Make filtering more robust to inputs on varying indexes
mdeceglie opened this issue · 2 comments
mdeceglie commented
The existing filtering framework, including that in TrendAnalysis doesn't work very well if filters or their inputs are on different date time indexes from energy_normalized
, for example, irradiance is 30 minute, and energy_normalized
is 15 minute. It would be good to provide some functionality to handle these cases.
mdeceglie commented
@kanderso-nrel and I talked through the following thought experiment about desired behavior for different use cases of TrendAnalysis
when interp_freq=None
PV frequency | weather frequency | behavior |
---|---|---|
15 min | 15 min | existing |
15 min | hourly | interpolate weather to pv index, then filter |
hourly | 15 min | all contributing weather has to pass filter |
weekly | hourly | don't filter based on weather (turns on when PV is daily or lower freq) |
hourly | weekly | raise error and don't do analysis (turns on when weather is daily or lower freq) |
mdeceglie commented
Capturing the outcome of a phone call on this one: For the time being in TrendAnalysis
let's always interpolate onto the pv index, either with or without regularizing PV (depends on whether interp_freq
is set)