REAL-TIME TRANSIENT STABILITY STATUS PREDICTION SCHEME AND COMPARATIVE ANALYSIS OF THE PERFORMANCE OF SVM, MLPNN AND RBFNN

Abstract

This paper presents a simple and effective technique for real-time prediction of transient stability status following a large disturbance and compares the performance of three artificial intelligence (AI) techniques commonly employed as decision making tools. The (AI) techniques compared are support vector machine (SVM), multilayer perceptron neural network (MLPNN), and radial basis function neural network (RBFNN). The stability status prediction scheme samples rotor angles of all system generators and extracts the absolute value of the first sampled rotor angle value of each generator. The extracted absolute rotor angle value of all generators are summed and fed as input to a decision tool. The scheme was tested using simulations carried out on the IEEE 39-bus test system. One hundred percent prediction accuracy was obtained when SVM and MLPNN were each employed as decision tools. The use of the RBFNN as decision tool resulted in only sixty-three percent prediction accuracy.

Keywords

• neural networks
• out-of-step
• support vector machine
• transient stability