摘要
基于BP神经网络算法和遗传算法优化的BP神经网络算法(GA-BP算法)建立了燃煤锅炉内H2S质量浓度的预测模型,以燃烧调整试验得到的数据作为训练样本和测试样本,对所建立的H2S质量浓度预测模型进行评价。结果表明:基于BP神经网络算法建立的H2S质量浓度预测模型出现过拟合现象,而基于GA-BP算法建立的H2S质量浓度预测模型具有较好的逼近能力和泛化能力,可用于锅炉内H2S质量浓度建模预测,为运行优化和控制燃煤锅炉内H2S质量浓度奠定了基础。
Prediction models for the H2S concentration in a coal-fired boiler were established respectively based on BP neural network algorithm and GA-BP algorithm, which were evaluated with the samples obtained from combustion adjustment experiments. Results show that overfitting happens when the model built on the basis of BP algorithm is used to predict the H2S concentration, while the model based on GA-BP algorithm has better approximation and generalization capability, which can be used to effectively predict the H2S concentration in a boiler, and therefore may serve as a reference for further control of H2S concentration in the boiler and for operation optimization of the unit.
作者
相明辉
梁占伟
孙贻超
王新钢
XIANG Minghui;LIANG Zhanwei;SUN Yichao;WANG Xingang(Yantai Longyuan Power Technology Co.,Ltd.,Yantai 264006,Shandong Province,China;Shenhua Guohua(Beijing)Electric Power Research Institute Co.,Ltd.,Beijing 100024,China)
出处
《动力工程学报》
CAS
CSCD
北大核心
2020年第6期433-439,共7页
Journal of Chinese Society of Power Engineering
基金
国家重点研发计划资助项目(2017YFB0603900,2017YFB0603904)。
