Case studies of extreme values of residual load.

The residual load is defined by energy demand minus energy production from renewable sources (wind and solar power). High values of residual load correspond to a high energy demand that cannot be covered by renewable energies and must be compensated with fossil energies. On the opposite, low/negative values of residual load correspond to a surplus of renewable energies with a too low demand. In this case, energy must be exported, or wind turbines and solar panels curtailed.

We compute the residual load using a demand model and a solar and wind power model. For the solar power model, we use the CAMS solar energy dataset interpolated to the position of all single solar panels installed in Germany and model the solar power output with the PVlib Python package. The wind power model is using ERA5 windspeed and air density interpolated to turbine location and turbine hub height to be later fed in the power curve of all single turbines in Germany. The demand model is a linear regression of 2m-temperature from ERA5, industry production index, day-of-week, public holidays and installed solar capacity. The residual load is then given by: ResLoad = Demand - (Wind + Solar)

We look at both tails of the distribution of the residual load. High values of residual load are mostly observed in winter when temperatures and renewable energy production are low (mostly during blocked regimes). These events correspond to high energy prices. Negative values of residual load are mostly observed on Sundays, public holidays and in summer, where demand is low and temperatures and solar and wind power production are high. They usually correspond to negative energy prices. Both events described above are critical for the energy industry. Skillful forecasts of these events in the sub-seasonal range help to reduce stress to the energy market. The results presented herein lay the foundation for a more comprehensive, statistical analysis of the predictability characteristics of residual load extremes.