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
张珂 1曹姗姗 1孙春华 1夏国强 1吴向东2
作者信息
- 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
- 折叠
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.
Abstract
It is proposed to add sliding time window algorithm based on the method for identifying abnormal room temperature data.Combined with exam-ples,the optimal sliding parameters(sliding window width and sliding step size)and room temperature data acquisition interval were screened to verify the credibil-ity 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 tempera-ture 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 deter-mined.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Key words
indoor temperature/sliding time window algorithm/abnormal data identification/data filling引用本文复制引用
出版年
2024
国家科技期刊平台
NSTL
万方数据