基于改进 BP 的神经网络模型参考自适应控制

摘要

Due to the traditional BP algorithm’s some defects like slow convergence speed, easily falling into local minimum, the process is made more complex by the improved BP algorithm. Thus model reference adaptive control based on BP neural network also has slow convergence, poor real-time and low accuracy. In this paper, combined with the improved BP algorithm and the reversibility of nonlinear system, the author puts forward a kind of model reference adaptive control based on BP network that transfer function can optimized by itself. And the Matlab simulation result shows that, in the case of meeting the control precision, the control effect of the identifier and the controller are very ideal. Therefore, this method is valuable for practical application of engineering.%由于传统 BP 算法存在收敛速度慢，容易陷入局部极小值等弊端，目前的 BP 优化算法又使得控制过程变得复杂，继而基于 BP 神经网络的模型参考自适应控制过程也存在实时性差，收敛性慢，精度不高等不足。现针对改进的 BP 算法和非线性系统的可逆性，分析设计了一种基于激励函数自寻优的 BP 网络模型参考自适应控制，并通过 Matlab仿真结果表明，在满足控制精度的情况下控制系统中的辨识器和控制器效果都很理想。因此，对工程应用有很大的实际参考利用价值。