AN ADVANCED WAVELET–AI FAULT CLASSIFICATION FRAMEWORK FOR ENHANCED PROTECTION OF SERIES-COMPENSATED TRANSMISSION SYSTEMS
Keywords:
Series Compensation, Discrete Wavelet Transform (DWT), Daubechies Wavelet, Artificial Neural Network, Fault Classification, Multilayer Perceptron, Radial Basis Function, Probabilistic Neural Network, Transient StabilityAbstract
Series-compensated transmission systems play a vital role in enhancing power transfer capability and improving steady-state stability in long-distance AC networks. However, their dynamic behavior under fault conditions introduces complex transient phenomena, including high-frequency oscillations and sub-synchronous resonance (SSR), which challenge conventional protection strategies. Traditional phasor-based relaying techniques often exhibit reduced reliability due to waveform distortion introduced by compensation devices and nonlinear protective elements.
This paper presents an advanced hybrid Wavelet–Artificial Intelligence (AI) framework for intelligent fault classification in a series-compensated transmission system modeled in MATLAB Simscape Electrical. Multi-resolution analysis using the Daubechies-5 (db5) wavelet at level-5 decomposition is employed to extract transient signatures from three-phase current signals. Statistical features derived from wavelet coefficients are utilized to train and comparatively evaluate Multilayer Perceptron (MLP), Radial Basis Function (RBF), and Probabilistic Neural Network (PNN) classifiers.
The proposed system is validated under single-line-to-ground (SLG), line-to-line (LL), and three-phase-to-ground (LLLG) faults with varying fault resistance and compensation levels. The PNN classifier achieved the highest classification accuracy of 98.5%, outperforming the MLP and RBF networks. Results demonstrate enhanced classification accuracy, robustness, and rapid detection capability, supporting improved protection performance in modern series-compensated transmission networks.
