Statistical Modeling for Real Domestic Hot Water Consumption Forecasting

Abstract

In this paper, we study the real domestic hot water (DHW) consumptions from single family houses equipped with solar hot water tank. We model it to understand and forecast the daily needs of inhabitants. Thus, the forecasts can be integrated in a control strategy to optimize the energy cost by heating only the necessary DHW volume. At first, we realize a data analysis from real uses of several dwellings to lay assumptions of the statistical model. This study highlights a weekly periodicity, random fluctuations and the different profiles of consumption following the residence, the season, and the day of the week. Otherwise, having no prior information like the location or the number of residents, we propose an adaptive time series model which does not require strong a priori and computational time. Then, we develop an ARMA model to forecast the daily DHW volume and we apply it on each individual installation. This model allows to take into account the periodicity of one week, the consumption of the previous days and random fluctuations. The results obtained on real data show that this approach is very promising.