PEV Submetering pilot data issues Nov/Dec 2015
Dear Pilot participant -
This page details the potential data issues with your submeter energy data that could affect your electrical bills for the months of November and December of 2015.
eMotorWerks server-side data processing software suffered an issue causing incorrect reporting of some charging data between October 26, 2015 and January 8, 2016
We estimate that this issue could result in 16-24% of the EV charging load during that period to be billed as general house load for some customers
We have received the first complaints from our customers in the second half of December last year as the customers were reviewing their prior cycle bills. Across the 3-5 initial complaints in December, participants have generally reported apparent overbilling by $20-30 / month.
Towards the end of December, following Christmas holidays, eWM has launched an investigation into this issue. In the first week of January, we have tracked down the issue to the change in the behavior of our SQL server following our server migration on October 26, 2015. Net effect of this change was a 24-hour shift of a fraction of 15-min intervals - e.g., charging data for Tue 3:00-3:15am interval were reported as Wed 3:00-3:15am interval.
The issue was then addressed via server-side fixes on January 8 & 9, with thorough testing during the following week.
Billing impact was largely contained due to high degree of similarity of the charging load for all days of the same type (e.g., weekdays are similar to other weekdays and weekends - to other weekends):
For Mon charging load (peaking 12am - 3am, reported as Tue load), no significant impact due to Tue actual load occurring at the same time of night and generally being higher than Mon load.
For Tue-Fri charging loads (peaking 12am-3am, reported as Wed - Sat load), no significant impact due to very similar load patterns for these days with respect to the next day's load. We estimate the impact at the 10-15% level due to occasional mismatches
For Sat charging load, there was no significant impact since Sat and Sun charging loads are quite similar, as well. We estimate the impact at the 10-15% level due to occasional mismatches
However, any Sun daytime charging load (reported as Mon daytime load) most likely got rejected as bad data and the energy got booked into the house load instead. Fortunately, Sun charging loads are typically only ~7-10% of the weekly load so the impact was limited to that level
Therefore, our assessment of the total impact for affected customers is 10-15% of 90% of the charging load (Mon-Sat) plus 7-10% of the weekly load that fell on Sun, for the total impact of 16-24%.
UPDATE May 2016:
Since our email notification to all pilot participant, ~25 customers reported potential discrepancies in their bills for Nov / Dec 2015. For every such report, we have completed a thorough investigation using the customer billing data (provided by customers) and our archived second-by-second submetering data. The resulting reimbursements were calculated to average at $23.04 for the period of two months in Nov / Dec 2015.