Size a demand response management system to your metered, controllable load at the worst 15-minute interval of the year — then add 15-20% for growth. Not to your connected load. Not to the peak on your utility bill. I've reviewed enough of these specs to tell you the gap is usually 30-50% of the system cost, and close to half the time it's in the wrong direction: overbuilt, oversold, and quietly wasting money.
That's the short answer. Here's why it matters, how to get the number right, and where the method breaks down.
Why the Wrong Number Is Expensive
I'm the quality and compliance manager at an electrical equipment company — GridSense, in our case — and part of my job is reviewing every DRMS and energy management system spec that goes out before it ships. Four years of doing this, roughly 200 specs a year. In Q1 2024 we reviewed 14 demand-response proposals. Two got rejected outright, and both times for the same reason: the "controllable load" figure had been pulled from a three-year-old single-line diagram, not from metered data.
Here's the thing: connected load and controllable load are not the same number. They're not even close. I learned that one the hard way.
We had a facility with 2,000 kVA of total connected load. The spec assumed controllable load was about 1,800 kVA — "everyone knows you can switch most of it off." When we actually metered the switchable branches for 12 months, the real figure was closer to 1,100 kVA. That's a 39% miss, and it meant the DRMS was sized to respond to a load the site could never deliver on its worst day.
That mistake was ours, and it cost the customer roughly $18,000 in unnecessary hardware plus a failed utility enrollment. After that we stopped accepting any connected-load-derived figure on a spec over 500 kVA.
Everyone warned me to validate against metered intervals before signing off on a size. I only believed them after that rejection. Classic lesson learned at full price.
Why does the number matter so much? Under FERC Order 745 (2011), demand response resources can be compensated in wholesale markets — but verification happens at the 15-minute interval. Any gap between what you promised and what you actually delivered (or failed to deliver) gets reconciled against you, and the penalties claw back far more than the oversized hardware ever cost.
How to Actually Size It
Four steps, in this order.
1. Define what you're actually allowed to curtail
Not everything you can switch off counts. If your HVAC is serving a cleanroom, if your compressor feeds a process that can't tolerate a restart, if the site carries a medical or food-safety load — that capacity doesn't count toward demand response, no matter what the one-line diagram says.
Ask two questions per load: can it be interrupted, and will operations actually let it be interrupted on a hot Tuesday in July? Most overbuilt DRMS specs fail the second question.
2. Measure by interval, not by month
Monthly bills tell you almost nothing useful here. You need 15-minute interval data — preferably 12 months of it, ideally from the utility meter or a dedicated power quality analyzer. Then find the worst 15-minute interval of the year for the controllable subset only. That's your floor. Size to that. Not to the annual average, not to the single peak day that happened to land on a production shutdown.
If you're specifying the monitoring side, make sure the platform can export clean interval data and timestamp it against the utility's clock. Two of the rejections I've handled came from clock drift between the meter and the DRMS head-end — the interval math was correct, but the intervals didn't line up with the utility's dispatch window. On the GridSense energy monitoring IoT platform (or any equivalent), insist on NTP-synced timestamps in the acceptance test. It sounds pedantic. It isn't.
3. Add a margin, not a fudge factor
Add 15-20% to the measured controllable load for growth and measurement uncertainty. That's it. If a vendor is quoting 50%+ headroom "for safety," ask them to show you the metered intervals that justify it. In my experience, most of that extra capacity gets sold at full price and sits idle for the life of the asset. That's not resilience. That's a sales margin dressed up as engineering.
4. Confirm the switchable loads are power-quality-friendly
This is the step people skip. Sizing for capacity is meaningless if your largest controllable motor drags voltage sag below the IEEE 519 thresholds every time the contactor opens. In Q3 2024 we had a customer whose DRMS worked exactly as specified — capacity matched, dispatch matched — but the site's power quality degraded enough that operations shut the program down after two events. That's an $11,000 control system doing nothing.
Before you sign off on size, run a short power quality assessment on the switchable branches. Check THD, check voltage sag on contactor open, check inrush on restart. If any of that is marginal, you've found your real constraint — and the DRMS size is smaller than the math suggested.
Small Sites Get Sized Right Too
Worth saying clearly: this method doesn't change for small facilities. A 400 kVA workshop deserves the same measured-interval treatment as a 40 MW plant. Maybe with fewer data points. Same logic.
Small doesn't mean unimportant — it means potential. I've watched small customers get pushed toward oversized systems because they couldn't argue with the sales sheet, and I've watched a few others get the right size and stay with that supplier for years. When I was starting out in this role, the vendors who took my $2,000 specs seriously are the ones I still use for $200,000 ones. If a vendor won't put half a day of engineering into a modest project, that tells you everything about how they'll behave when the project is bigger.
No DRMS is so small that it doesn't deserve a correctly sized spec.
When This Method Doesn't Apply
Honest caveats, because the method has edges.
You don't have interval data yet. Then you're not sizing, you're estimating — and you should say so on the spec. Install a power quality analyzer for 30-90 days first. Two months of rental data is cheaper than being wrong by 40% on a permanent system.
Your program is a capacity market, not a per-event one. If your utility or RTO runs a committed curtailment model instead of an event-based one, the obligation is a firm MW figure at dispatch, not an annual energy total. The margin math changes. Read the tariff before you finalize the number.
The site is single-process. One large load doing one job — an electric furnace, a single large chiller serving one loop — doesn't leave you much to curtail. DRMS might not be the right investment at all. There's no shame in that answer, and I'd rather say it upfront than sell you capacity you can't use.
To be fair, there are markets where oversized DRMS still pencils out, because the enrollment incentive is generous enough to cover the extra hardware. That's a financial calculation, not an engineering one. Don't let the incentive talk you out of a correct size.
Prices, tariffs, and program rules shift — verify the current figures with your utility and your regional transmission operator before committing. The sizing logic won't move. The financials will.