加入滑动时间窗算法室温异常数据识别与填补  

Adding Sliding Time Window Algorithm for Identification and Filling of Abnormal Room Temperature Data

作　　者：张珂 曹姗姗 孙春华[1] 夏国强[1] 吴向东 ZHANG Ke;CAO Shanshan;SUN Chunhua;XIA Guoqiang;WU Xiangdong

机构地区：[1]河北工业大学能源与环境工程学院,天津300401 [2]河北工大科雅能源科技股份有限公司,河北石家庄050000

出　　处：《煤气与热力》2024年第8期17-23,共7页Gas & Heat

基　　金：国家重点研发计划“基于可再生能源热泵利用的复合型区域供热供冷系统关键技术研究与示范”(02021YFE0116100)。

摘　　要：提出在室温异常数据识别方法的基础上加入滑动时间窗算法。结合算例,对最佳滑动参数(滑动窗口宽度、滑动步长)、室温数据采集时间间隔进行筛选,验证KNN算法对剔除数据进行填补的可信性。加入滑动时间窗算法,可以提高3σ准则、四分位数法、K-means聚类法对室温异常数据识别的准确性。滑动窗口宽度、滑动步长、室温数据采集时间间隔对室温异常数据识别准确性均有影响,应合理确定。由KNN算法填补的数据可信性比较高,尤其是剔除数据占比较小时。It is proposed to add sliding time window algorithm based on the method for identifying abnormal room temperature data.Combined with examples,the optimal sliding parameters(sliding window width and sliding step size)and room temperature data acquisition interval were screened to verify the credibility of the KNN algorithm in filling the excluded data.Adding the sliding time window algorithm can improve the accuracy of 3σ criterion,quartile method,and K-means clustering in identifying abnormal room temperature data.The sliding window width,sliding step size,and room temperature data acquisition interval all have an impact on the accuracy of identifying abnormal room temperature data,and should be reasonably determined.The credibility of the data filled by the KNN algorithm is relatively high,especially the proportion of excluded data is relatively small.

关 键 词：室内温度 滑动时间窗算法 异常数据识别 数据填补 

分 类 号：TU832[建筑科学—供热、供燃气、通风及空调工程]