用于SCR喷氨量模型参数辨识的辅助变量递推最小二乘法

Recursive instrumental variable estimation algorithm for ammonia flow model in SCR denitration reactors

为了对火电厂选择性催化还原(SCR)烟气脱硝系统中喷氨量-出口NOx质量浓度控制对象模型参数进行在线辨识,针对现场有色噪声数据,采用辅助变量对递推最小二乘法进行了改进,构成辅助变量递推最小二乘法,并对含噪声二阶惯性加纯延迟模型参数进行辨识。同时,结合变回归估计(FMVRE)方法得到最小化损失函数,通过两步辨识得到系统的惯性时间、比例系数及迟延时间。以某350MW机组SCR烟气脱硝系统为对象,采用标准递推最小二乘法和辅助变量递推最小二乘法进行参数辨识。结果表明,辅助变量递推最小二乘法的辨识参数收敛更快,辨识精确度更高,所得模型能够较好地表征实际系统的动态和静态特性。因此,基于辅助变量递推最小二乘法的参数辨识方法可用于SCR烟气脱硝系统控制对象的一类二阶惯性加纯延迟模型参数的在线辨识。

In order to estimate the parameters of selective catalytic reduction（SCR）denitration thermodynamics system in power plants,concerning the fact that the relevant data contains colored noises,this paper introduces an on-line estimation method based on standard recursive least squares with instrumental variable for the estimation of both time delay and other parameters.Combining with the variable regression estimation method,the parameters and time delay can be obtained by minimizing the loss function in a two-step estimation algorithm.Moreover,estimation is conducted with both standard and instrumental variable recursive least square for the SCR system in a 350 MW unit.The simulation results show that the introduced estimation algorithm improves the results’ convergency and accuracy,indicating this method is appropriate for estimating the parameters of the model of second-order inertia transfer function plus a time delay.