2 house climate dataset
Oliver Kennedy edited this page 2016-11-24 20:30:25 -05:00


Example Data Errors

Misattributed Readings

When the office sensor (id 2) was added it was initially misconfigured, and several of its first readings were reported as being from the armory sensor (id 1).

Sensor failures

Power failures or other errors have caused the sensors to stop producing readings until they were reset. Each sensor has intermittent periods that last up to 2 days where the sensor produces no readings. Other times, there's a failure in the sensor itself and it reports -999 degrees celsius.

Granularity change

Sometime between late October 2017 and early November 2017, I bumped up the sampling rate from every 30min to every 5min. This did not happen uniformly for all sensors --- the office sensor (id 2) ran for a little longer

Moire patterns

The furnace (and presumably also the AC unit) produces a distinctive spike pattern in the temperature trace (e.g., see Mid-Late November) as the heating unit turns on/off at the low/high water-mark temperatures. That is, there's a rapid temperature increase followed by a gradual cooling (resp., warming) period. Note that this pattern also manifests in the relative humidity.

The error occurs on colder days, when a full round of heating/cooling takes less than 5 minutes. As a result, the 5 minute sampling period may entirely miss one full heating cycle, and it appears as if the house is taking longer to cool off. Several examples of this behavior occur around Wed Nov 23 (e.g., at 10:20 EST). Note that these aren't strictly speaking data errors as the temperature readings are technically correct, but the data may still be misinterpreted.