Specify the maximum number of predictors in “Max model size”. If you enter the value:
-1 – you will get a forecast based directly on the day type pattern. The program will skip the regression completely. If “none day type” is selected, the last specified predictor will be used as forecast values.
0 – you will get a constant forecast with the mean load during the base period as constant forecast value.
1 – Only one explanatory value is used. Since the hourly day type pattern for the load is generally the most significant predictor, the model will usually select this predictor. If you want to exclude explicitly weather parameters from the model, you must however use one of the models “Weather independent – A” or “Weather independent – C” (see below).
2 – The next most important variable after the social load is generally the air temperature – often expressed as a moving average of the temperature during the last day or days. With model size 2, generally only the day type pattern plus a temperature variable will be included.
If you set the value to more than two, the result of the automatic procedure will result in different model selections depending on the on the type of load series and the current time of year. During March, for instance, global radiation may be incredibly significant on cold days with strong sunshine, while the average hourly temperature may be more important on cold days in November and December.