催化剂粉尘浓度软测量建模研究与应用

Research on Soft-sensor Modeling of Catalyzer Particle Concentration Measurement and Its Application

采用基于Mie理论的激光散射法测量催化剂粉尘浓度时,催化剂粉尘浓度与监测参数——入射光强、散射光强、出射光强以及烟气流量之间存在着复杂的非线性关系,给粉尘浓度的准确测梁带来困难。利用支持向量机优良的非线性映射和强大的泛化能力,建立了一个基于最小二乘支持向量机的催化剂粉尘浓度软测量模型,给出了相应的系统结构和算法,并通过网格搜索和交叉验证的方法对支持向量机进行参数选择。采用遗忘因子法和数据滑动时间窗技术对工业软测量模型进行在线校正,克服了工况条件发生改变时的估计偏差,提高了估计精度。仿真和实际运行结果表明基于LS-SVM的软测量模型具有较高的估算精度与泛化能力,为催化剂粉尘浓度的在线测量提供了一种简单、可靠的新方法。

Being better nonlinear approximation and generalization capability of support vector machine（SVM）, a soft-sensor model based on least square support vector machine （LS-SVM） was introduced to realize the complex nonlinear relation between the particle concentration and its effects such as energy of scattering light, energy of incidence light, energy of transmission light and flue gas flow of oil refining equipment, and to provide a solution to on-line measuement of catalyzer particle concentration. The cross validation and gridding search method were used to select parametrs of LS-SVM model, and the data in sliding time window, incorporating with forgetting factor in the queue was used to updates the model in terms of both the new data and the old model to improve the estimate precision. The simulation results and the performance of practical application in industrial field show that the LS-SVM model gives better estimate precision, and has good generalization capabilities of on-line catalyzer particle concentration measurement.