Identification of power quality disturbances based on dynamics and artificial neural networks

This paper proposes a novel identification method of power quality (PQ) disturbances based on Dynamics and artificial neural networks (ANN). Firstly, the PQ disturbance data are preprocessed, and then, feature vectors as the inputs of the ANN classifier are extracted through the Dynamics algorithm, and finally, PQ disturbance types are automatically identified by the ANN classifier. The Dynamics algorithm can detect all the extremum points of the signal, identify distortion points according to the great difference between the Dynamics of distortion points and those of peak or valley points. Elements of the feature vector include energy feature, duration feature, amplitude range feature and amplitude variation feature. The parallel structure is adopted in the ANN classifier. Each sub-ANN, which can only identify a type of disturbance, is constructed with BP network and adopts the momentum-adaptive learning BP algorithm to improve its convergence performance. Simulation and test results certify the accuracy, validity, and high correct identification rate of the proposed method.