Research On Mining Daily Energy Consumption Patterns Of Room And Evaluation Of Large Office Buildings

Different occupancy patterns may lead to huge differences in building energy consumption and require different energy management policies and building energy conservation standards to reduce building energy consumption.However,due to the complexity of occupant behavior,potential privacy issues lead to the difficulty of collecting a large number of occupant behavior data,and the Hawthorne Effect may lead to systematic deviations between experiment and truth,establishing accurate,quantitative,and practical occupant behavior models is difficult.Therefore,it is increasingly important and urgent to find an approach that it can be used to apply occupant behavior to engineering.Energy consumption of indoor directly reflects the occupancy energy consumption patterns.So,extracting it's characteristic from room energy consumption data is an effective research method.With the rapid development of China's building energy monitoring platform,a large amount of energy consumption data has stored.Therefore,daily energy consumption patterns of room and evaluation have studied in this paper by a data mining framework based on a large amount of room lighting and socket energy consumption data which is collected from a completed large government building energy monitoring platform by our research group.The main research of this paper is including:(1)Energy consumption data is easily affected by noise,sensor failure,communication interruption,and other factors on steps of collection,communication,transmission,storage and others.So,some abnormal datas that seriously affect data quality are produced,such as missing data and cumulative energy consumption data.For this situation,a recognition method based on the characteristic of accumulative length of time and multiple of the average of non-zero energy consumption is proposed to identify abnormal data by accumulated based on statistical surveys.(2)Taking the offset Euclidean distance as the similarity measure distance,using the clustering algorithm,combining with the characteristics of day-to-day cycles of energy consumption in a building,5 typical room daily energy consumption patterns are obtained through the preliminary cluster classification which determined by Davies-Bouldin indicator.Random distribution models of the daily energy consumption patterns have been found by hypothesis test statistic method.Which provides a basis and quantitative modeling method for distinguishing abnormal energy consumption patterns and energy simulation of this type of buildings.(3)Energy consumption benchmarks are respectively determined by quartile and cluster analysis method for different daily occupancy energy consumption patterns under different types of room.Based on the energy consumption benchmarks,discussing the applicability of the two methods at different stages of energy conservation management.What's more,evaluation of indoor lighting and socket energy consumption,the association rules mining by multiple minimum supports have been done for further energy saving and refined energy management.