Using smart grid data to predict next-day energy consumption and photovoltaic production

The rise of sustainable energy production is a challenge for grid operators, who need to balance consumer demand with an increasingly volatile supply that is heavily dependent on weather conditions and environmental factors. Smart gird data provides fine-grained insight into consumer behavior as well as local renewable energy producers. We use data from an electric company in a region of South Tyrol to model both energy consumption as well as energy production. With a simple nearest-neighbor approach, we predict next-day load profiles for local power stations with relative error rates as low as 3 %. The energy production at these local power stations (in the form of photovoltaic power plants) can be predicted by adapting an ideal irradiation model to actual production data, stratified by varying weather conditions. Using this approach, we achieve relative errors in predicting next-day power production of 3–9% for favorable weather conditions.