基于数据挖掘技术的冷水机组与冷却塔模型辨识方法  

Identification Method of Chiller and Cooling Tower Model Based on the Data Mining Technology

作　　者：何磊[1] 雷波[1] 毕海权[1] 王晓亮[1] 

机构地区：[1]西南交通大学机械工程学院,成都610031

出　　处：《建筑科学》2015年第2期92-96,共5页Building Science

基　　金："十二五"国家科技支撑计划课题"高原气候适应性节能建筑关键技术研究与示范"(2013BAJ03B04)

摘　　要：本文通过对支持向量机的分析,采用数据挖掘技术建立了冷水机组和冷却塔的支持向量回归机辨识模型;以某公共建筑冷水机组、冷却塔历史运行数据为样本,分析了不同核函数、模型结构参数对辨识精度的影响,确定了适用于冷水机组和冷却塔采用历史运行数据辨识其特性的核函数及归一化方法。结果表明,采用[-1,1]规整样本数据能够提高模型辨识精度;冷水机组能耗模型适宜采用多项式核函数,而冷却塔释热量模型适宜采用径向基核函数;冷水机组及冷却塔的SVR辨识模型精度均高于经验模型。Based on the analysis of data mining technology of the support vector regression, the identification model of chiller and cooling tower are constructed. The operation data of a chiller and cooling tower in a public building are employed to analyze the effect of various kernel functions, parameters and normalization methods to identification precision. The results of analysis show that normalizing the sample data with [ - 1, 1 ] is conductive to improve the identification precision. And it is revealed that polynomial kernel function is a more appropriate choice for chiller identification and the radial basic function makes a batter fit for cooling tower identification. Moreover, the identification precisions of SVR models are superior to the empirical models for both the chiller and cooling tower.

关 键 词：数据挖掘 模型辨识 支持向量回归机 冷水机组 冷却塔 

分 类 号：TU831.3[建筑科学—供热、供燃气、通风及空调工程] TK523[动力工程及工程热物理—热能工程]