基于最小二乘支持向量机的电站锅炉燃烧优化  

Combustion Optimization for Utility Boiler Based on Least Square-support Vector Machine

作　　者：顾燕萍[1] 赵文杰[2] 吴占松[1] 

机构地区：[1]电力系统国家重点实验室(清华大学热能工程系),北京市海淀区100084 [2]华北电力大学自动化系,河北省保定市071003

出　　处：《中国电机工程学报》2010年第17期91-97,共7页Proceedings of the CSEE

摘　　要：高效、低污染是电站锅炉燃烧优化的目标。该文基于最小二乘支持向量机,建立了电站锅炉燃烧模型,实现了飞灰含碳量、排烟温度、NOx排放量等参数的软测量和锅炉效率的预测;对比了最小二乘支持向量机和BP神经网络模型的性能,对比结果表明,最小二乘支持向量机具有训练时间短、泛化能力高等优点。提出2种锅炉燃烧优化方式,并以所建立的燃烧模型为基础,采用遗传算法对锅炉运行工况进行寻优,为分散控制系统基础控制层提供最佳的操作变量设定值。算例表明,文中所提出的燃烧优化方案可以有效提高电站锅炉效率和降低NOx排放量。High efficiency and low NOx emission are the two main goals of boiler combustion optimization.This paper applied least square-support vector machine （LS-SVM） to build the utility boiler combustion model,for the prediction of carbon content of fly ash,exhaust flue gas temperature,NOx emission,boiler efficiency,and so on.As a novel modeling method,the performance of LS-SVM was compared with that of the traditional BP neural network;the corresponding results indicate that LS-SVM needs less training time and has better generalization ability.Based on the above boiler combustion model,two combustion optimization modes were proposed and genetic algorithm was selected to solve the multi-objective and multi-constrained boiler combustion optimization problem.The results of boiler combustion optimization process provide best settings of manipulate variables for DCS.Numerical examples show that the proposed optimization control scheme in this paper is effective on improving utility boiler efficiency and reducing its NOx emission.

关 键 词：燃烧优化 锅炉效率 NOX排放 最小二乘支持向量机 遗传算法 

分 类 号：TK223[动力工程及工程热物理—动力机械及工程]