Photo by American Public Power Association on Unsplash.
Search “demand side response” today and the entire first page is trying to sign you up. Drax, E.ON, Enel X, GridBeyond, NESO, gov.uk, the UK Parliament and the Association for Decentralised Energy all answer the same question: how does a building get paid for reducing grid demand on request. That’s a legitimate question, and each of those sources answers it well. It is also not the question most building operators are actually stuck on.
The question none of them answer comes earlier and is harder: which of my loads can I actually shed or shift, without breaking comfort, process, or a compliance obligation I don’t get to renegotiate? An aggregator can tell you what a flexibility contract pays. It cannot tell you, from your own building’s data, whether the chiller you’d nominate for a peak event will actually come off without a comfort complaint an hour later, or whether a load that looks flexible on a spec sheet is load-bearing in ways the spec sheet never shows.
That’s the gap this piece sits in. Not another walkthrough of what DSR is, NESO and gov.uk already own that ground, and there’s no reason to re-explain what they’ve already published well, but the operational question that has to be answered before any aggregator relationship makes sense in the first place.
What demand side response actually asks of a building
At the mechanism level, DSR is a commitment: reduce or shift consumption by a defined amount, within a defined window, when called, in exchange for payment or an avoided charge. The obligations attached to that commitment, metering accuracy, response-time verification, penalty exposure for a missed or partial event, are documented in detail by NESO’s flexibility markets material and by the aggregators selling participation. If what you need is the mechanics of enrollment, settlement windows, or which programme pays what this quarter, that’s their material to read directly, not ours to paraphrase.
What that material assumes, reasonably, is that you already know which loads you’re offering. That assumption is where most of the actual difficulty lives.
The loads that are genuinely flexible, and the ones that only look that way
A generator or a battery is unambiguously flexible. It either holds stored capacity to release or it doesn’t, and that number is known in advance. Almost nothing else in a commercial building is that clean.
HVAC precooling and setpoint flex is the example every DSR primer reaches for, and it’s often a real one, but only inside a band the building’s own thermal mass sets. Precool a floor two degrees below setpoint ahead of an afternoon event and you’re borrowing thermal inertia the building will need to give back over the following hour. Push the precool further than that buffer supports and you haven’t flexed a load. You’ve created a comfort complaint with a delay built in.
Chiller sequencing is a second real candidate. A lead-lag plant with genuine headroom on the lag machine can often absorb a shed on the lead unit without the space noticing, provided the staging deadband and the downstream thermal buffer are both sized to cover it. That’s a building-specific answer, not a category-wide one: the same plant configuration in a different building, on a tighter deadband or a smaller buffer tank, may have no headroom at all.
Non-critical process loads are the most consistently overlooked candidates, and often the best ones. A batch compressor with slack in its schedule, a pump running on a VFD well below rated duty, a refrigeration case whose defrost cycle can shift by minutes without risk. None of these show up on a typical DSR shortlist, because the shortlist usually starts and stops at HVAC.
On the other side of the line: life-safety systems, ventilation rates tied to occupancy or indoor-air-quality compliance, and refrigeration running near a food-safety threshold aren’t candidates. Neither is process equipment where an interruption costs more than any realistic flexibility payment could offset. The mistake worth naming directly is treating “large load” and “flexible load” as the same thing. A big chiller plant is a large number on a spec sheet. Whether it can actually come off the grid for forty minutes without consequence is a question about that specific building’s thermal buffer, that day’s occupancy, and that hour’s outdoor conditions, not about the machine’s nameplate rating.
How you’d actually know, from your own building’s data
This is where a walkthrough inspection runs out of answers. A facilities team can list candidate loads by eye. What eye-level inspection can’t do is tell you, for a specific event window on a specific day, whether shedding a given load will hold or fail, and by how much margin.
Anomaly detection is the first input worth having: knowing what a load’s normal operating pattern looks like well enough to notice when it’s already running outside that pattern. A chiller already running hot, or a compressor already short-cycling, is a worse candidate for a shed today than its historical profile suggests, and a shortlist built from history alone won’t catch that.
Load forecasting with confidence bounds is the second input. Not “this load typically draws X kW at 3pm” but a forecast with an error band honest about how much that estimate can actually be trusted, for this building, on this day, under this weather. A flexibility commitment made against a bare point estimate, with no stated uncertainty, is a commitment made blind.
What-if simulation is the third input, and the one that actually answers the operator’s question. FrostDynamics™, the physics-grounded model underneath Explore’s forecasts, exists to answer exactly this: if this load sheds for this window, what happens to comfort, to the next load in the sequence, to the building’s ability to recover before the next occupied hour. That’s a simulation run against the building’s own measured thermal and electrical behavior, not a rule of thumb carried over from a different building’s spec sheet. That forecasting work runs on the same energy management software Explore already provides for a building’s everyday cost and consumption picture; DSR readiness is one more question it answers, not a separate module bolted on for the occasion. It’s what turns “this looks flexible” into “this is flexible, by this much, under these conditions” before anyone commits to an event.
