Particle Swarm Optimization Based Model Predictive Control with Constraints

摘要

Model Predictive Control (MPC) algorithms can control a large scale of systems with many control variables, and. most importantly, can handle constraints on inputs and states systematically. In MPC. these constraints are accounted for explicitly by solving a constrained optimization problem in real-time to determine the optimal predicted inputs. However, when solve these constrained optimization problems by Quadratic Programming (QP) algorithm, the predicted and actual responses may differ and can not get correct results. For this problem, a novel Chaotic Particle Swarm Optimization (CPSO) method is introduced and applied to MPC, solving the control problems with constraints on inputs and states systematically. PSO is newly developed evolutionary technique which has gained much attention and wide applications in different fields. However, the standard PSO greatly depends on its parameters and exists as the premature phenomenon, especially in solving complex multi-hump problems. Chaos is a kind of characteristic of nonlinear systems. Due to the unique ergodicity and special ability to avoid being trapped in local optima, here, chaotic dynamics is incorporated into the PSO and a more advanced optimization method. CPSO is generated and then is applied to MPC. Finally, two constrained optimization problems on discrete-time linear systems arc introduced and solved by QP and PSO respectively. By comparing the simulation results, the advantages of PSO based MPC algorithm arc fully illustrated.