Hybrid approach using counterpropagation neural network for power-system network reduction

Abstract

A hybrid counterpropagation neural network and Ward-type equivalent approach for power system network reduction is proposed for improving the conventional external system equivalent technique. The proposed Ward-type equivalent technique not only possesses the good properties of the extended Ward equivalent, but can also update the parameters of the equivalent model for representing real-time topology changes of the external system. Another improvement is that a counterpropagation neural network is used to match the boundary equivalent power injections. The new hybrid approach combines the simplicity of Ward-type equivalent techniques with the speed of artificial neural networks. Test results demonstrate that the hybrid approach is very efficient and highly accurate compared to the external system equivalent.