基于GA-BP的电站锅炉NOx排放特性建模研究

NOx排放模型是电站锅炉实时控制系统的基础。针对普通BP神经网络建模方法收敛速度慢和易陷于局部极值点的问题,提出基于遗传算法（GA）优化BP神经网络的建模方法。通过电站锅炉热态试验获取样本数据,对BP网络隐节点数进行优化后,建立了GA-BP模型。相比BP神经网络模型,该模型训练时间短,拟合误差大大降低。仿真试验表明：GA-BP模型性能得到改善,泛化能力明显提高,能准确预测NOx排放。GA-BP模型可为运行人员提供指导,也可作为电站锅炉实时控制系统的基础模型。

A study of GA-BP-based modeling of NOx emission characteristics of utility boilers

The NOx emission model is the foundation of the real-time control system for utility boilers.The traditional back propagation（BP）network algorithm has the limitations of poor convergence and is easy to fall in local optima.In allusion to the defects,a modeling method which uses the Genetic Algorithm（GA）to improve BP neural network is proposed.The GA-BP model is established after the Network of Hidden Layer Nodes were optimized through collecting sample data from the utility boiler in running state.Compared with the model of BP neural network,the training time of GA-BP is shortened.Meanwhile,the average relative error is reduced greatly.Simulation results show that performance and generalization abilities of the GA-BP model are improved obviously,and the model can accurately predict NOx emissions.GA-BP model can not only provide guidance for operators,but also be serviced as a basic model of utility boiler real-time control system.