During the regression and the forecast period, data is obtained from the day type pattern based on the following priority:
1. First choice selection
2. Second choice selection
For the regression: If none of these day types are found, the time in question is ignored.
For a forecast: If the second choice selection is used, the user receives a warning. If neither the first choice selection nor the second choice selection is found for a date, the program tries to redo the entire forecast process based on the second choice selection. If this also fails, the program will redo the entire forecast process a third time based on the weekly pattern.
The estimated models are saved in the memory, which means that if you use the same observation series for several weather station locations, no new model is calculated. This can reduce calculation time by 70-80% if you use multiple weather forecast modules. The observation series can be evaluated by producing long-term forecasts of observed weather and comparing them with the actual outcome.
An enhanced data check can be performed if you have entered values in “Max relative delta”, “Max hours to ignore” and “Max relative mean diff”.
3. “Max relative delta” indicates the amount of relative change that is to be accepted from one value to another. If a value is found that differs more than the threshold value, the program steps forward and sees whether the next values also differ more than permitted. If the number of immediately subsequent values that differ more than permitted from the last “accepted” value is fewer than or equal in number to “Max hours to ignore”, these are ignored during the forecasting. Otherwise a change is deemed to have occurred in the load level and the user will receive a warning to that effect.
4. During the day type estimation, the program first calculates the mean load for all data used in the year in question, or, if the year in question is not to be used, the previous year. This mean load is saved as a reference load. The mean load is then calculated for each year’s data that might possible be used in the day type estimation. If this differs by more than “Max relative mean diff” from the reference mean, no data from the year in question or further back in time is used. This is to avoid the use of non-representative data.