Maximal overlap discrete wavelet transform and deep learning for robust denoising and detection of power quality disturbance

Abstract

This study presents a new technique for power quality (PQ) disturbance detection. The technique focuses on voltage sags and interruptions that are related to various faults, i.e. transmission line, feeder, and transformer faults. A maximal overlap discrete wavelet transform‐based PQ detection algorithm is proposed to provide accurate points of disturbance initiation and recovery. The proposed PQ detection algorithm is robust even without a detection threshold and independent of the sampling frequency of PQ recording. In consideration of the presence of noise conditions, the preprocessed PQ waveforms are converted into 2D binary vectors using space vector transformation. Then, an improved stacked sparse denoising autoencoder combined with supervised backpropagation training is proposed as a robust classifier. Results show that the proposed method is suitable for detecting various types of PQ disturbances and possesses high recognition accuracy despite insufficient training samples.