To understand which weather variables most influence each target, we trained a Random Forest model on the full set of weather features from all five cities and extracted permutation-based feature importances. On the vertical axis are the weather variables by city. On the horizontal axis is the percentage of variation in the target explained by each variable.

Here we can see that the humidity in Valencia, Bilbao and Seville have a significant impact on overall electricity demand, as does the temperature and wind degree in Barcelona. Similarly, the wind speed in Madrid and pressure in Bilbao give us the greatest insight into the total onshore wind generation for Spain. This helps us identify which variables to look at when making evaluations.