Instead of taking the same proportion at each step, proportion could be based on variance of residuals. Periods of day with more reliable predictions can take in more solar as there is less risk. Same idea could probably apply to reducing the peak - focus on more confident high peak times, perhaps?
I'm not sure if this will work or not. Need to think about it more.
Instead of taking the same proportion at each step, proportion could be based on variance of residuals. Periods of day with more reliable predictions can take in more solar as there is less risk. Same idea could probably apply to reducing the peak - focus on more confident high peak times, perhaps?
I'm not sure if this will work or not. Need to think about it more.