采用炉膛辐射信号的锅炉燃烧系统模糊神经网络预测控制的仿真研究

针对火电厂燃烧过程中主蒸汽压力控制系统的大时滞、大惯性和非线性 ,采用以炉膛辐射信号为中间被调量的串级控制系统 ,并设计一个基于神经网络预测模型的模糊神经网络控制器作为主控制器。该控制器首先将神经网络与预测控制相结合 ,采用递阶遗传算法对神经网络的结构和权值分别进行训练 ,以实现非线性、大时滞系统模型的精确预测 ;然后将模糊控制与神经网络相结合 ,实现模糊神经网络预测控制。考虑到炉内剧烈的湍流燃烧造成炉膛辐射信号包含随机分量 ,又设计了一个附加判断器的二自由度 PID控制器作为副控制器。仿真结果表明 ,该方案显著提高了非线性、大时滞燃烧系统的控制品质 ,且易于工程实现

Simulation of Fuzzy Neural Network Forecast Control Based on Radiation Signal of Furnace

Considering that main steam pressure of burning system is a big-lagged, big-inertial and nonlinear object, a cascade control system is designed, which uses the radiation signal as intermediate regulated variable. A fuzzy neural network controller is applied to outer loop, the controller, first, combines forecast control with neural network optimized by a layer hierarchical genetic algorithm and realizes accurate forecast to the nonlinear and big-lagged system; then, combines fuzzy control with neural network and realizes forecast control to the system. Considering that the radiation signal of the furnace contains random variable, a double-degree PID controller with judger is applied to inner loop. Simulation result shows the project is effective to control nonlinear and big-lagged system and is easy to be applied.