We've noticed that for some EnaSolar inverters, the inverter reports what appears to be an anomaly in the accumulating kWattHoursToday value. The anomaly is that the reported value goes down from the previous reading. For example, we've seen the following data captured:

select id,local_created,watts,watt_hour 
from solarnet.sn_power_datum 
where node_id = 36 and source_id = '4' and local_created between '2011-12-15 14:17:00'::timestamp and '2011-12-15 14:23'::timestamp
order by local_created;

   id    |      local_created      | watts | watt_hour 
---------+-------------------------+-------+-----------
 5281189 | 2011-12-15 14:17:00.772 |   807 |      5920
 5281199 | 2011-12-15 14:18:00.819 |   976 |      5940
 5281220 | 2011-12-15 14:21:01.854 |     0 |      5760
 5281228 | 2011-12-15 14:22:04.473 |    16 |      5950
(4 rows)

Notice the watt_hour value in row 2 goes from 5940 down to 5760 in row 3, then back up to 5950 in row 4. The backwards drop of the watt_hour value causes the aggregation query to assume the accumulating watt_hour value has reset, so it treats the total energy for that time period as 5760 Wh. This is based on the assumption that accumulating watt_hour values are reset by the inverter each day, and thus the first value that is lower than the previously read value must be a "new day" and to treat the reading as the first reading of the day, assuming it started over from zero.

We must tweak our algorithm to work around this, and inquire from EnaSolar a possible reason for this scenario to see if this is expected.