火电厂CO2捕集系统的LSSVM-Hammerstein建模方法

The LSSVM-Hammerstein modeling method for CO2 Capture Systems in Thermal Power Plants

针对火电厂燃烧后CO2捕集过程的强非线性及大惯性等动态特性,研究准确性高的建模方法是系统优化设计的前提,本文从控制角度研究CO2捕集系统的非线性动态辨识建模方法。首先介绍火电厂CO2捕集系统工作原理,将系统抽象为二输入二输出的受控对象;然后采用基于最小二乘支持向量机的Hammerstein模型(LSSVM-Hammerstein)辨识方法,将CO2捕集系统表示为静态LSSVM模型与动态线性模型的组合。辨识结果表明该模型有很高的辨识精度和泛化能力,能精确描述CO2捕集系统非线性动态特性,为后续研究先进控制算法提供仿真平台和设计依据。

In view of the strong nonlinear and large-inertia dynamic characteristics of the postcombustion CO2 capture process in thermal power plant,studying a high-accuracy modeling method is the premise of the system optimization design.In this paper,the nonlinear identification modeling method of CO2 capture system is studied from the perspective of control.Firstly,the working principle of a post-combustion CO2 capture system in thermal power plant is introduced,and then the whole system is abstracted as a double-in-and-double-out controlled plant.Secondly,using the least-squares-support-vector-machines-based Hammerstein model(LSSVM-Hammerstein)identification method,the CO2 capture system is represented by the combination of static LSSVM model and dynamic linear model.Identification results show that the LSSVM-Hammerstein model has a higher identification precision and better generalization ability than the ARX model and the polynomial-based Hammerstein model.It can accurately describe the nonlinear dynamic characteristics of the CO2 capture system,and thus can be used for the simulation platform and as the design basis for the subsequent research on advanced control algorithm for CO2 capture process.