Optimum decision by artificial neural networks for reactive power control equipment to enhance power system stability and security performance

The purpose of this study is to explore how to utilize the artificial neural networks (ANNs) technique to optimize the number, sizes, and locations of reactive power control equipment in order to increase the power system global steady-state stability and security performance. In this paper, two widely used ANNs - multilayer perceptron neural network (MLP) and self-organizing feature map (SOFM) neural network are discussed and trained to evaluate different shunt capacitor banks (SCBs) installation plans. Based on the system global steady-state stability and security achievements, and the sizes and locations of the SCB installation plans, four acceptance levels are defined by the experts as the output variables of the ANNs: best, good, average and poor. The optimum plans are among those with best acceptance level. The optimum plan means that considering the economy of the MVAR sizes and the availability for the sites of SCBs, the installed SCBs will maximally improve the system global steady-state stability and security performance. In this research, the proposed ANNs are trained and tested under a 39-bus power system.