Christine L from Minneapolis, MN. Xcel Energy territory. Working on a large hospital campus with 4 separately metered buildings all on the A14-TOU rate. I pulled 24 months of 15-minute interval data for all 4 meters and compared the on-peak vs off-peak energy allocation on the bills to what the interval data shows. Three of the four meters match perfectly. But Meter 3 — the surgery center — has a systematic error. The billing system is allocating about 12% more energy to the on-peak period than the interval data supports. Every single month for at least 2 years. The other three meters on the same rate with the same TOU windows are fine. Something is wrong specifically with how Meter 3 data is being processed.
Interval data reveals systematic TOU errors on a hospital campus
Christine, a systematic percentage error on one meter but not the others on the same rate is unusual. Possible causes: the meter itself could have a timing error similar to what Stuart described with the DST issue, or the channel configuration in the billing system could be mapping the wrong interval registers to the TOU periods, or there could be a data processing error specific to that meter account.
I have seen billing system channel errors before on Duquesne Light accounts. Each interval meter stores data in multiple channels — one for total consumption, one for on-peak, one for off-peak, sometimes one for each TOU period. If the billing system maps channel 3 (off-peak) to the on-peak calculation and channel 2 (on-peak) to off-peak, you get exactly this kind of systematic percentage error because the channels have different totals.
Walt, that channel mapping theory fits perfectly. If the on-peak and off-peak channels were swapped, the billing system would be charging on-peak rates for what is actually off-peak usage and vice versa. Since this hospital surgery center does more work during daytime (on-peak), swapping the channels would overstate on-peak consumption on the bill. Let me test this theory against the data.
Christine, a quick way to test: add up the total on-peak kWh from the interval data and compare to what the bill shows. Then do the same for off-peak. If the bill on-peak number matches the interval off-peak number (or is very close), the channels are swapped.
Mike, just ran that comparison. For January 2015: interval data shows 142,800 kWh on-peak and 98,400 kWh off-peak. The bill shows 141,200 kWh on-peak and 99,600 kWh off-peak. Close but not an exact match, so it is not a clean swap. The bill on-peak is slightly less than the actual on-peak. That does not fit the swap theory.
What if the issue is not a swap but a timing offset? If the meter TOU clock is off by 30-45 minutes, the on-peak period would capture some off-peak usage and vice versa. That would cause a percentage shift without being an exact swap. The magnitude of the shift would depend on what load is running during the transition minutes.
Yuri, a partial timing offset makes more sense than a full swap. If the meter clock is 30 minutes late, the billing system would assign the 8:30-9:00am interval to off-peak (meter thinks it is 8:00-8:30, before the 9am peak start) when it should be on-peak. And it would assign the 9:00-9:30pm interval to on-peak (meter thinks it is 8:30-9:00pm, before the 9pm peak end) when it should be off-peak. The net effect depends on the load during those transition periods.
Christine, request the raw interval data timestamps from Xcel for Meter 3 alongside the billing period TOU assignments. If the timestamps show a consistent offset from the other three meters, you have your answer. The surgery center meter clock is drifting and Xcel needs to recalibrate and rebill.
Randy, requested the raw timestamp data. Also asked Xcel to pull the TOU programming parameters for all 4 meters so we can compare. If Meter 3 has different clock settings than the other three, that proves the error.
Got the data back from Xcel. Meter 3 TOU clock is 38 minutes behind the other three meters. The programming timestamp shows it was last synchronized in 2009 — over 5 years ago. The other three meters were all synchronized in 2014. Meter 3 was simply missed during the last calibration cycle. 38 minutes of drift over 5 years is about 7.6 minutes per year which is within the expected range for quartz clock drift on older metering equipment.
38 minutes of clock drift — that is substantial. Depending on the load profile during those 38-minute windows at the start and end of each TOU period, the billing impact could be significant on a surgery center that ramps up in the morning and winds down in the evening.
Exactly Walt. I calculated the billing impact of the 38-minute offset across 24 months. The surgery center was overcharged $31,400 because more energy was allocated to on-peak than actually occurred during the peak window. Xcel verified the clock error, recalibrated the meter, and agreed to rebill the full 24 months. Credit to the hospital: $31,400. The facilities director told me this is the first time anyone has ever analyzed their interval data at this level of detail.
$31,400 on a 38-minute clock drift. Christine, this is exactly the kind of analysis that separates professional auditors from people who just compare bills to tariff rates. The interval data told the whole story. Outstanding work.
Thanks Randy. I am now checking the interval data timing on every TOU account I audit. If one meter in a campus can drift 38 minutes, others can too. This is going to be a standard part of my TOU audit methodology going forward.