Where FrostLogic’s job ends and the aggregator’s begins
None of that is a market, and Explore doesn’t pretend otherwise. FrostLogic Explore does not trade flexibility, bid a load into a grid programme, or settle a DSR payment. It is a decision layer sitting upstream of that market, not a participant in it. The work of enrollment, of bidding into whichever programme fits a building’s profile, of verifying and settling an event, sits with the companies built to do exactly that: Drax, E.ON, Enel X, and GridBeyond among them, each running the aggregation and market side of DSR at a scale FrostLogic has no reason to try to duplicate.
The handoff is deliberately narrow. Explore’s job stops at producing an evidenced, ranked list of which loads in a given building are actually flexible, under what conditions, with what confidence, and what shedding each one costs operationally. What an aggregator does with that list, which programme to enroll it in, how to structure a bid, how settlement gets handled, is their domain, and better handled by a company built for exactly that than by a sensor-analytics platform improvising a trading desk it has no business running.
That division has a compliance edge worth naming directly. Participating in a DSR programme can intersect reporting obligations a building already carries, particularly where a shed event touches metered consumption that also feeds a sustainability or energy-performance disclosure. Where that overlap needs its own handling, it belongs alongside the rest of a building’s compliance work, not treated as a DSR-specific afterthought nobody owns.
Load flexibility, by type: what looks true and what holds up
Load type | Looks flexible on paper | Actually flexible (evidence-based) | What decides it |
|---|---|---|---|
HVAC setpoint / precooling | Yes, universally cited | Usually, within a band | Thermal mass, and how much inertia the building can lend back after the event |
Lead-lag chiller sequencing | Yes, if there is a second machine | Building-specific | Staging deadband width and downstream buffer capacity |
Non-critical process loads (VFD pumps, batch compressors) | Rarely considered | Often the best candidate | Schedule slack the process actually has, not its nameplate rating |
Refrigeration defrost timing | Sometimes | Yes, within minutes | Food-safety threshold and current case temperature |
Life-safety / IAQ-mandated ventilation | No | No | Regulatory floor, not a judgment call |
Critical process equipment (continuous, high cost of interruption) | No | No | Cost of interruption exceeds any realistic flexibility payment |
FAQ
What is demand side response?
Demand side response is a commitment to reduce or shift electricity consumption on request, in exchange for payment or an avoided charge, usually coordinated through an aggregator or directly with a grid operator. NESO and gov.uk publish the mechanics of UK flexibility markets in detail; this page focuses on the operational question underneath that commitment, which loads a building can actually offer.
Is FrostLogic a DSR aggregator?
No. FrostLogic Explore does not trade flexibility, bid loads into grid markets, or settle DSR payments. It identifies and evidences which loads in a building are actually flexible and forecasts the impact of shedding them. Enrollment, bidding, and settlement are handled by aggregators such as Drax, E.ON, Enel X, and GridBeyond.
How do I know which loads I can actually flex?
Start from the building’s own data rather than a generic candidate list. Anomaly detection shows whether a load is already running outside its normal pattern, load forecasting with confidence bounds estimates what a shed would actually save, and what-if simulation against the building’s physics shows what happens to comfort or the next load in sequence if you shed it. A load that looks flexible on a spec sheet isn’t confirmed flexible until it’s tested against the building’s own behavior.
Can HVAC precooling break comfort if I get it wrong?
Yes. Precooling borrows thermal inertia the building has to give back, and the safe band depends on that building’s thermal mass, occupancy, and the outdoor conditions on the day. Push a precool or setpoint flex further than the building’s buffer supports and the result is a comfort complaint, usually with a delay of an hour or more between the shed and the symptom.
What loads should never be offered to a DSR programme?
Life-safety systems, ventilation rates tied to occupancy or indoor-air-quality compliance, refrigeration operating near a food-safety threshold, and process equipment where interruption cost exceeds any realistic flexibility payment. These are not judgment calls; the regulatory or safety floor decides them before economics do.
Does participating in DSR affect my compliance reporting?
It can, where a shed event touches metered consumption that also feeds sustainability or energy-performance disclosures. That overlap is worth checking against a building’s existing compliance obligations rather than treating DSR participation as a purely commercial decision.
Does FrostLogic replace my aggregator relationship?
No. Explore produces the evidence, which loads are flexible, under what conditions, with what confidence, that makes an aggregator relationship more defensible. The aggregator still handles enrollment, market bidding, and settlement; Explore does not compete with that role or attempt to replicate it.
What data do I need to start identifying flexible loads?
Usually what a BMS, energy meters, and existing building sensors already collect: HVAC setpoints and zone temperatures, chiller and AHU load and sequencing data, and duty-cycle data for major process equipment. The gap is rarely instrumentation. It is usually that nobody has run a forecast or a what-if simulation against that data before committing a load to a DSR event.
FrostLogic Explore brings sensor intelligence, scenario simulation, and grounded-inference AI to commercial and industrial buildings. Learn more about Sensor Intelligence or talk it through with us.
Curious how this would look on your building?
What's your building not telling you?
Tell us what you're trying to figure out: energy drift, a BMS you don't trust, compliance you're chasing. We listen first, then tell you straight whether Explore helps. 30 or 60 minutes, your pick. No commitment either way.
