摘要
为保证燃煤电厂烟气脱硝系统的安全、稳定运行,需要制定科学合理的选择性催化还原(SCR)催化剂寿命预测方案。SCR催化剂失效是多个物理和化学因素共同作用的结果,难以用传统的物理模型或数学公式对其失活程度进行预测。本研究针对电厂大数据特性,对原始数据进行预处理,建立了曲线拟合、灰色预测、BP神经网络、灰色神经网络4种预测模型。实例对比分析发现:数据预处理可以提高预测精度;当数据满足等时距特性时,灰色神经网络优化后的直接输出模型预测精度较高;当数据不满足等时距特性时,使用BP神经网络模型预测效果更好。
In order to ensure the safe and stable operation of denitrification system in coal-fired power plants, a scientific and reasonable life prediction plan must be formulated for the SCR catalysts. The deactivation of the SCR catalysts is determined by the combined effects of multiple physical and chemical factors. Therefore, it is difficult to predict the catalysts5 service life by using conventional physical models and mathematical formulas. According to the characteristics of big data in power plants, this article preprocessed the raw data and established four prediction models, including curve fitting model, grey prediction model, BP neural network model and grey neural network model. Through case analysis, it is found that data preprocessing can improve the prediction accuracy. Generally, the optimized direct output model of the grey neural network shows high accuracy for the data that met the equidistant time requirement. Whereas, the BP neural network model can achieve better prediction results for the non-equidistant time data.
作者
唐诗洁
陆强
王则祥
吴昱廷
董长青
杨勇平
TANG Shijie;LU Qiang;WANG Zexiang;WU Yuting;DONG Changqing;YANG Yongping(National Engineering Laboratory for Biomass Power Generation Equipment,North China Electric Power University,Beijing 102206,China)
出处
《热力发电》
CAS
北大核心
2019年第3期61-68,共8页
Thermal Power Generation
基金
国家重点基础研究发展计划(973计划)项目(2015CB251501)
北京市科技新星(Z171100001117064)
霍英东教育基金会(161051)~~
