基于最小二乘支持向量机的锅炉对流受热面清洁吸热量预测  

Prediction of boiler convection heating surface clean heat absorption based on least squares support vector machine

作　　者：安连锁[1] 马美倩[1] 沈国清[1] 张世平[1] 吕伟为[1] 

机构地区：[1]华北电力大学电站设备状态监测与控制教育部重点实验室,北京102206

出　　处：《华北电力大学学报（自然科学版）》2013年第1期55-60,共6页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家自然科学基金资助项目(50976034);中央高校基本科研业务费专项资金资助(09QG43)

摘　　要：针对神经网络拓扑结构复杂、易出现过度训练、仅获局部最优解的问题,为提高锅炉对流受热面清洁时潜在吸热量预测的准确度,更好地进行受热面污染监测,提出了一种新的基于最小二乘支持向量机的对流受热面清洁时潜在吸热量预测方法。依据最小二乘支持向量机预测原理,建立对流受热面清洁时潜在吸热量最小二乘支持向量机预测模型,同时建立神经网络预测模型进行对比研究,实例研究结果表明,最小二乘支持向量机较神经网络具有更高的拟合度,预测各性能都高于神经网络,其在对流受热面清洁时潜在吸热量预测方面明显优于神经网络,将成为对流受热面清洁时潜在吸热量预测也即受热面污染监测方面更为有利的工具。As neural network suffers from the problems like complex network topology and overtraining, and the existence of local optimal solution, a new method based on least squares support vector machines （LS-SVM） was put forward to improve the prediction accuracy of the convection heating surface clean heat absorption and monitor the heating surface fouling better. According to the prediction principle of LS-SVM, LS-SVM prediction model of the convection heating surface clean heat absorption was established, at the same time, neural network prediction model was established for comparative studies. Case study shows that LS-SVM has higher degree of fitting and better performances than neural network, LS-SVM can become more favorable prediction tool for the convection heating surface clean heat absorption and the heating surface fouling monitoring.

关 键 词：清洁吸热量 最小二乘支持向量机 神经网络 预测 

分 类 号：TM621.2[电气工程—电力系统及自动化]