摘要
提出一种基于神经网络和参数优化的预测控制方法。首先利用带有动量项的改进BP神经网络辨识系统模型,在辨识过程中使用粒子群算法(PSO)对改进BP网络的初始权值/偏置、学习率、动量系数等辨识参数进行学习优化,解决这些参数的取值问题;然后将辨识得到的模型用于隐式广义预测自校正控制中,使用遗传算法(GA)对控制过程进行优化,寻找最优的控制参数(预测时域、控制时域、控制加权系数、柔化系数)。将该方法应用在热工系统中,仿真结果表明了方法的有效性。
A kind of predictive control based on neural network (NN) and parameter optimization is put forwend in this paper. Firstly, system model is identified by using improved BP NN with momentum coefficient. Particle swarm optimization (PSO) algorithm is used to optimize the parameters of NN weighting initial value and offset initial value, speed of learning and momentum coefficient in identification process of improved BP NN, solving the problem that it' s difficult to determine the value of them. Subsequently, NN model optimized by PSO is used in implicit self - correcting generalized predictive control process. Genetic algorithm (GA) is used to optimize predictive control process and to find the optimal control parameters (predictive length, control length, control weight number and soft coefficient). The application in thermal process shows the method is effective.
出处
《华北电力大学学报(自然科学版)》
CAS
北大核心
2009年第5期66-72,共7页
Journal of North China Electric Power University:Natural Science Edition
关键词
改进BP网络
隐式广义预测自校正控制
粒子群算法
遗传算法
参数优化
improved BP neural network
implicit self correcting generalized predictive control
particle swarm optimization algorithm
genetic algorithm
parameter optimization