An Independent Component Analysis Classification for Complex Power Quality Disturbances With Sparse Auto Encoder Features

This paper introduces a method to detect multiple power quality (PQ) disturbances of power system features based on an independent component analysis (ICA) and a sparse auto encoder (SAE). Seven basic single PQ disturbances are extracted as typical features from numerous training samples of PQ disturbances by the SAE method, which can automatically gain the training features rather than manually selecting features as in the conventional approaches. An ICA is adapted to conduct basic separate signals from the blind original disturbance sources. The experimental results indicate that the proposed method has not only achieved a significant improvement for detecting single PQ disturbances but is also effective for detecting and classifying multiple PQ disturbances.