摘要
我国已经建立了大量的建筑能源系统互联网(iBES)能耗监测平台,然而平台底层网络和数据传输与处理过程中的数据质量问题较为突出。本文针对iBES建筑能耗监测平台数据异常问题,提出了一种将本征正交分解法,线性随机估计法以及分型关联维数相结合的POD-LSE-FCD异常数据检测方法,并将该方法应用于大连市某高校能耗监测数据,分别模拟计算了同一时间跨度不同步长的情况。结果表明,异常值出现的时间步骤,其直接POD和POD-LSE时间系数的FCD标准差明显高于其余计算步骤的标准差,可根据异常值检测准确度和计算量两个条件确定一个最优步长。POD-LSE-FCD法可以准确并快速的检测出能耗异常数据,适宜应用于i BES建筑能耗监测平台异常数据诊断。
China has established a large number of i BES building energy consumption monitoring platforms. However,the data quality problems in the underlying network, data transmission and processing process of i BES building energy consumption monitoring platform are particularly prominent. Aiming at the problem of abnormal data, this paper proposes a method called POD-LSE-FCD to detect the abnormal data. This technology combines the Proper orthogonal decomposition method(POD), Linear stochastic estimation(LSE) and Fractal correlation dimension(FCD). It is applied to the energy consumption monitoring data of a university in Dalian, and the situations of different steps in the same time span are simulated. The results show that the standard deviation of FCD of POD and POD-LSE time coefficients is significantly higher than that of other calculation steps in the time step of outlier occurrence. An optimal step can be determined according to the accuracy of outlier detection and the amount of calculation. POD-LSE-FCD method can accurately and quickly detect the abnormal data of energy consumption, which is suitable for the abnormal data diagnosis of iBES building energy consumption monitoring platform.
作者
马忠娇
张吉礼
MA Zhong-jiao;ZHANG Ji-li(Faculty of Infrastructure Engineering,Dalian University of Technology)
出处
《建筑热能通风空调》
2022年第9期1-5,47,共6页
Building Energy & Environment
基金
国家重点研发计划项目(2017YFC0704200)。
关键词
本征正交分解
建筑能源系统互联网
异常数据检测
数据质量
proper orthogonal decomposition
internet of building energy system
abnormal data detection
data quality