一种燃煤质量的综合预测方法  

AN INTEGRATED FORCASTING MERHOD FOR FUEL COAL QUALITY

作　　者：吴德会[1] 

机构地区：[1]九江学院,江西九江332005

出　　处：《热力发电》2008年第1期26-30,49,共6页Thermal Power Generation

基　　金：国家自然科学基金资助项目(70272032)

摘　　要：建立了一种新的基于多元线性回归与最小二乘支持向量机(LS-SVM)的燃煤发热量综合预测模型,并给出了建模步骤和具体算法。考虑到燃煤实际发热量与其工业分析组成(水分、灰分和挥发分)之间的关系,由初步预测子模型和误差修正子模型构成综合模型。用煤样实测发热量与其工业分析组成的原始数据构造训练样本集,用多元线性回归算法对发热量进行初步预测,用LS-SVM修正子模型获得误差补偿量。综合模型最终的预测值为初步预测子模型的输出值加上误差补偿量。实际预测结果表明,这种综合预测方法耗时少,预测效果优于常用方法,具有一定的应用价值。A new integrated forcasting model for calorific value of coal has been established based on multi- linear regression and least square support vector machine （LS -SVM）, and steps for establishing the said model and the concrete algorithm being given, consid- ering the relationship between the practical calorific value of fuel coal Qb.ad and the industrial analysis data （Wad ,Ad ,Vd,f）, the integrated model is composed of two parts, mamely the preliminary sub- model and the error correction sub - model. The process of modelling is as in the following： firstly, the sample sets to be constructed by using the ofviginal data, then, the preliminary forcasting is obtained by a multivariate linear regression algorithm, and finally the compensated error is obtained from correction sub -model based on I.S- SVM, so the prediction value will be the output value of the preliminary forecasting sub- model plus the compensated error. The practical prediction results show the intgrated model can obviously improve the prediction precision compared to other al- gorithms, being less time- comsuming, and effectiveness of prediction being superior than other commonly used methods, having certain applying value.

关 键 词：燃煤质量 发热量 多元线性回归 最小二乘支持向量机 综合预测 预测模型 

分 类 号：TQ533.4[化学工程—煤化学工程]