A fault detection technique with preconditioned ANN in power systems

This paper presents a hybrid method of a data precondition technique and an artificial neural network (ANN) for fault detection to estimate the fault location and the type in the transmission systems. FFT is used as a data precondition technique to extract the features of input variables. Also, the radial basis function network (RBFN) is employed to approximate a nonlinear relationship between input and output variables as ANN. To enhance the model accuracy, this paper proposes a new RBFN called D-RBFN that makes use of DA clustering in determining the center vector and the width of the radial basis function. The D-RBFN has a global structure obtained by global clustering. The proposed method is successfully applied to a sample system.