An Expert System Based on Least Mean Square and Neural Network for Classification of Power System Disturbances

Abstract

This paper proposes a new solution method for power quality (PQ) classification using least mean square (LMS) and neural network (NN). The proposed hybrid LMS-NN method comprises of LMS based effective feature extractor and PQ classifier based on a multi layer perceptron neural network (MLP-NN). First, the LMS method is employed to estimate the efficient features such as amplitude, slope, and harmonic indication from the measured voltage signals where the developed structure is merely simple. Further, the PQ classification is executed with the aid of MLP-NN. The different voltage signals analyzed for this research work are pure sine, sag, swell, outage, harmonics, sag with harmonics, and swell with harmonics. The performance and efficiency of the presented hybrid LMS-NN classifier is assessed by testing total 1400 voltage samples which are simulated based on PQ disturbance model. The rate of average correct classification is about 96.71 for the different PQ disturbance signals under noise conditions.