基于变量筛选的烟气流量软测量研究

Research on Soft Measurement of Flue Gas Flow Based on Variable Selection

为了解决烟气流量主要依靠硬件传感器测量,高温腐蚀使得传感器损坏,烟气湍流、紊流现象的存在,使得测量不准确等问题,提出一种基于投影重要性分析和前向搜索筛选辅助变量,使用支持向量机建模的软测量方法,测量SCR反应器入口烟气流量。以电厂阵列式传感器测量出的烟气流量作标准,使用最小二乘支持向量机方法建立模型预测输出,对比投影重要性分析筛选辅助变量和不筛选变量结果表明,投影重要性分析和前向搜索筛选辅助变量的方法对建模有较好优势,能够减小系统结构,增加软测量建模的泛化能力。

To solve the problems of the dependence of flue gas flows on hardware sensor damage made by high temperature corrosion flue gas turbulence and inaccurate measurement, this paper proposed a method based on the analysis of projection importance and the auxiliary forward searching screening variables. By using support vector machine modeling of soft measuring method, the SCR reactor inlet gas flow is measured. Based on an array of sensors for measuring gas flow standard in current power plant, by applying the least squares support vector machine （SVM） method and the comparison method, the detailed analysis is conducted. Results show that projection analysis of the importance and the auxiliary forward searching screening variables are superior in measuring methods, and this method expands the new way of flue gas flow measurement.