基于多神经元模型的非线性系统预测控制

Multiple Neuron Models Predictive Control for Nonlinear Systems

利用单神经元来逼近非线性系统在平衡点邻域内的泰勒展开式的直至二次项,首次提出了一种用多个单神经元模型来拟合非线性系统的建模方法,引入多模型参考轨迹,得到一种新的多模型预测控制。仿真结果表明,基于二阶泰勒级数得到的多神经元模型的预测控制器的性能要优于采用泰勒级数一阶线性项得到的多模型预测控制器,但计算量并未显著增加。

One neuron is trained to approximate the Taylor expansion eouation to the second order at each equilibrium point. And a new modelling method is proposed in which multiple neuron models are to approximate a nonlinear system. A new multi-model predictive control algorithm is presented by given multiple reference trajectories. Many test results show that the method is feasible, and it is characterized by its simple structure, small calculating amount and fast speed. Performance comparison by simulation demonstrates that the predictive controller based on the second-order Taylor nonlinear expansion equation performs better than that based on the one-order linear part, and calculation is only slightly increased.