Development of a Fuzzy Impedance Distance Protection Scheme to Enhance Fault Detection and Mitigate Power Swings in Electrical Power Grid

In power systems, distance relays are key protection devices due to their simplicity and ability to locate faults using local voltage and current measurements, but they may fail during power swings, making power swing blocking (PSB) vital for proper operation [1]. A number of techniques have been put forth by researchers to identify power swing (PS) and initiate PSB function. The most popular way to do this is to use the rate at which impedance varies in a certain location [2, 3]. Accordingly, monitoring the temporal variation of this rate enables a detector to separate PSs from faults. Shift periods in each zone are at least as long as the nominal frequency, and impedance shifts occur exceedingly slowly for slow PSs (1 Hz) [4]. This method works well to detect a slow PS and turn the relay off. However, it is unable to detect faults that occur during a PS or quick PSs (5 Hz) [5, 6]. A “double-blinder” arrangement was proposed to address the shortcomings of simple impedance-rate detectors. Despite some downsides, it often achieves the desired outcome. It requires in-depth offline analysis to determine the right PSB values, and fails to distinguish between faults and quick power swings, as well as handle obstructions during faults [7]. Distance relays are crucial for protection, particularly in scenarios involving "under-reach" and "over-reach", as they rely on impedance measurements between the relay fault location [8, 9]. Impedance fluctuations may be a dependable predictor of fault occurrence because distant relay impedance changes during power system disturbances but stays constant during faults [10]. Among FACTS technologies, the Static Synchronous Compensator (STATCOM) is widely used; however, its presence poses additional difficulties for distance relays, affecting response accuracy and complicating discrimination between faults and power swings [11, 12]. Relay settings have been adjusted to mitigate these effects, as devices such as the Unified Power Flow Controller (UPFC) also affect the impedance, causing problems for impedance-based protection units in differentiating between power swings and faults [13]. The importance of PS detection is highlighted using a combination of concentric impedance characteristics and continuous apparent impedance monitoring due to impedance change throughout the transient period [14]. The reduced-resistance technique is capable of distinguishing power swings from faults without requiring detailed network analysis, and it can also identify faults that occur during swings. Nevertheless, its performance declines when a three-phase fault happens simultaneously with a delayed swing [15]. A detection method is proposed that utilizes trajectory monitoring to identify the separation point between stable and unstable PSs, improving the ability to block the relay at the correct moment [16]. Faults during power swings were detected using a fast Fourier transform (FFT) of the DC component of the current. A proposed FFT-based method was developed for fault detection, but its ability to detect faults relies on specific values without a straightforward procedure [17]. To overcome FFT limitations, wavelet analysis was employed [18]. Relay logic then processed data acquired from numerical relays at a 40.96 kHz sampling rate (fs) [19]. The detection of symmetric faults has been addressed by a high-speed method as given in [20]. A method is proposed that applies the wavelet transform (WT) to incorporate impedance measurement and detect changes caused by swings and faults [21]. Wavelet transform (WT) has been used to analyze current changes in order to detect faults with various resistances during power swings (PSs) [22]. For power system protection, one study proposed a technique grounded in Prony analysis [23]. An analysis of several states revealed that the Prony method performs better than FFT-based methods [24]. Additionally, an algorithm has been proposed to detect three-phase faults during PSs by using the current waveform, with this method of operation relying on the reverse decline in the DC current rate [25]. Hybrid renewable energy systems, including solar and wind, aid in reducing power swings in times of emergency, enhance system stability, and reduce losses [26, 27]. Using steam turbines with several masses during major transitions improves system stability by preventing synchronisation loss [28]. It has been confirmed that changes in system characteristics do not alter the effectiveness of artificial intelligence-based methods for system security in studies [29-31]. Several PS detection methods have been developed and are based on the adaptive neuro-fuzzy inference system (ANFIS). In addition to being unaffected by variables like system characteristics and fault location, these methods distinguish PSs from faults with exceptional speed and precision [32-34]. A new relay configuration employing fuzzy logic has been introduced to lessen the impact of UPFC and to improve both the speed and accuracy of fault detection [35]. Transient stability in a two-machine, three-bus system was improved using a fuzzy logic–controlled Static Var Compensator (SVC), reducing rotor angle overshoot by 11.94%, improving settling time by 4.7%, and compensating about 16.2% of the system’s equivalent kinetic energy [36]. Phasor quantities and network admittance have been employed to discriminate between power swings and coincident faults. The analysis indicates that fundamental-frequency voltage and current phasors exhibit discriminative features that are especially useful for protective decision-making when symmetrical three-phase faults occur during scheme operation [37]. A detection method based on moving window averaging was proposed. A low-pass filter is applied to the distance relay using the moving window averaging approach. One of the major problems with this method is its inability to detect multi-mode PSs. Additionally, PSs that arise in a single phase cannot be detected by this method [38]. The proposed detection method is based on the rate of change in the root-mean-square value of the current. When PS happens in a single phase, this method wrongly kills the line, but it can detect multi-mode PSs and other kinds of high-resistance faults during PSs [39]. PSs and faults have been distinguished using the instantaneous power frequency. Unstable and stable PSs, as well as simultaneous single-phase high-resistance faults, may be accurately detected using this method; however, three-phase high-resistance faults prevent the relay from being stopped [40]. A detection method based on the rate of change of the average apparent power has been proposed. PSs that arise in a single phase cannot be detected by this method [41]. For determining the rate of faults that occur during PSs, the proposed method takes advantage of changes in the negative component of apparent power. Multi-mode PSs cannot be detected by this method, which has a somewhat slow response time in detection [42]. Deep learning models like Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks have demonstrated strong potential in power system protection by enhancing fault detection, differential protection, and anomaly identification with faster and more accurate responses [43]. Hybrid CNN-LSTM frameworks, optimized with genetic algorithms, can also detect false data injection attacks by capturing spatial and temporal patterns, achieving higher precision than conventional methods [44]. The impact of fault resistance and line length on protection mechanism effectiveness was analyzed. Results showed significant effects on relay performance and system stability. These findings align with the current study's assumptions, reinforcing the chosen approach's reliability and the importance of considering these factors for protection improvements [45].

