Research On The Application Of Adaptive Dynamic Programming In Power System

Power system is a nonlinear time-variant and high-dimension system. How can weguarantee the security and stability of its running by efficiently control, which is the issuemany scholar researching. With the increasing of power system’s scale and the random newenergy source bringing in, the challenge of power system’s security and stability becomesmore severely.Traditional optimal control methods have their limitations. Calculus of variations canonly solve problems with no-limitation on control variable. Maximum principle can onlysolve the problems described by ordinary differential equation. Although dynamicalprogramming can solve more general problems, it has the curse of dimensionality.Adaptive dynamic programming is the production of artificial intelligent and controltechnology, its essence is using neural networks to approximate the solution ofHamilton-Jacobi-Bellman. ADP may not need the model of the controlled objection, alsomay not need the accurate cost-to-go function, and can learn on-time, so bringing it intopower system may afford a new solution of nonlinear optimal control.Firstly, the thesis introduce the development of ADP, summarize the current researchsituation, then introduce the classification、advantages、the selection method of parametersand the method of normalization.Secondly, the thesis deduce the state equation of a single machine infinite bus system,introduce the principle of ADHDP and PSO, apply ADHDP to control exciter, and comparethe results to a PID controller which is optimized by PSO.Lastly, the thesis introduce the model and control strategies of STATCOM, alsointroduce the principle of GrHDP, then apply GrHDP to control a STATCOM, and comparethe results to a PID controller.