基于LSSVM-GPC的流化床锅炉多变量协调控制方法  

Multivariable coordinated control method of FBC boiler based on LSSVM-GPC

作　　者：孙立[1] 潘蕾[2] 沈炯[2] 

机构地区：[1]清华大学热能工程系,北京100083 [2]东南大学能源热转换及其过程测控教育部重点实验室,南京210096

出　　处：《东南大学学报（自然科学版）》2013年第2期312-316,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金重点资助项目(51036002);国家自然科学基金资助项目(51106024)

摘　　要：为了获得良好的流化床锅炉控制品质,提出了一种基于最小二乘支持向量机广义预测控制(LSSVM-GPC)的多变量协调控制方法,以适应流化床锅炉多变量、强耦合、大滞后的动力学特性.在研究流化床机理模型的基础上,采用LSSVM算法辨识流化床模型,并将所得的决策函数转化为广义预测模型.对比结果显示,LSSVM预测模型能够准确描述对象输出特性,并有效去除测量噪声.为了解决FBC多变量预测控制中易出现的病态矩阵以及调节量动作频繁等问题,进一步利用关联分析法,设计了基于LSSVM-GPC的流化床协调控制方法.仿真结果表明,结合该方法能使锅炉负荷响应具有良好的快速性和稳定性,同时能保持床温基本稳定,并有效避免调节量的频繁变化,达到了节能优化控制的目的.In order to achieve good control performance of fluidized bed combustion （FBC） boilers with the dynamic characteristics of multi-variables, strong coupling, and time delays, a coordinated generalized predictive control method based on the least-squares support vector machine （LSSVM- GPC） is developed. First, a precise identification model from the FBC mechanism model is obtained by using the LSSVM approach, and then the generalized predictive model is derived from the LSS- VM decision function. The comparisons among several modeling approaches show that the LSSVM prediction model can accurately describe the output characteristics of the plants and effectively re- move the measurement noises. For avoiding ill-conditioned matrixes and frequent varying of manipu- lating variables in the control decisions, a coordinated control strategy based on the LSSVM-GPC al- gorithm is developed by using the correlation analysis on the FBC process. Simulation results show that the approach obviously improves the rapidness and stability of the FBC load control; meanwhile it keeps the bed temperature settled well. Further, it avoids the frequently varying of the actuators; thus the control strategy is optimal and energy-saving.

关 键 词：流化床 机理模型 支持向量机 协调预测控制 

分 类 号：TK323[动力工程及工程热物理—热能工程]