A NEURAL•FUZZY CLASSIFIER FOR RECOGNITION OF POWER QUALITY DISTURBANCES

Abstract

Power quality monitoring has advanced from strictly problem solving to on going monitoring of system performance. The increased amount of data being collected requires more advanced analysis tools. New intelligent system technologies using expert systems and artificial neural networks provide some unique advantages regarding fault analysis. Advances in signal processing and artificial intelligence tools will be examined for their role in the detection and classification of events, the application of various mathematical transforms and the implementation of rules-based expert systems. This report focuses further on the review on several implementation methodologies, and a performance comparison of existing implementations and proposed methodology. A system for the identification of power quality violations is proposed. It is a two-stage system that employs the potentials of the wavelet transform and the adaptive neuro-fuzzy networks. For the first stage, the wavelet multiresolution signal analysis is exploited to denoise and then decompose the monitored signals of the power quality events to extract its detailed information. A new optimal feature-vector is suggested and adopted in learning the neuro-fuzzy classifier. Thus, the amount of training data needed is extensively reduced. Extracting fuzzy rules from the optimal feature-vector allows relationships in the data to be modeled by "if-then" rules that are easy to understand, verify, and extend. This report presents methods for extracting fuzzy rules for pattern classification. The rule extraction methods are based on estimating clusters in the data; each cluster obtained corresponds to a fuzzy rule that relates a region in the input space to an output class. After the number of rules and initial rule parameters are obtained by cluster estimation, a combination of back propagation and least square methods are used to tune. the parameters of the membership functions enabling the fuzzy system to best track the given input/output data. Simulation results confirm the aptness and the capability of the proposed system in power quality violations detection and automatic diagnosis. Recommendations for future study are also outlined. This report opens the path for researchers to future comparative studies between different architectures, and as a reference point for developing more powerful and flexible structures. iii