Application of probabilistic neural network with transmission and distribution protection schemes for classification of fault types on radial, loop, and underground structures

Abstract

This paper proposes the development of transmission and distribution protection schemes to classify faults along transmission and distribution systems. The systems under consideration are composed of a 500-kV radial (two-bus single circuit), loop (three-bus double circuit) structure transmission systems, and a 115-kV radial structure underground distribution system. This complex system shows the advantage of the proposed method. A decision algorithm based on a discrete wavelet transform (DWT) and probabilistic neural network is investigated for inclusion in a transmission and distribution protection system. Fault signals in each case are extracted to several scales on the DWT to decompose high-frequency components from fault signals using the mother wavelet daubechies4. The maximum coefficients of a DWT at 1/4 cycle that can detect faults are used as input patterns for the training process in a decision algorithm. In addition, coefficients from a DWT technique and a back-propagation neural network algorithm are also compared with the proposed algorithm in this paper. Moreover, the real signal from an experimental set-up was investigated. Based on the accurate results for the simulation signal and real signal, it can be concluded that the proposed algorithm is adequate for other power systems with different line models and that it can be applied to actual systems even if the accuracy is slightly reduced by the effect of noise in an actual system. Thus, the overall results show that the proposed algorithm can detect a faulty bus and classify types of fault with satisfactory results.