of that condition at that moment is shown in Figure 4. The bus frequency and bus angle of cases A2 and A3 showed approximately similar behavior, according to Figure 3 and Figure 4.

Table 4. Threshold setting results of cases A1, A2, and A3.

Cases	Slip Angle (Deg/s)	Acceleration Angle (Deg)	Time
A1	169.815	5490.75	20 ms
A2	90.85	4542.5
A3	91.81	4590.5

presents the thresholds results of cases A1, A2, and A3. The calculation was conducted using Equations (1) and (2) from the Figure 4 voltage angle data. From Table 4, the highest slip angle and acceleration was figured out for case A1, and the lowest one was seen in case A2 at a time of 20 ms.

4.2. Cases B1, B2, and B3

Cases B1, B2, and B3 represented the bus frequency of the islanding conditions when the interconnection points went from Bus 4 to Bus 12, from Bus 9 to Bus 10, and from Bus 6 to Bus 9, and the DG was less than the connected feeder load or the under-frequency condition.

Figure 5 and Figure 6 illustrate the bus frequency and voltage angle of the DG and utility bus for case B1. Figure 5 is after incepting the island at 2 s frequency, going downward over time and creating an under-frequency condition. The voltage angle profile of that condition at that moment is presented in Figure 6. The bus frequency and bus angle of cases B2 and B3 showed an approximately similar pattern, according to Figure 5 and Figure 6.

For cases B1, B2, and B3 the values were comparatively smaller than cases A1, A2, and A3 due to the distance between the interconnection point and the placement of the PMU. The calculation in

Table 5. Threshold setting results of cases B1, B2, and B3.

Cases	Slip Angle (Deg/s)	Acceleration Angle (Deg)	Time
B1	0.525	26.25	20 ms
B2	1.5	75
B3	2.54	127

for each case was conducted using Equations (1) and (2) from the Figure 6 voltage angle data. The maximum and minimum slip angle and acceleration angle in Table 4 are for case B3 and case B1, respectively.

4.3. Case F1 and F2

Cases F1 and F2 represent the single line to ground fault and three-phase balance fault event in the distribution network. The single line to ground fault was applied at the highest fault current (3.31900 P.U) bus, which was Bus 3, and the three-phase fault was applied at Bus 4, where the fault current was 6.47348 P.U.

Figure 7 and Figure 8 show the transient fault incepting point and the DG and utility bus voltage angle behaviors of cases F1 and F2.

Table 6. Threshold results of case F1 and case F2.

Cases	Slip Angle (Deg/s)	Time
F1	14.57	20 ms
F2	503.945

presents the transient fault thresholds of case F1 and case F2, where the slip angle was calculated using Equation (1) from the voltage angle data in Figure 7 and Figure 8.

4.4. Threshold Setting

Considering islanding cases A1 to B3, we came out with a common threshold that will be applicable for all cases to differentiate between islanding events and transient fault events.

Table 7. Threshold settings for the islanding detection algorithm.

Slip Angle (Deg/s)	Acceleration Angle (Deg)	Thresholds
0.525	26.25	Islanding threshold
14.57	NA	SLG fault threshold
503.945	NA	3PB fault threshold

shows the islanding threshold and the transient fault threshold for the islanding detection algorithm.

4.5. Results Comparison

Table 8. Comparison of results with other islanding detection methods.

Ref.	Methods	Transient Faults	Detection Time
[45]	WSE	Not considered	<10 ms
[46]	ST	Not considered	26 ms
[47]	EMD	Not considered	<35 ms
[48]	TMF	Not considered	55 ms
[49]	VMD	Not considered	10 ms
[50]	ANN	Considered	0.5 s
[51]	SVM	Considered	50 ms
[52]	HoGF+ SVMs	Considered	218 ms
[53]	ELM	Considered	NA
[54]	Ensemble tree	Considered	NA
[55]	RF	Considered	NA
[56]	MIMF	Considered	<70 ms
[41]	Slip frequency and acceleration	Considered	500 ms
	Proposed method	Considered	20 ms

shows the results of a comparative analysis of the proposed algorithm with other recent methods. It shows that in [45,46,47,48,49], the transient fault events that are crucial in islanding detection were not analyzed, whereas [41,50,51,52] showed a higher detection time and [53,54,55] was unable to figure out the islanding detection time. However, [56] shows less detection time by considering transient faults but compare to the proposed algorithm that detection time [56] is bit higher.

As shown in Table 8, [50,51,52,53,54,55,56] considered transient faults outside the utility grid, where the proposed method focused on the inside of the utility grid. As island occurs when the utility side opens from the distribution side, so it is important to scrutinize the transient fault inside the utility grid. However, [41] considered transient faults inside the utility grid but managed a higher island detection time of 500 mile seconds. On the other hand, the proposed method showed a better island detection time of 20 mile seconds compared to the other methods. In addition, like other methods, the proposed method follows IEEE standard 1547, which is 2 s, so it can be used in a real power system to detect islands.

5. Conclusions

The motivation of this research work was to detect islanding events for under-frequency and over-frequency systems. To detect islanding events, it is very important to set a proper threshold, because an inappropriate threshold can increase the islanding detection time. The following are the important discussion points:

• This research work proposed a novel islanding threshold setting method using the slip angle and acceleration angle.
• The proposed research work overcame the limitation of the local area-based islanding detection threshold setting.
• The proposed research method can distinguish the actual islanding event and the transient fault event.

Future work will develop an islanding detection algorithm to detect the “big island” in the Utility Kerteh, Malaysia system using this proposed threshold setting methodology.