神经网络广义预测控制在锅炉燃烧系统中的应用

针对锅炉燃烧系统的非线性、大延迟、时变、干扰频繁等特点,以煤粉浓度为中间被调量,将神经网络、广义预测控制、串级控制相结合,设计了基于神经网络模型的广义预测串级控制系统.该控制方法克服了单纯PID控制对大惯性大延迟对象调节品质差、抗干扰性弱的缺点,神经网络预测器有效地补偿了传统预测控制基于线性模型的局限性.将该控制算法用于燃烧系统中主汽压力对象的控制,仿真结果表明该方法具有较强的跟踪性能和抗干扰能力及良好的动静态性能指标.

Application of Neural Network Generalized Predictive Control in Combustion System of Boiler

In accordance with the features of combustion system of boiler, e.g. nonlinearity, large time delay, time variation and frequent interference, the generalized predictive cascade control system based on neural network model is designed by combining neural network, generalized predictive control and cascade control. In this system, the concentration of powdered coal is an intermediate controlled variable. The disadvantages of pure PID for objects with large time delay and large inertia, i.e. poor control quality, weak capability of anti-interference are overcome by this method. The limitation of traditional prediction control based on linear model is compensated effectively by neural network predictor. The control algorithm is utilized to control the main steam pressure in combustion system, the simulation results show that this method features good tracking performance, anti-interference capability as well as dynamic and static performance indexes.