Performance Improvement of Power Quality Disturbance Classification Based on a New De-noising Technique

Identification, localization and classification of power quality disturbance is the precondition of appropriate mitigation actions that can be taken. However, the signal under investigation is often corrupted by noises, especially the ones with high frequency signal produced by EMI. A new de-noising method is proposed to improve the classification performance for power quality disturbance. The proposed method first estimate the threshold of different decomposition level of wavelet transform for power frequency sinusoidal signal contained the noises. Secondly, each detailed coefficient of decomposition level of wavelet transform thresholded by the stored threshold value before. Wavelet transform and Parseval's theorem are used to extract the feature of the modified signal at different resolution levels. The extracted features can be characteristic of the disturbances by choosing the nice threshold value, so as to enhance the correct classification rate of power quality disturbance polluted by noises.