A new fault location technique on radial distribution systems using artificial neural network

In this paper, an effective fault location technique is proposed. Using the current samples of the distribution feeder measured at the substation, the proposed technique first determines the type of fault. Furthermore, an artificial neural network (ANN) is trained for each type of fault. The ANNs are trained to estimate the fault distance to the substation (FDANN).The Inputs of the ANNs are data of 3 phase voltages, currents and active powers of the feeder are measured at the substation in pre-fault and fault stages. The proposed method does not need data of loads of consumers. The proposed method is tested on IEEE 34-bus test feeder. Each ANN is trained by operating patterns. In order for ANNs cover the total operating space of the radial distribution network; fault location, fault resistance and loads are changed in each pattern. The outputs of ANNs for the operating test patterns, not presented in the training stage, are shown the accuracy of the ANNs. The trained FDANNs can estimate fault distance to the substation; even the structure of the distribution network is changed. Proposed method is effective while the input data are contained errors of measuring.