基于PCA-PSO-SVM的供热负荷预测研究  

Research on heat supply load forecasting based on PCA-PSO-SVM

作　　者：李明柱[1] 程丹 王梓玮 LI Mingzhu;CHENG Dan;WANG Ziwei(Jilin Jianzhu University,Jilin 130119,China)

机构地区：[1]吉林建筑大学,吉林长春130119

出　　处：《区域供热》2023年第5期146-153,共8页District Heating

基　　金：吉林省科技发展计划项目-基于BIM的公共建筑运行能耗大数据挖掘应用研究与示范(20190303059SF)。

摘　　要：以长春某热力公司的集中供热系统为研究对象,提出了可用于热负荷预测的PCA-PSO-SVM模型。首先利用PCA进行降维,然后通过PSO优化算法选取最优参数c、g和ε,从而构建PCA-PSO-SVM的预测模型。仿真结果表明,经过PCA降维处理的模型预测精度略低,但模型的预测速度可以提高20%~40%左右;此外也验证了基于PSO优化模型的预测精度较高,模型拟合度较好。PCA-PSO-SVM model for heat load prediction is proposed based on the central heating system of a heat company in Changchun.Firstly,PCA is used for dimensionality reduction,and then the optimal parameters c,g andεare selected by PSO optimization algorithm to construct the PCA-PSO-SVM prediction model.The simulation results show that the prediction accuracy of the PCA model is slightly lower,but the speed of the model can be improved by 20%~40%.In addition,the prediction accuracy of PSO-based optimization model is higher and the model fit is better.

关 键 词：供热负荷预测 主成分分析 粒子群优化 支持向量机 

分 类 号：TU995[建筑科学—供热、供燃气、通风及空调工程]