基于改进即时学习算法的吸收塔pH值软测量技术研究

Research on Soft Sensor Technology of the pH Value of Absorption Tower Based on Improved Just-in-time Learning Algorithm

针对火电厂脱硫系统工作过程中存在非线性、时变性、多变量等问题,提出了一种基于改进即时学习算法的脱硫系统吸收塔浆液pH值软测量模型。在选择即时学习算法相似样本时,为了充分考虑输入变量与输出变量之间的相关性,采用一种基于偏最小二乘法(partial least squares,PLS)的相似准则确定系统当前工作点的建模邻域,利用得到的建模邻域建立基于最小二乘支持向量机(least squares support vector machine,LSSVM)的局部模型预测当前pH值。将该方法应用于脱硫系统吸收塔pH值建模仿真,结果表明,该软测量模型具有良好的预测性能。

To solve the problems such as nonlinear,time-varying and multivariable in the desulfurization system of thermal power plant,a soft sensor model of the pH value of the absorption tower in desulfurization system based on improved just-in-time learning algorithm was proposed.In the selection of similar samples in just-in-time learning algorithm,in order to fully consider the correlation between input and output variables,a similarity criterion based on partial least squares(PLS)was used to determine the modeling neighborhood of the current operating point of the system.Through the obtained modeling neighborhood,a local model based on least-squares support vector machine(LSSVM)was established to predict the current pH value.This method was applied to the pH value modeling and simulation of absorption tower in desulfurization system.The results show that the soft sensor model has good prediction performance.