基于改进FCM算法的办公建筑负荷曲线聚类分析  

Cluster Analysis of Office Building Load Curve based on Improved FCM Algorithm

作　　者：齐家业 胡翀赫 王喜春 于兵 Qi Jiaye;Hu Chonghe;Wang Xichun;Yu Bing(Shanghai DFYH Tech Services Co.,Ltd.,Shanghai,20060;Shanghai Yanhua Smartech Group Co.,Ltd.,Shanghai,20060;Postdoctoral Station of Mechanical Engineering,Tongji University,Shanghai,201804)

机构地区：[1]上海东方延华节能技术服务股份有限公司,上海20060 [2]上海延华智能科技(集团)股份有限公司,上海20060 [3]同济大学机械工程博士后流动站,上海201804

出　　处：《建设科技》2021年第19期77-81,94,共6页Construction Science and Technology

基　　金："国家重点研发计划政府间国际科技创新合作/港澳台科技创新合作"(编号:2019YFE0100300-21);"住房和城乡建设部科技示范项目"(编号:S20200064);"上海市信息化发展专项资金(信息化建设和应用)"(编号:201901065)。

摘　　要：为探究夏热冬冷地区办公建筑在不同季节的日负荷特征,拟采用FCM算法进行聚类分析。考虑到模糊C均值聚类算法(FCM)所存在的易受初始聚类中心影响,且易出现局部收敛的情况,提出基于量子行为约束的天牛群算法对其进行改进,以对FCM算法初始聚类中心进行优化并提升其全局寻优性能。UCI数据集算例表明基于改进天牛群算法的FCM算法具有较好的准确性及稳定性,所得出办公建筑负荷特性曲线对建筑能效提升、负荷预测和用户侧管理均有积极意义。In order to explore the daily load characteristics of office buildings in different seasons in hot summer and cold winter areas,it is proposed to use FCM algorithm for cluster analysis.Considering the existence of fuzzy C-means clustering algorithm(FCM)that is easily affected by the initial clustering center and is prone to local convergence,a BSO algorithm based on quantum behavior constraints is proposed to improve the initial FCM algorithm in order to optimize the cluster center and enhance its overall search performance.The UCI data set example shows that the FCM algorithm based on the improved Tianniu group algorithm has good accuracy and stability,and the derived office building load characteristic curve has positive significance for building energy efficiency improvement,load forecasting and user-side management.

关 键 词：办公建筑 负荷聚类 量子行为 天牛群算法 FCM算法 

分 类 号：TU831.2[建筑科学—供热、供燃气、通风及空调工程]