小波神经网络在过程控制中的应用研究

非线性过程的建模和控制问题尚无通用的模型结构和方法可用于过程控制.结合小波和神经网络方法进行过程控制的通用模型和方法的研究,提出了一种仅用尺度函数逼近的小波神经网络模型,并用它来实现非线性预测控制方案.该模型能用线性最小二乘方法进行拟合,因而具有易于实现和通用的优点.由于简化了在线优化方法,所提出的非线性预测控制方案已在线实现.用该方法进行了两个非线性系统的模型辨识和一个双线性系统的控制仿真,模型的通用性、辨识和控制方法的简单易用和仿真结果表明,所提出的方案对过程工业和非线性、高阶系统有很好的实用性,有着比标准PID控制器好得多的控制效果.

Application research of wavelet neural networks in process control

The modeling and control problems of nonlinear processes have always been difficult problems. There were not any general models or methodologies to be used in process control. This paper is devoted to the research of practical general models and methods of process control by combining wavelets and neural networks. It presents a wavelet neural netowrk model approximated only by scaling function. The model is adopted to implement nonlinear model predictive control strategy with nonlinear programming. The model can be identified by linear least square (LS) learning algorithm, therefore it is easy to be realized as a general model. Since a simplified on-line optimization method is developed, the proposed control strategy is very easy to implement. Using the proposed identification and control strategy, the identification results of two kinds of systems and the simulation of a bilinear process control system are presented. The generality of the model and the simplification of identification and control method together with the simulation results show that the above strategy has strong possibility to nonlinear system and high order system identification for process industries and the excellent control performances superior to a standard PID controller for the same process.