Applying neural networks in adaptive distance protection

Classic distance relaying measures the apparent impedance of the transmission line. Fault resistance and the remote source infeed to the fault branch significantly modifies the apparent impedance, which causes a possible overreach or underreach error. The error is influenced by the actual power system conditions so that it is difficult to select a universal operating characteristic. In this paper we show how neural networks can be designed to achieve adaptive distance relaying for the above problem. A realistic, complicated 500 kV system is considered. The research is concentrated on designing, training and testing of the neural distance relays for line-to-ground faults