Particle Swarm Optimization for Adaptive-Critic Feedback Control with Power System Applications

Considering the complete dependence of traditional adaptive dynamic programming (ADP) algorithms on gradient information and the lack of theoretical guarantee of particle swarm optimization (PSO), we develop an evolution-explored ADP algorithm based on PSO to realize optimal regulation for discrete-time nonlinear systems. This algorithm combines the value iteration method in ADP with PSO for policy improvement to seek out the optimal control policy, which enhances the algorithm applicability while ensuring the control performance of the system. Compared with the method using only PSO, it can speed up the search of particles for the optimal value and reduce iteration errors. Finally, the advantages and control effects of the proposed algorithm are verified through comparative experimental simulations on power systems.