The statistical component of Achelous assumes that the output from a power plant, P, at hour t of the forecast can be written as
P(t)= αMPt + βH(t) + γ
Where αMPt = average forecasted production at hour t and βH(t) estimates the sensitivity for differences in runoff, H.
The model training period for the statistical part of the Achelous model is set in Reference period and the Training period field.
In the Achelous model setting “Reference period”-field the length of the reference period, Length (d), is determined. The reference period length indicates number of days in historic production values to search for past similar intervals. That is; similar intervals in terms of increasing or decreasing production for that period.
The number of hours compared when searching for past intervals similar to the interval preceding the forecast period is set in the “Compare (h)” field. In order to statistically determine if the production is in a rising or downward state all sub-intervals of the length “Compare (h)” are compared with the period that immediately precedes the forecast.