Optimal neuro-fuzzy based islanding detection method for Distributed Generation

Utilizing wavelet transform, in this paper an intelligent and optimum method is presented for islanding detection. The recommended relay is based on neuro-fuzzy system which is trained by using four different heuristic algorithms and finally among all of them, PSO with having the best results is elected for proposed relay. In this procedure, selection of mother wavelet and also proper input signals for islanding detection is done optimally. To do so, firstly, various islanding and switching in different loading conditions of network are simulated. Then, the eligible signals for islanding detection are sorted based on wavelet analysis and receiving operating characteristic (ROC) concept. Afterward, using four different algorithms, training of the designed relay is done by the sorted signals and the optimum condition of the fault detector is recognized. Reduction of non-detection zone approximately to zero, selection of optimum number of relay׳s input signal, fast detection, solving of threshold selection, appropriate signal selection and fast training via least information are the prominent features of the proposed method. Simulation results and comparison of them for different cases approve the efficiency of the proposed method.