粒子群优化算法在神经网络控制中的应用

考虑粒子群优化算法在不确定系统的自适应控制中的应用。神经网络在不确定系统的自适应控制中起着重要作用。但传统的梯度下降法训练神经网络时收敛速度慢,容易陷入局部极小,且对网络的初始权值等参数极为敏感。为了克服这些缺点,提出了一种基于粒子群算法优化的RBF神经网络整定PID的控制策略。首先,根据粒子群算法的基本原理提出了优化得到RBF神经网络输出权、节点中心和节点基宽参数的初值的算法。其次,再利用梯度下降法对控制器参数进一步调节。将传统的神经网络控制与基于粒子群优化的神经网络控制进行了对比,结果表明,后者有更好逼近精度。以PID控制器参数整定为例,对一类非线性控制系统进行了仿真。仿真结果表明基于粒子群优化的神经网络控制具有较强的鲁棒性和自适应能力。

Applications of Particle Swarm Optimization to Neural Network Control

This dissertation studies the problem of applications of particle swarm optimization in adaptive control of uncertain systems.Neural networks play an important role in adaptive control of uncertain systems.However,it has several shortcomings,such as slow convergence speed,local minimum as well as desperate sensitivity of the initial parameters when using the traditional gradient algorithm to train the neural networks.In order to overcome these shortcomings,proposed a method that regulates the weights of the neural networks by using particle swarm optimization algorithm.First,this paper present an algorithm to optimize and get the initial parameters of the neural networks based on fundamental principle of the particle swarm optimization,then uses the neural networks to regulate the parameters of the controller by gradient algorithm.A comparison between traditional neural network control and based-particle-swarm-optimization neural network control is drawn. The comparison shows that the latter can approximate uncertain nonlinear function with higher accuracy.