【摘要】 本文研究了基于径向基函数神经网络(RBFNN)和多目标优化算法的非线性模型预测控制。RBFNN神经网络在每个控制间隔预测被控变量的实时值,多目标模型预测控制以多目标方式考虑每个局部模型的目标函数,同时将所选目标函数作为主要考虑因素,其他目标函数被认为是主要的附加约束。随着附加约束的值在最小和最大之间变化,可以通过选择非支配结果来获帕累托最优选择。并将提出的算法和单模型单目标函数非线性模型预测控制方法在CSTR对象上进行了比较。仿真结果表明该算法的有效性和可行性。更多还原

【Abstract】 The multiple objective model predictive control of nonlinear plant based on Radial basis function neural networks(RBFNN)and multiple objective optimization algorithm is considered.The RBFNNs predict the plant outputs at every control interval.Multiple objective model predictive control considers the objective functions of each local model simultaneously in a multiple objective way.Putting selected objective function as the main one,other objective functions are considered as additional constraints of the main.With the values of additional constraints varying between the minimum and maximum,the Pareto surface can be obtained by selecting the non-dominated results.And a non-dominated sorting method is applied to solve the computing problem of control moves.Comparison between the proposed algorithm and single model single objective function nonlinear model predictive control method is investigated on a simulated CSTR.The simulation results show the efficiency and feasibility of the algorithm.更多还原