摘要
针对目前锅炉飞灰含碳量难以准确测量的问题,应用支持向量回归和粒子群算法相结合建立了飞灰含碳量软测量模型;建模中以某电厂提供的1 000 MW超超临界机组的测试数据为研究对象,对数据进行了预处理,对各种变量进行了关联度分析,采用粒子群算法优化了模型的惩罚参数C和核函数参数g,建立了飞灰含碳量软测量模型;同时利用测试数据和另选的随机数据验证了模型的准确性和泛化能力;仿真结果表明,飞灰含碳量软测量模型的预测精度较高,相对误差被控制在±1%以内,而且泛化能力较强,为锅炉飞灰含碳量的测量提供了一种有效的途径。
A soft sensor model of the carbon content of fly ash is established by applying support vector regression and article swarm al gorithm about the problem that the carbon content of fly ash is difficult to measure accurately. In this work, carried on data preprocessing and the various variables' correlation analysis, and adopted article swarm algorithm to optimize the parameters C and g of the model according to the real time data of a Power Plant IO00MW ultra--supercritical unit. Moreover, the model's accuracy and generalization capability were Identified by using the test data and the extra random data. The simulation results show that the carbon content in fly ash soft sensor model simulation has a higher prediction accuracy that relative error is controlled within ±1% and better generalization ability, which provides an effective way for measurement of fly ash carbon content.
出处
《计算机测量与控制》
北大核心
2014年第2期345-348,共4页
Computer Measurement &Control
基金
河南省科技计划资助项目(132102110173)
关键词
支持向量回归
粒子群优化算法
数据预处理
飞灰含碳量
软测量
support vector regression
particle swarm optimization algorithm
data preprocessing
carbon content in fly ash
soft sensor