电站锅炉神经网络建模及性能优化

Neural Network Modeling and Performance Optimization of Power Plant Boilers

锅炉系统是一种非线性的多变量耦合系统,具有一定的滞后性,导致无法获得精确的数学模型。目前,燃煤电站的锅炉控制系统主要为DCS,该系统虽然对单输入、单输出系统具有较高的控制精度,但是由于锅炉系统的复杂性,系统整体控制很难达到最优。利用神经网络对锅炉系统建立黑箱模型,通过遗传算法对锅炉模型参数进行辨识,用辨识获取的参数指导运行人员对锅炉进行调节。最后在一电厂320 MW循环流化床锅炉上进行了试验,结果显示,所提出的方法能有效地降低供电煤耗。

Boiler system is a non-linear multi-variable coupling system, which has a certain lag, so it is impossible to obtain an accurate mathematical model. At present, the boiler control system of coal-fired power plant is mainly DCS. Although DCS system has higher control precision for SISO system, the complex characteristics of boiler system make it difficult to achieve the optimal overall control of the system. The black box model of boiler system was established by using neural network, the parameters of boiler model were identified by genetic algorithm, and the parameters obtained by identification were used to guide operators to adjust the boiler. Experiments on a 320 MW circulating fluidized bed boiler in a power plant showed that the proposed method could effectively reduce the coal consumption of power supply.