摘要
工业过程的模型通常具有非线性强、系统时变明显、工况变化大等特点。传统的滑动时间窗选择方法容易包含大量关联度低的数据,影响了数据模型的建模精度和建模效率。本文提出了一种带状态约束的滑动时间窗口选择算法,应用于工业连续过程软测量模型的训练数据选取。将该时间窗口选择算法与最小二乘支持向量回归算法结合,利用电站锅炉的历史运行数据,建立了燃煤电站锅炉尾部烟气含氧量的软测量模型。研究结果表明,相比于传统的滑动时间窗选择算法,利用该算法进行训练样本选取后,提高了所建立的最小二乘支持向量机模型的模型精度和运行效率。利用该方法建立的烟气含氧量软测量模型具有较高的精度,可以在氧化锆传感器发生故障时代替其工作,保证了氧量信号的稳定性和可靠性。
Most industrial process has features like strong nonlinearity,time-varying and frequent adjustment of work condition.The traditional moving window algorithm can take many irrelevant data to build data-driven model,which leads to the inaccuracy and badperformance of the model.In this paper,a state constrained moving window algorithm is proposed to select better training data for softsensing model.Using the proposed algorithm and partial least support vector machine,a soft sensors model is built.The result demonstrates that the LSSVM model is improved with the data selected by state constrained moving window algorithm.And the soft-sensing model of oxygen content is accurate and could replace the ZrO2 sensors in the failure time,to ensure the reliability of the oxygen content signal.
出处
《控制工程》
CSCD
北大核心
2014年第S1期115-117,120,共4页
Control Engineering of China
基金
国家自然科学基金项目(61174059
60934007
61233004)
国家973项目(2013CB035406)
上海市青年科技启明星跟踪计划(11QH1401300)
