Application of artificial neural networks for series compensated line protection

This paper investigates the feedforward neural network (with a quasi-Newton method for minimization of the error function) for on-line calculation of the voltage across a nonlinear capacitor installation. This technique is incorporated into a new scheme which is independent of the series capacitor installation, operation of the capacitor protection, and the surrounding power system elements. The proposed scheme is simple and accurate and requires only local voltage and current at the bus to improve the reach measurement of distance relays and fault locators. Detailed testing using EMTP has been done to show the benefits, robustness and generalization capabilities of the ANN technique.