All the previous studies have not included and collected the intelligent fuzzy distance relay that distinguishes faults and power swings of various types and performs load shedding based on the stability detection curve. The goal of this paper is to design two fuzzy systems: one to improve fault detection through a fuzzy impedance distance protection system, and another to prevent critical power swings by shedding load and restoring system stability. It also aims to activate the main protection in case load shedding is insufficient or faults occur, ensuring continued system reliability.

In the event of a load loss or a malfunction in the protection scheme, remedial actions—such as disconnecting the line or isolating the faulted section are applied. Figure 2 shows, in red, the time-domain characteristic of the planned impedance element for a 700 km transmission line. This characteristic extends over 20%-80% of the line length and is represented by a quadrilateral zone. The inclined black boundary indicates the suggested stability limit and offers an early signal of whether the system, immediately after a fault, is moving toward a stable or unstable power swing. During the fault, the observed impedance trajectory (violet) is taken as the reference signal in the proposed approach, allowing both fault identification and assessment of whether the path tends toward a stable or unstable region. Although not central to this work, the purple star designates the first intersection (T1) with the stability margin. It also marks the entrance instant (T2) into the distance-relay zone, while the green marker corresponds to the exit instant (T3). The time span between T2 and T3 reflects the relay’s fault sensitivity. The green star then highlights the second crossing with the suggested stability–swing boundary, denoted T4, and the interval ΔT = T4 − T2 is applied to evaluate the power-swing status of the system, indicating if it is moving toward stability or instability.

2.jpg

Figure 2. Impedance trajectory and distance relay characteristic with stability limitation

The present study proposes a line or curve for the stability constraints that indicate the occurrence of a malfunction. The approach proposed here begins by first locking the relay characteristic and the stability limits. When a severe disturbance or fault occurs, the X/R ratio surpasses the stability threshold, and the first intersection with the stability boundary labelled as T1 is recorded. Afterwards, the X/R trajectory keeps shifting until it enters the relay zone, where the initial crossing (T2) is logged as the fault-detection instant. The X/R value continues to change until it goes beyond the relay-zone limits, at which moment the second crossing of the relay characteristic is registered as T3. Relay fault-detection sensitivity is assessed using the interval T3–T2. A larger T3–T2 implies higher relay sensitivity. Until it crosses the stability limits once again, the value of X/R keeps changing automatically; at that time, the second intersection denoted T4 is recorded. Then compute ΔT = T4 − T2. If ΔT lies in 70–80 ms and T4 is in 80–90 ms, it indicates that the critical swing occurs in the fifth cycle, which is the proposed stability threshold at which the isolation signal will be sent. If instead ΔT is 80–90 ms and T4 exceeds 100 ms, it indicates that the critical swing occurs in the sixth cycle. The proposed stability threshold will shift to the sixth cycle instead of the fifth. Subsequently, the regulated load isolation procedures will begin to avoid losing synchronisation. The system will then be monitored to see if the isolation procedures have led to system stability. If the answer is yes, the load will be returned to the system after it stabilizes. In case of failure, a command will be given to disconnect the lines to avoid damage caused by losing synchronisation. If ΔT < 70 ms, the timestamps are cleared and the evaluation is repeated until the criterion is satisfied. Figure 3 illustrates these operations in the form of a flowchart.

