Statistical indicator for the detection of anomalies in gas, electricity and water consumption: Application of smart monitoring for educational buildings

摘要

Building facility managers are increasingly equipping their buildings with extensive sets of sensors. This article aims to develop an analysis decision-making methodology based on the production of statistical indicators. The tracking of such indicators allows detecting any systems performance problems. The automatic pinpointing of malfunctions can serve to activate alerts. Our approach focuses on the processing of data stemming from secondary schools managed by departmental services in the Pas-de-Calais, where 117 secondary school buildings have been instrumented with various sensors and supplying data since 2015. This article starts with a close-up on data mining for water, gas and electricity consumption. Data mining and machine learning methods, including the Clustering approach (K-Means), have been used to extract information from the measurements conducted in 2015 and 2016. This information is used to classify the 2017 measurements according to supervised approaches (SVM). The specificity of this work is to delve deeper into the analysis by combining into the same algorithm a set of various sensors related to both energy use and building occupancy. The data classification results have allowed highlighting "atypical" operations during the daytime, through interpreting data classification results in an effort to define the status of every day in year 2017. (C) 2019 Elsevier B.V. All rights reserved.