VSTCF of a Household for the Integration of Smart Grids
The recent integration of smart grid systems to present electric power systems and the increasing penetration of renewable energy sources make electrical energy consumption forecasting not only a prominent subject but also an arduous challenge due to nonlinear and nonstationary characteristics of electric loads which can be affected by seasonal effects, weather conditions, socioeconomic dynamics, and random effects. Very short-term electrical energy consumption forecasting (VSTCF), which includes few minutes to an hour ahead forecasting of electrical energy consumption, ensures monitoring energy consumption, identifying base and peak loads, making feasible decisions for renewable energy investments such as photovoltaic (PV) systems, and improving energy management quality of a household for the smart grid integration.
In this paper, for the first time in Turkey, electrical energy consumption data of a household with an averaging period of 10-minute is obtained by an energy logger during a 1-month period in order to perform VSTCF by using several artificial intelligence (AI) techniques including decision trees (DT), genetic algorithm (GA), artificial neural networks (ANN), and support vector machines (SVM) in the literature. After data pre-processing, various AI techniques will be applied to real-time data obtained from a household in the Mediterranean Region of Turkey for the calculation of mean absolute error (MAE) performance metric. Results indicate that gradient boosted decision trees (GBDT) have the best performance in comparison with other techniques for VSTCF.