Comprehensive Intellectual and Statistical Analysis of Water Consumption

This article deals with the water consumption regime in a residential building. The study is based on data of cold and hot water hourly consumption in a multi-storey apartment building. The measurement period is one month. The study comprehensively uses statistical analysis of water consumption and data mining of group outliers. Statistical data analysis is designed to determine the distribution pattern of different data samplings. The analysis is carried out for three different samplings of apartment water consumption data. As a result, group outliers of hourly water consumption are identified. Machine learning methods are used to identify group outliers. The task boils down to clustering the hours of the day to find hours with reduced (nighttime) water consumption. Clustering is carried out using five methods, and clustering quality is assessed by three metrics. As a result, nighttime consumption periods are determined for different samplings of water consumption data in apartment buildings. In general, comprehensive intellectual and statistical analysis of water consumption is useful for solving the tasks of designing water supply and sanitation systems, adjusting the operating modes of engineering equipment, and clarifying the calculated parameters of water consumption in apartment buildings.