For the creation of hourly load profiles for most European countries, ENTSO-E uses a load forecasting toolbox that incorporates uncertainty analysis under various climate conditions. The toolbox comes in a software application developed by an external provider.
The demand forecasting toolbox (DFT) combines historical load data, weather variables (such as temperature and irradiance), calendar effects (weekdays, weekends, holidays, and special days), and scenario-based structural adjustments to model future electricity consumption patterns.
In addition, DFT is able to integrate electrification drivers and demand-side flexibility assumptions, including the impact of:
- electric vehicles (EVs),
- heat pumps (HPs),
- air-conditioning growth,
- sanitary water heating,
- temperature-dependent demand growth
DFT contains three different algorithms for modelling the hourly electric load time series. These are popular machine learning algorithms that can provide precise estimates of future load curves. At the same time these methods can give interpretable results so the main drivers of load evolution can also be inferred. Currently available algorithms in DFT: Generalised Additive Models (GAM), Random Forest, Linear regression.
