摘要
以北京市212栋大型公共建筑样本为例,对电力部门提供的建筑电耗数据中常见的缺失数据和异常数据分别进行诊断和处理。预处理方法包括电耗数据的完整性识别及补全、年单位面积电耗相对极差检测、异常样本箱线图检测、采用多参数预测回归模型补全整月电耗数据等过程。完整的处理过程能为建筑电耗数据的预处理工作提供参考。
Taking the monitored power consumption data consumed in 212 large-scale public buildings in Beijing as the basis, the missing and abnormal data sets were diagnosed and processed respectively in this paper. The preprocessing method consists of the completeness identification and completion, the detection of the relative range of power consumption per unit area, the detection of the abnonnal sample by box plot, and the multi-parameter prediction regression model to complete the monthly power consumption data. This proposed complete framework provides a reference when dealing with the monitored building power consumption data.
作者
周优
陈义波
谭洪卫
仲敏
顾中煊
罗淑湘
ZHOU You1, CHEN Yi-bo1,2, TAN Hong-wei1,2,3, ZHONG Min4, GU Zhong-xuan4, LUO Shu-xiang4(1.School of Mechanical Engineering Tongji University, 201804, Shanghai, China; 2.UNEP-Tongji Institute of Environment for Sustainable Development, 200092, Shanghai, China; 3.Research Center of Green Building and New Energy of Tongji University, 200092, Shanghai, China; 4.Beijing Building Technology Development Co., Ltd., 100055, Beijing, China)
出处
《建筑技术》
2018年第5期469-472,共4页
Architecture Technology
基金
国家重点研发计划项目"基于全过程的大数据绿色建筑管理技术研究与示范"(2017YFC0704200)
关键词
电耗监测数据
数据预处理
数据异常
monitored power consumption
data preprocessing
data exception