Time series method for short-term load forecasting using smart metering in distribution systems

Recently, the research community gave a particular attention to the "smart grid" concept since it enables a higher efficiency to answer users' energy needs. Smart grid networks relying on the exploitation of smart meters enable the design of more accurate forecasting models on the distribution grid. Within this context, this paper presents a time series forecasting model based on real measurements. It basically relies on the description of the power load through two components: a trend component and a cyclic (periodic) component. The two components are identified separately using, on the one hand a regression algorithm and on the other hand spectral techniques. The carried out simulations are based on real measurements collected from the French distribution network. These results are promising and show better results compared to a naive model.