基于ISSA-DELM算法的CSTR系统广义预测控制研究

CSTR system′generalized predictive control based on ISSA-DELM algorithm

连续搅拌反应釜(CSTR)作为典型的聚合反应化工生产用到的设备,其在工作运行时具有强非线性、大滞后性和不确定性,用传统的方法难以建立精准的数学模型。文中根据一类CSTR反应过程采用Hammerstein-Wiener模型,使用高斯径向基函数的LS-SVM分别对模型的两个非线性模块进行建模,并使用其建立的Hammerstein-Wiener模型作为广义预测控制的预测模型;针对广义预测控制的滚动优化环节,采用多策略改进的麻雀算法(ISSA)优化深度极限学习机(DELM)的混和优化算法策略,并利用基准函数测试改进麻雀算法的优越性;最后将混合优化算法应用在非线性CSTR对象上,经过实验证明,所提出的ISSA-DELM混合优化算法对CSTR系统具有较好的控制效果,并与未改进的SSA-DELM算法和DELM算法进行仿真结果对比,结果显示,文中算法控制效果明显优于SSA-DELM算法和传统的DELM算法。

The continuously stirred tank reactor(CSTR),as a typical equipment used in chemical production of polymerization reaction,has strong nonlinearity,large lag and uncertainty during operation,so it is difficult to establish an accurate mathematical model with the traditional methods.In this paper,the Hammerstein-Wiener model is adopted according to a class of CSTR reaction processes,and the two nonlinear modules of the model are modeled by Gaussian radial basis function LSSVM,and the established Hammerstein-Wiener model is used as the prediction model for generalized predictive control(GPC).For the rolling optimization of GPC,a hybrid optimization algorithm strategy based on deep extreme learning machine(DELM)which is optimized by multi-strategy improved sparrow search algorithm(ISSA)is adopted,and the superiority of the ISSA is tested by reference function.The hybrid optimization algorithm is applied to nonlinear CSTR objects.The experimental results show that the proposed hybrid optimization algorithm ISSA-DELM has a good control effect on CSTR system.The simulation results of the SSA-DELM algorithm and the DELM algorithm show that the control effect of the proposed algorithm is significantly better than those of the SSA-DELM algorithm and the traditional DELM algorithm.