基于支持向量机和BP神经网络的燃煤锅炉NO_x排放预测  

Modeling of NO_x Emission from Coal Fired Boiler based on Intelligent Algorithm

作　　者：李鹏辉 刘冉[2] 余廷芳[2] 

机构地区：[1]华电电力科学研究院,浙江杭州310030 [2]南昌大学热能与动力工程研究所,江西南昌330031

出　　处：《热能动力工程》2016年第10期104-108,129-130,共5页Journal of Engineering for Thermal Energy and Power

基　　金：国家自然科学基金燃煤电站锅炉经济运行与污染排放多目标优化研究(61262048)

摘　　要：基于某660 MW燃煤锅炉运行时的热态实验数据,应用BP神经网络方法和支持向量机回归的方法对该燃煤电站锅炉NO_x排放特性分别进行建模,针对BP神经网络存在的问题,采用动量法对其进行改进,而对SVM预测模型进行了核函数及相应参数c和g进行了选优。两种模型仿真结果的平均相对误差为2.75%和1.37%,证明模型的准确性和泛化能力比较好。引入神经网络模型评价指标,对这两种模型的仿真和预测结果进行对比分析,结果表明采用支持向量机方法建立的NO_x排放模型比BP神经网络模型收敛速度快,准确度高,性能更优。Based on the experimental data of the thermal state in the 500 MW ~ 600 MW load range of a coal fired boiler,BP neural network and support vector machine regression were used to model the NOxemission characteristics of a coal-fired power station. To address the problems of BP neural network,the momentum method was adopted,but for the prediction model of SVM,the kernel function and the corresponding parameters c and g were selected through optimization. The average relative errors of the simulation results of the two models were 2. 75% and 1. 37%,respectively,indicating a reasonable accuracy and generalization ability of the model. With the evaluation index of neural network model,the simulation and prediction results of the two models were compared and analyzed. The results show that the NOxemission model established by the support vector machine method has faster convergence speed,higher accuracy and better performance than the one based on BP neural network model.

关 键 词：锅炉 NOX排放 BP神经网络 支持向量机 

分 类 号：TK223[动力工程及工程热物理—动力机械及工程]