基于小波神经网络的二次再热汽温建模

Modeling of Secondary Reheat Steam Temperature Based on Wavelet Neural Network

二次再热机组再热汽温模型精度对其稳定控制有着重要的影响。由于二次再热系统多变量、强耦合、大迟延的特性,传统机理建模或数学模型辨识的效果并不理想。本文提出了一种小波神经网络、动量梯度法、粒子群和免疫算法相结合的二次再热机组再热汽温建模方法,在免疫粒子群开始时使用动量梯度法优化初始种群,在免疫粒子群算法收敛缓慢时使用动量梯度法进行局部细调。与其他几种网络权值优化算法进行仿真对比表现良好,改善了小波神经网络权值系数易陷入局部最优的缺陷,提高了网络性能,最后用于二次再热汽温建模中。

Due to the multivariable,strong coupling and large delay characteristics of the secondary reheat system,the traditional mechanism modeling or mathematical model identification is not ideal.In this paper,a reheating steam temperature modeling method based on wavelet neural network,momentum gradient method,particle swarm optimization and immune algorithm is proposed.The simulation performance is good,which improves the defect of the wavelet neural network weight coefficient which is easy to fall into local optimum,improves the network performance,and finally is used in the secondary reheat steam temperature modeling.