If it would be relevant for the models (e.g. to prepare for prolonged cloudy periods where the net has to step in more), I suppose you could account for rooftop generation in your models by comparing energy consumption from the net on both cloudy and sunny days.