针对传统的解耦方法对实际工业生产过程中的多变量、非线性、强耦合系统解耦效果不理想的问题,提出了改进的简化粒子群(isPSO)算法与PID神经网络(PIDNN)相结合的方法。PIDNN训练用于消除回路间的耦合,其连接权值由简化粒子群算法学习优化。该isPSO算法能克服PIDNN易陷入局部收敛的缺点,而且与基本PSO算法相比,搜索到最优值的概率更高。采用的小步长线性递减惯性权重和增加的极值扰动算子,则加速了对PIDNN权值的优化。通过对强耦合对象火电厂锅炉燃烧系统的仿真表明,该方法具有更好的控制品质:鲁棒性强、跟踪快、解耦效果好,为实际应用中强耦合系统的改进提供了理论依据。 Aiming at the problem that the traditional decoupling method is not ideal for the decoupling effect of multivariable, nonlinear and strongly coupled systems in real industrial processes, an improved simplified PSO algorithm is proposed in combination with PID neural network (PIDNN) Methods. PIDNN training is used to eliminate the coupling between loops, and its connection weights are optimized by learning from the simplified PSO algorithm. The isPSO algorithm can overcome the shortcoming that PIDNN is easy to fall into local convergence, and the probability of finding the optimal value is higher than that of the basic PSO algorithm. The small step size used linearly decreasing inertia weight and increasing extremum perturbation operator accelerate the optimization of PIDNN weights. Simulation results show that this method has better control quality. It has strong robustness, fast tracking and decoupling effect, which provides a theoretical basis for the improvement of strong coupling system in practical applications.