Detect passing valves

[carlosduarteroa]

No description provided.

[praftery]

@carlosduarteroa - Misc comments from reviewing the code

1. The calculation of long_t should account for the transient behavior before calculating. That is, only values X minutes after the valve has closed should contribute to the average value of long_t. As currently calculated, long_t will be skewed high due to heat remaining in the coil after the valve has closed. here.

2. Similarly, I don't think you need to identify the bad valves solely using the temp delta above the long term average delta T when valve is closed (i.e. long_t ). see here. If long_t is high just by itself, it could be a perpetually passing valve. e.g. if long_t   is >10F, it seems certain that either that valve has an issue, or there is some other major issue with that VAV box or sensor.

3. There should be some mechanism for checking whether the AHU serving the reheat coil was running to ensure there was airflow through the VAV box. Otherwise there will be lots of erroneous values where the temp difference is high and the valve is closed because the DAT is reading the return plenum temp, and the AHU SAT is reading whatever that sensor has decayed away to (and they are often on a rooftop, relatively exposed to outside conditions). This check could be done using the airflow sensor at the box, or some fan/VFD/enable indicator at the AHU. In the absence of that info, maybe a time of day filtyer on the data would do a reasonably good job (e.g. subset so only data Mon-Fri 9-5 is used in one of the analyses).

4. More of a feature_request than a bug. If you end up pulling in the zone airflow sensor, you could estimate the reheat energy cost of the passing valve. i.e. do a modified heat balance where you subtract a reasonable long_t from the temp difference used in the heat balance. either assume some threshold for a reasonable long_t due to heat gain, fan gain, etc., or calculate an average value from the 'good valves' served by the same AHU.

5. I think it would be better if the denominator for bad_ratio was the total number of records where the valve was closed, vs the the total number of records here. We can't say too much about passing valves when the valve is open, so it seems less helpful to include that in the denominator.