基于PSO算法的多变量PID型神经元网络在球磨机控制上应用

Application of PSO-based multivariable-PID neuron network in ball mill control

为提高多变量、非线性和强耦合系统的动态特性和解耦能力,基于PID控制的简单结构和良好性能优势以及神经元网络的自调节和自适应的特长,设计了具有PID结构的多变量自适应的PID型神经元网络控制器。给出了这种控制系统的结构和算式,其为一种3层前向神经网络,其隐层单元分别为比例(P)、积分(I)和微分单元(D),各层神经元个数、连接方式、连接权值的初值是按PID控制规律确定的。神经元网络参数采用了粒子群优化(PSO)学习算法,并给出了相关算式。分析了球磨机制粉控制系统的特点,并应用提出的控制方法对其进行了仿真研究,比较了文中控制方法与传统的PID控制方法的控制效果。仿真结果表明了所提方法具有较好的控制品质、良好的自适应解耦能力和自学习功能。

To improve the dynamic property and decoupling capability of muhivariable nonlinear coupled system,a multivariable adaptive neuron network controller with PID structure is introduced, which has both the simple construction and good property of PID controller and the self-regulation and adaptability of neuron networks. It is a 3- layer feed forward neural network, with proportional neuron （P） ,integral neuron（I） and derivative neuron（D） in its hidden layer. The neuron amounts at each layer,the connection modes and the initial weights are determined by the rules of PID control. PSO（Particle Swarm Optimization） algorithm is applied in neuron network parameter learning and its corre- lative formulas are given. Characteristics of ball mill pulverization control system are analyzed,and both the proposed controllon and conventional PID controller are applied in simulative research of control performance. Simulation results prove that the controller proposed has better control quality, adaptive decoupling ability and self-learning function.