基于粒子群优化的有约束模型预测控制器

Particle-swarm optimization algorithm for model predictive control with constraints

研究了模型预测控制(MPC)中解决带约束的优化问题时所用到的优化算法,针对传统的二次规划(QP)方法的不足,引入了一种带有混沌初始化的粒子群优化算法(CPSO),将其应用到模型预测控制中,用于解决同时带有输入约束和状态约束的控制问题.最后,引入了一个实际的带有约束的线性离散系统的优化控制问题,分别用二次规划和粒子群优化两种算法去解决,通过仿真结果的比较,说明了基于粒子群优化(PSO)的模型预测控制算法的优越性.

We investigate the optimization algorithms for solving the constrained optimization problems in model predictive control（MPC）. To deal with the disadvantage of the quadratic programming（QP） algorithm, we introduce and apply the chaotic particle-swarm optimization（CPSO） algorithm to solve the control problem with simultaneous constraints on inputs and states. A practical constrained optimization problem of the discrete-time linear system is solved by QP and PSO, respectively. By comparing the simulation results, we show the advantages of the PSO-based MPC algorithm.