I have the following problem:
Imagine we are at time t, and we have a forecast model who predicts the power production of a wind farm for the next hour (i.e at t+1), and this procedure is repeated hourly.
At time t, we already know the real value of power production at t-2, t-3, t-4, ...(but for some reason not at t-1!), hence we have information about the previous errors committed by our model.
Do you know some kind of algorithm (control system, filter, anything) able to modify and improve our predictions knowing that previous information (errors committed in the past?
Thanks!
The Kalman filter is used for filtering (removing noise from current observations), predicting future observations, and smoothing (removing noise from past observations). It is the tool of choice for what you are trying to achieve, at the very least the 1st technique to be tried.