论文部分内容阅读
针对非线性预测控制中,系统建模和目标函数求解的问题,提出了一种基于粒子群优化的非线性预测控制策略(PSO-NPC)。首先,将时间因素引入到即时学习算法中,提高了基于即时算法的最小二乘支持向量机(LS-SVM)对非线性系统的建模精度。其次,针对单目标优化的常规PSO-NPC算法不足之处,将系统的第一步预测和最后一步预测输出作为主要优化目标,提出了多目标粒子群优化的非线性预测算法。最后,将目标函数中的误差权重作为粒子群优化的目标,根据系统耦合程度自适应调整误差权重,消除了系统回路之间耦合。仿真结果验证了改进算法的可行性和有效性。
Aiming at the problem of system modeling and solving objective function in nonlinear predictive control, a particle swarm optimization based non-linear predictive control strategy (PSO-NPC) is proposed. Firstly, the time factor is introduced into the instant learning algorithm, which improves the modeling precision of LS-SVM based on real-time algorithm for nonlinear systems. Secondly, aiming at the shortcomings of conventional PSO-NPC algorithm for single-objective optimization, the first-order prediction and the last-step prediction output of the system are taken as the main optimization objectives. A nonlinear prediction algorithm based on multi-objective particle swarm optimization is proposed. Finally, the error weight in the objective function is taken as the target of particle swarm optimization, and the error weight is adaptively adjusted according to the degree of system coupling, thus eliminating the coupling between system loops. Simulation results verify the feasibility and effectiveness of the improved algorithm.