基于先验知识和神经网络的非线性建模与预测控制

Nonlinear Modeling and Predictive Control Based on Prior Knowledge and Neural Networks

神经网络模型是模拟非线性系统的有力工具,它的缺陷是难以利用已有的先验知识。利用通用学习网络的建模方法,提出了一种利用先验知识和神经网络建立非线性系统模型的方法,具有简化神经网络结构、减小计算量的优点。基于这种模型利用改进的遗传算法进行优化计算,从而实现了基于先验知识和神经网络的非线性建模和预测控制。对一个悬吊系统的仿真实验说明了该算法的有效性。

Neural network model is a powerful tool in modeling nonlinear system, and its shortcoming is that it can not utilize prior knowledge. Using universal learning networks, this article has proposed a new method to model nonlinear system, in which prior knowledge is combined with neural networks. This method has the advantage of simplifying the network construction and reducing the computation load. Based on this model, we use improved genetic algorithm (GA) to implement the optimization, and then provide a nonlinear predictive control algorithm. A simulation experiment dealing with a crane system demonstrates the efficiency of the algorithm.