On ADHDP-based optimization control for STATCOM

Aimed at the contradiction between the rapidity and the stability, the problems of the dependence upon precise mathematical model and the poor adaptability and robustness in the traditional PI control for STATCOM, this paper introduces a novel nonlinear optimal neuron-controller which is based on adaptive critic design and uses the structure of action-dependant heuristic dynamic programming ( ADHDP). The controller is composed of an action network and a critic network, which are built up by BP neural networks. By using the system's voltage change and the reactive power to train its action network and critic network alternately, the STATCOM controller can be optimized. The ADHDP neuron-controller has a simple framework and is independent from the mathematical model and can study online too. A simulation example of a single machine infinite bus system is carried out in Matlab/ Simulink, and the control results of traditional PI control are compared too. The simulation results show that the ADHDP neuron-controller gets better in overshoot, speed of response, steady state error, adaptability and robustness.