A study of power network stabilization using an artificial neural network

In power network, low frequency oscillation become a major concern for many years. In order to depress low frequency oscillation, the power system stabilizer parameters must be adjusted when there are changes in power network conditions. This paper presents the application of neural network to tune the power network stabilizer parameters. For training neural network, generator real power and reactive power are chosen as the input signals and the output are the desired power network stabilizer parameters. A popular type of neural network, the multi-layer perceptron with error-back-propagation training method, is employed. The neural network, once trained, can yield proper power network stabilizer parameters under any generator loading conditions. Simulation results show that the neural network based power system stabilizer yield better dynamic performance than conventional power system stabilizers in the sense of having large damping in responds to a step disturbance.