Day-ahead optimal scheduling of smart electric storage heaters: A real quantification of uncertainty factors

Abstract

Optimized controls are particularly promising for flexible and efficient management of space heating and cooling systems in buildings. However, when controls are based on predictive models, their effectiveness is affected by the reliability of the models used. In this paper we propose a quantification analysis of some of the main uncertainty factors that can be observed in an optimal control really implemented in a building. A day-ahead optimal scheduling was applied to the heating system (composed of smart electric heaters with thermal storage) of a single room in an office building located in Osimo (Italy). The control algorithm is formulated to determine the charging periods of the heaters with the objective of minimizing the withdrawal of energy from the grid. The control takes into account the electricity produced by a photovoltaic plant and must maintain the internal air temperature close to an imposed setpoint.Firstly, the actual application of the control is shown during two selected days. Secondly, the analysis is extended to quantify the impact on the control performance of the prediction uncertainty of the input variables. The variable that has the greatest impact is the weather forecast and, specifically, the cloudiness index, which determines the solar gains. The different moment in time in which the weather forecast is predicted has proved to have a significant impact on the charging periods of the heaters (expected variation ranges from -50% to + 100%) and on the prediction of the indoor air temperature (variations observed up to 40%).