3.jpg

Figure 3. Overall flow chart for the proposed procedure

Through simulation and practical experience, fuzzy rules were designed to improve the system's performance. The Sugeno fuzzy inference method was chosen for the stability detection curve, using 625 rules and 25 membership functions, and for the distance phase with 1225 rules and 35 membership functions. Based on the number of membership functions utilized in the modeling, the number of rules was established. To achieve the lowest error rate during the training process, the number of membership functions was manually entered and tested using the "Neuro Fuzzy Designer" tool. The number of rules was determined by multiplying the number of membership functions, where 25 × 25 membership functions result in 625 rules, and for 35 membership functions, 35 × 35 results in 1225 rules. In a fuzzy inference system (FIS), the process of transforming fuzzy output into precise output is referred to as defuzzification. Within the framework of the defuzzification technique, the output of fuzzy rules is converted into a numerical output that is not fuzzy. This particular stage is incredibly significant as it is responsible for generating the final, accurate output of the fuzzy logic controller, which is achieved through the adept utilization of fuzzy inputs that have been carefully processed. Figures 7 and 8 serve to illustrate the three-dimensional output surface of the fuzzy inference system, which is specifically employed for the purposes of analyzing the stability detection curve and assessing the distance stage, thereby providing valuable insights into the system's operational dynamics and performance characteristics.

7.jpg

Figure 7. 3D input-output dataset mapping for the proposed fuzzy distance relay

8.jpg

Figure 8. 3D input-output dataset mapping for the proposed fuzzy stability limitation curve

This study introduced a combined protection system for a power network with three buses and two generators, looking at how three-phase faults affect the system's transient stability. Detailed analysis indicated that fault resistance and its location play a decisive role in the timing of stability loss, where the loss of synchronization differs between the sixth and seventh cycles based on fault resistance values and their location.

A sophisticated relay and an adaptive stability detection curve predicated on fuzzy logic were formulated, facilitating rapid and precise identification of faults alongside various manifestations of power swings. The investigation encountered obstacles in the categorization of power fluctuations owing to the analogous nature of resistance (R) and reactance (X) values during the fault interval, thereby complicating the differentiation between stable and unstable fluctuations; however, enhancements were made to the stability detection thresholds in targeted regions to augment detection precision.

The proposed protection system depends on precise timing measurements (T2 and T4) and monitoring of phase angles (θ1 and θ2) to improve the timing of breaker-tripping and load-shedding procedures, which support system stability under different fault conditions. Simulation results confirmed the proposed strategy's effectiveness in detecting faults early and determining whether power swings lead to stability or loss of synchronization after fault clearance. If the difference time value (T4-T2) indicates that the load-shedding procedures are insufficient, a signal to disconnect the line should be issued to prevent damage caused by losing synchronization.

The study showed that the proposed system can adapt to different fault resistances and their locations. The system implements advanced protection procedures that help improve the electrical network's reliability and stability. The study opened the door to developing more advanced protection systems in the future. These new systems can rely on artificial intelligence and fuzzy logic to ensure effective responses in future power networks. It is important to note that these results indicate the possibility of significantly improving the performance of electrical networks.

As future work, the performance of the proposed scheme can be applied to a large power system for enhancing overall transient stability, and also can be implemented on a scaled-down real-world prototype, with an analysis of clock cycles and hardware feasibility. Furthermore, systematic comparisons can be conducted with conventional protection methods, such as impedance relays and neural network–based approaches, to enhance the reliability of the results and assess the practical applicability of the proposed method.