You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
When program runs over midnight stations that run before midnight use the weather data from the last day but the stations that run after midnight only use weather data after midnight. For some locations there will be significant weather differences during the day (humidity and temperature for example). This results in very different run times for stations that are open one after another in the same program. For me for example the stations before midnight opened on 90% and the remaining stations open on 47%.
Possible solution: Use 24h moving window instead of midnight cutoff for calculating run times. So when calculating runtimes consider the weather data for up to 24 hours before the run instead of the weather data from the calendar day.
Possible (even better) solution: Use all weather data (with some time cap) since last time the same station has run. This will ensure more accurate run estimation especially when stations are run on few days cadence instead of daily (a common scenario).
The text was updated successfully, but these errors were encountered:
When program runs over midnight stations that run before midnight use the weather data from the last day but the stations that run after midnight only use weather data after midnight. For some locations there will be significant weather differences during the day (humidity and temperature for example). This results in very different run times for stations that are open one after another in the same program. For me for example the stations before midnight opened on 90% and the remaining stations open on 47%.
Possible solution: Use 24h moving window instead of midnight cutoff for calculating run times. So when calculating runtimes consider the weather data for up to 24 hours before the run instead of the weather data from the calendar day.
Possible (even better) solution: Use all weather data (with some time cap) since last time the same station has run. This will ensure more accurate run estimation especially when stations are run on few days cadence instead of daily (a common scenario).
The text was updated successfully, but these errors were encountered: