Average Voltage and Multilayer Perceptron Neural Network Based Scheme to Predict Transient Stability Status

Abstract

This paper presents a technique that predicts the transient stability status of a power system after a disturbance. It uses generator bus voltage as input parameter and a trained single-input multilayer perceptron neural network (MLPNN) as decision tool. When activated, the scheme samples voltages of all generator buses. Two sets of voltage values are extracted from each sampled generator bus voltage. For each set, the minimum voltage value is obtained. An average value is computed from the minimum voltage values extracted from the first sample sets of the various generator buses. The average value is then used to compute the deviations of the minimum voltage values from the second sets of data. The deviations are then summed and used as input to a trained MLPNN which indicates the stability status. The technique was tested using the IEEE 39-bus test system and its accuracy found to be 98.97%.