Optimized Artificial Neural Network Method for Underground Cables Fault Classification

the electrical power network is the biggest system made by human in the world. With increasing the demand for this type of energy, several problems have been emerged in electrical power network. In this condition the new and complicated problems have been emerged in these large networks. One the most important problems in these networks are the occurred faults in underground cables. This study investigates the efficient approach for detecting these faults with high accuracy. In this study we consider four different states in cables that are normal condition, one phase fault, two phase fault and three phase fault. This study proposes the application of multilayer Perceptron (MLP) neural networks as a classifier. MLP neural networks are powerful and efficient classifiers among other classifiers. In the MLP, the parameters of number of hidden layer and number of neurons have high effect on its performance. These parameters must be selected by accuracy. Thus this paper proposes the application of imperialist competitive algorithm (ICA) for finding the optimum value of these parameters. Simulation results show that the proposed intelligent method has very good performance and accuracy.

Keywords:

ICA , MLP , Accuracy , Classifier , Detection