A ROBUST WAVELET–ANN FRAMEWORK FOR NOISE-AWARE DETECTION AND CLASSIFICATION OF POWER QUALITY DISTURBANCES
Keywords:
Power quality disturbances, discrete wavelet transform, multiresolution analysis, feature extraction, artificial neural networks, MLP, RBF, and PNNAbstract
The rapid penetration of power-electronic-based loads, renewable energy sources, and sensitive digital equipment has significantly increased the occurrence and complexity of power quality disturbances (PQDs) in modern electrical power systems. Accurate and automated detection and classification of PQDs remain challenging due to the non-stationary nature of disturbance signals and the presence of measurement noise. This paper presents a comprehensive, noise-aware hybrid framework integrating discrete wavelet transform (DWT) based multiresolution analysis (MRA) with artificial neural network (ANN) classifiers for reliable detection and classification of PQDs.
Standardized single and combined PQ disturbance signals are generated in accordance with IEEE Std. 1159 and sampled at 10 kHz. DWT–MRA is employed for denoising, decomposition, and extraction of discriminative statistical features from multiple resolution levels. A systematic evaluation of diverse mother wavelet families is conducted to identify the most suitable wavelet for PQD representation. The extracted features are classified using multilayer perceptron (MLP), radial basis function (RBF), and probabilistic neural network (PNN) classifiers. Performance is evaluated under varying signal-to-noise ratio (SNR) conditions ranging from 20 dB to 50 dB.
Simulation results demonstrate that the proposed framework achieves superior and consistent classification accuracy across all disturbance types and noise levels. Comparative evaluation with recent state-of-the-art techniques confirms that the proposed wavelet–ANN approach provides a computationally efficient, interpretable, and highly accurate solution suitable for real-time power quality monitoring applications.
