摘要
提出了一种基于径向基函数的非线性过程预测控制策略。首先,开发过程径基函数网络模型:根据过程特性,选择模型阶次、径基函数类型,用K均值法确定基函数中心位置,统计F检验确定基函数中心的数目,迭代最小二乘法确定径基函数网络权系数。然后,利用网络模型抽取非线性预测控制器(NLPC)特征样本训练构造径基函数网络预测控制器(RBFPC)。仿真结果表明,与NLPC比较,由于RBFPC不必在线解非线性最优化问题,易于在线快速实施;与PI控制器比较,RBFPC具有更好的跟踪设定值性能和抗干扰性能。
A predictive control strategy for nonlinear processes based on radial basis function (RBF) models is proposed in this paper. First, an RBF network model of nonlinear processes is developed according to the following steps: (1)The model orders and the type of RBF are determined based on the characters of the processes. (2)The positions and the numbers of the RBF centers are fixed by the K-means method and F-statistical test respectively. (3) The weighting coefficients of RBF network are estimated with stepwise least squares algorithm. Then, using the developed RBF process model, the characteristic samples of the nonlinear predictive controlled NLPC) are extracted by solving an optimization problem off-line and an RBF predictive controller is constructed. The simulation shows that since no optimization problems have to be solved on-line, this RBFPC can be implemented more easily than the NLPC can. And it provides more excellent setpoint tracking and disturbance rejection performances when compared with conventional PI controller.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2000年第4期361-365,共5页
Pattern Recognition and Artificial Intelligence
基金
高等学校骨干教师资助计划资助项目
关键词
非线性过程
预测控制
径向基函数
神经网络
Nonlinear Process, Predictive Control, Neural Network Model, Radial Basis Function