A real-time hardware fault detector using an artificial neural network for distance protection

A real-time fault detector for the distance protection application, based on artificial neural networks, is described. Previous researchers in this field report use of complex filters and artificial neural networks with large structure or long training times. An optimum neural network structure with a short training time is presented. Hardware implementation of the neural network is addressed with a view to improve the performance in terms of speed of operation. By having a smaller network structure the hardware complexity of implementation reduces considerably. Two preprocessors are described for the distance protection application which enhance the training performance of the artificial neural network many fold. The preprocessors also enable real-time functioning of the artificial neural network for the distance protection application. Design of an object oriented software simulator, which was developed to identify the hardware complexity of implementation, and the results of the analysis are discussed. The hardware implementation aspects of the preprocessors and of the neural network are briefly discussed