针对现有非线性系统辨识超调较大和预测控制计算量繁琐等问题,提出了改进的RBF神经网络线性预测控制算法.该方法通过在传统性能指标函数中增加误差微分项,以优化跟踪效果;利用辨识模型作为预测模型,对输出设定值进行线性逼近的反向优化,并实时给出优化控制量.该方法简化了传统预测控制算法,在加快寻优速度的同时,有效地抑制了超调.通过非线性系统仿真实例,验证了该方法的可行性和有效性.%Aiming at the problems of large overshoot in existing nonlinear system recognition and complicated calculation of predictive control, the improved RBF neural network linear predictive control algorithm is proposed. In this method, error differential term is added in traditional performance index function to optimize the tracking effect; with the recognition model as the predictive model, the reverse optimization algorithm that linear approaching to the output setting point is adopted, and the optimized control value is given in real time. The method simplifies the conventional predictive control method, it efficiently eliminates the overshoot while accelerating the optimization speed. The simulation of nonlinear system shows that the system is feasible and efficient.
展开▼