Sunday 06 April 2025
The quest for more reliable and efficient fault detection in power grids has led researchers to explore innovative approaches, including machine learning techniques. A recent study published in IEEE Transactions on Power Systems presents a novel method that leverages wavelet coefficients and recurrence plots to identify internal faults and high impedance faults (HIFs) in medium-voltage distribution lines.
The problem of HIF detection is particularly challenging due to the subtle nature of these events, which often exhibit low current levels and resemble normal load currents. Traditional protection schemes may fail to detect such faults, leading to extended periods of undetected operation and potential cascading failures. The proposed method aims to overcome this limitation by analyzing wavelet coefficients derived from 3-phase current measurements.
Wavelet analysis is a powerful tool for extracting relevant features from non-stationary signals like power grid currents. In this study, the researchers employed a Continuous Wavelet Transform (CWT) to decompose the current signals into different frequency bands. This allowed them to identify characteristic patterns associated with internal faults and HIFs.
The CWT coefficients were then converted into recurrence plots, which provide a visual representation of the recurrence of states in the signal. Recurrence plots have been widely used in dynamical systems analysis to identify periodic or chaotic behavior. In this context, they help researchers distinguish between fault-free operation and faulty conditions.
To evaluate the effectiveness of their approach, the researchers conducted extensive simulations using PSCAD/EMTDC software. They simulated a range of internal faults, including HIFs, as well as external faults and normal operating conditions. The results showed that the proposed method achieved high accuracy in identifying internal faults (99.24%) and HIFs (98.26%).
The study also compared the performance of several machine learning algorithms, including Random Forest, Decision Tree, Gradient Boosting, AdaBoost, k-Nearest Neighbors, and Multi-Layer Perceptron. The tree-based classifiers, particularly Random Forest and Decision Tree, emerged as the top performers in both internal fault detection and HIF identification.
The authors’ approach offers several advantages over traditional protection methods. For instance, it can handle non-linear relationships between current signals and adapt to various fault scenarios. Additionally, the use of wavelet coefficients and recurrence plots enables the method to distinguish between different types of faults.
While further research is needed to validate the proposed method in real-world settings, this study demonstrates the potential of machine learning techniques for improving power grid reliability and efficiency.
Cite this article: “Unlocking Fault Detection in Microgrids: A Machine Learning Approach to High-Impedance Fault Identification”, The Science Archive, 2025.
Power Grids, Fault Detection, Machine Learning, Wavelet Analysis, Recurrence Plots, Internal Faults, High Impedance Faults, Medium-Voltage Distribution Lines, Power System Protection, Signal Processing
