摘要
火电厂燃煤锅炉产生的NO_x是大气污染物的重要来源之一,建立有效的NO_x排放预测模型是降低NO_x排放的基础。针对火电厂控制系统数据的海量化和高维化及燃煤锅炉多参数多变量相互耦合的特点,首先利用主成分分析法对火电厂分布式控制系统(DCS)数据进行特征提取,消除各特征变量间的耦合性;然后将提取的特征作为长短期记忆(LSTM)神经网络的输入,得到火电厂NO_x排放预测模型。将该模型与传统循环神经网络(RNN)模型、最小二乘支持向量机(LSSVM)模型应用于某超超临界660 MW机组燃煤锅炉对NO_x排放质量浓度进行预测。结果表明:LSTM神经网络和RNN模型预测效果均优于LSSVM模型;本文提出的LSTM神经网络模型预测准确率达到79%,均方根误差为0.398,优于其他2种模型;LSTM神经网络模型数据跟踪效果明显优于RNN模型,预测结果波动较小,模型稳定性和准确率较高。
The NOx produced in the combustion process of coal in thermal power plants is one of the most important sources of air pollutants.An effective prediction model should be established for reducing the NOx emissions.In view of the massive and high dimensional system data of thermal power plants and the multivariable mutual coupling characteristics of the coal-fired boilers,the components analysis method was applied to extract the data from the DCS and eliminate the coupling of each feature variable.Then,the extracted data were used as input of the long short term memory(LSTM)neural network to establish the prediction model for NOx emission.Moreover,this model,the conventional circulating neural network(RNN)model of NOx emission prediction and the least square support vector machine(LSSVM)model were applied to the actual working condition data of an ultra supercritical 660 MW coal-fired boiler.The results show that,the prediction effect of the LSTM and RNN models was better than that of the LSSVM model.The prediction accuracy of the proposed model was 79%and the root-mean-square error was 0.398,indicating the proposed LSTM model is superior to the conventional RNN model and the LSSVM model.Furthermore,the data tracking effect of the model is obviously better than that of the RNN model,the prediction result fluctuates less and the model has higher stability and accuracy.
作者
杨国田
张涛
王英男
李新利
刘禾
YANG Guotian;ZHANG Tao;WANG Yingnan;LI Xinli;LIU He(School of Control and Computer Engineering,North China Electric Power University,Beijing 102206,China)
出处
《热力发电》
CAS
北大核心
2018年第10期12-17,共6页
Thermal Power Generation
