If, for example, the ”Auxiliary – A” model is used, the adjustments for the weather dependencies will be estimated for the same season as the forecasted period but by using about one year old data. The adaption, which is optional, makes a test forecast for the forecasted series for the last period where data is available and then compare this to the outcome. An adjustment is made so that the final equation can be expressed as
F=αA(αQA+βW+c)+cA
In the equation, αA and cA denotes the adaption coefficient and adaption constant respectively.
Adaption is turned on by the ”Use adaption” checkbox. The length for the test forecast is set in ”Length (h)”. ”Tolerance” specifies how large deviation between the test forecast and the outcome that is tolerated, i.e. if abs((outcome-forecast)/outcome) > tolerance, this outcome value will be ignored.
”Min correlation” specifies a threshold value for the correlation between the corrected test forecast and the outcome after the applied adaption. If the correlation is below this value, the user will get a warning message.
The check box ”Separate weekdays/weekends” denotes if the system will have separate adaptions for working days and weekend/holidays respectively.