应用集成极限学习机的电站SCR脱硝系统建模与分析
Modeling and analysis of SCR denitration system in power plant based on ensemble of extreme learning machine
马宁 1尚勇 1尤默 1杨振勇 1李展 1刘磊 1康静秋1
作者信息
- 1. 华北电力科学研究院有限责任公司,北京 100045
- 折叠
摘要
针对火电机组普遍采用的选择性催化还原(selective catalytic reduction,SCR)脱硝反应容易受到环境因素影响,且具有非线性、大迟延、强扰动特点,难以建立准确反应模型的问题,提出一种基于集成极限学习机的SCR脱硝系统建模方法.首先,选择 4种具有不同激活函数的极限学习机和核极限学习机作为基学习器,分别建立SCR脱硝系统模型;然后,利用偏最小二乘算法将各基学习器的结果进行集成;最后,采用某 1 000 MW超超临界机组实际运行数据与所提建模方法相结合建立SCR脱硝系统模型.实验结果验证模型的有效性;与其他方法所建模型对比,结果表明基于集成极限学习机的SCR脱硝系统模型具有更好的模型泛化能力.
Abstract
Aiming at the problem that the selective catalytic reduction(SCR)denitration reaction commonly used in thermal power units is easily affected by environmental factors,and has the characteristics of nonlinearity,large delay and strong disturbance,so it is difficult to establish an accurate reaction model,a modeling method of SCR denitration system based on ensemble of extreme learning machine is proposed.Firstly,four kinds of extreme learning machines and kernel extreme learning machines with different activation functions are selected as the base learners to establish the SCR denitration system model respectively,and then the results of each base learner are integrated by using the partial least squares algorithm.Finally,the SCR denitration system model is established by combining the actual operation data of a 1 000 MW ultra supercritical unit with the proposed modeling method.The experimental results verify the effectiveness of the model.Compared with other modeling methods,the results show that the SCR denitration system model based on integrated limit learning machine has better model generalization ability.
关键词
火电机组/选择性催化还原/极限学习机/集成Key words
thermal power unit/selective catalytic reduction/extreme learning machine/ensemble引用本文复制引用
基金项目
河北省自然科学基金(E2018502111)
中央高校科研基金(E2018502111)
出版年
2024
NETL
NSTL
维普
万方数据