基于GA-SVM的电站锅炉烟气含氧量软测量模型

发布时间：2018-03-15 11:41

本文选题：烟气含氧量　切入点：支持向量机　出处：《热力发电》2017年04期 　论文类型：期刊论文

【摘要】：针对电厂烟气含氧量测量存在的投资大、精度低等问题,在烟气含氧量理论研究的基础上,选择合理的二次变量,引入支持向量机(SVM)建立二次变量与烟气含氧量的软测量模型。利用遗传算法(GA)对模型中的惩罚系数和核函数参数进行寻优,进而利用最优值构建了GA-SVM烟气含氧量软测量模型,并对比利用粒子群算法及网格搜索法对参数的寻优结果,对模型的准确性、泛化性进行测试。仿真结果表明:遗传算法比粒子群算法和网格搜索法更易找到全局最优解;GA-SVM软测量模型误差在±0.2%以内,相对误差在±4%以内,能满足不同负荷、不同时间段锅炉烟气含氧量的预测要求,其对烟气含氧量的测量更准确。
[Abstract]:Aiming at the problems of large investment and low precision in the measurement of flue gas oxygen content in power plant, the reasonable secondary variable is selected on the basis of theoretical study of flue gas oxygen content. This paper introduces support vector machine (SVM) to establish a soft sensor model of secondary variable and oxygen content in flue gas. The penalty coefficient and kernel function parameter in the model are optimized by genetic algorithm (GA), and the soft sensing model of oxygen content in GA-SVM flue gas is constructed by using the optimum value. The accuracy of the model is compared by using particle swarm optimization algorithm and mesh search method. The simulation results show that the global optimal solution of GA-SVM is less than 卤0.2%, and the relative error is less than 卤4%. The simulation results show that the genetic algorithm is easier to find the global optimal solution than particle swarm optimization algorithm and grid search method, and the relative error is less than 卤4%, which can satisfy different loads. The prediction of flue gas oxygen content in different time periods requires that the measurement of flue gas oxygen content is more accurate.
【作者单位】：华北电力大学能源动力与机械工程学院;华北电力大学自动化系;国网北京市电力公司;
【分类号】：TP18;TM621.2

【相似文献】