I'm analyzing interval data for a Green Mountain Power customer in Burlington and finding some readings that don't make physical sense. The 15-minute intervals show the facility going from 150 kW to 890 kW and back to 160 kW within a single hour, multiple times per day. The customer is a small office building with maybe 200 kW of total connected load. Either their meter is completely malfunctioning or there's some serious data corruption happening. Has anyone seen demand spikes this dramatic that turned out to be legitimate?
Load profile analysis showing impossible demand readings
Fiona, those numbers are definitely suspicious for a small office building. I've seen similar impossible spikes with smart meters that have communication errors or voltage sensor issues. The meter might be multiplying readings incorrectly due to a CT ratio problem. Have you verified the meter installation details - CT ratios, PT ratios, wiring configuration? Sometimes a simple wiring error can cause readings to be off by a factor of 5 or 10.
I work with Puget Sound Energy and we see this occasionally with newer AMI meters. Could be a firmware bug or data transmission error. The key question is whether these spikes show up consistently in the same time periods or if they're random. If they're random, it's likely a communication or processing error. If they're consistent, there might be some equipment you're not accounting for. Either way, those readings would trigger massive demand charges that no small office could actually be causing.
Vera raises a good point about consistency. Looking at the data more carefully, these spikes are completely random - different times, different days, no pattern whatsoever. And you're right about the demand charges - if these readings were accurate, the customer would be paying an extra $15,000+ per month. I'm definitely challenging this with GMP and demanding a meter investigation. The customer's actual usage patterns show very stable, predictable loads typical of office space.
Fiona, document everything and get GMP to provide you with the raw meter logs, not just the processed interval data. Sometimes the processing software introduces errors that don't exist in the actual meter readings. I had a similar case with AEP Ohio where their billing system was doubling certain intervals due to a database error. Once we got the raw data, it was obvious the meter was reading correctly but their processing was flawed.
This is exactly why we need to be skeptical of interval data, especially with newer smart meters. I've seen LG&E meters in Louisville report completely bogus readings for weeks before anyone notices. The automated billing systems just process whatever data they receive. Kevin's right about getting the raw logs - that's the only way to determine if it's a meter issue or a data processing problem. Don't let them brush this off as a one-time glitch.
I've been following this thread with interest because we're seeing more of these impossible readings with smart meter deployments. The problem is that utilities often don't have adequate validation checks in their billing systems. They assume the meter data is always correct and bill accordingly. PPL had to issue over $200,000 in credits last year for similar meter errors. Make sure you're documenting the financial impact - that gets their attention faster than technical arguments.
Update: GMP finally sent a technician to investigate the meter. They found a loose connection in the CT secondary wiring that was causing intermittent voltage spikes in the metering circuit. The meter was interpreting these as massive load increases and recording them as legitimate demand. They're replacing the meter and providing credits for the past 8 months of erroneous charges. Total credit amount: $47,300. This is why we do what we do, folks.
Fantastic outcome Fiona! A loose CT connection - that's a classic issue that can cause all sorts of bizarre readings. I'm glad you pushed GMP to investigate rather than just accepting their initial response. $47,300 is a significant recovery that probably wouldn't have happened without persistent interval data analysis. This case should be a lesson for all of us about the importance of questioning suspicious data patterns.
Great detective work Fiona! Loose CT connections are more common than people think, especially in older installations or after maintenance work. The fact that you recognized the readings were physically impossible for that facility size was key. Many auditors might have just accepted the utility's data at face value. This is a perfect example of why we need to understand the actual operations of our clients' facilities, not just look at numbers on a spreadsheet.
Excellent work Fiona. Cases like this make me wonder how many other customers are getting billed for phantom demand due to metering errors. Most small businesses would never notice or question their bills, especially with the complexity of interval data. It's up to professionals like us to catch these errors and protect our clients. That $47,300 recovery just paid for a lot of future auditing services!