Cooling is frequently the largest single contributor to a summer afternoon peak, and it has a property most loads do not: the service can be produced at one time and consumed at another.
That is the whole idea. Run the chillers at night, store the result as chilled water or ice, and draw on the store during the expensive hours. The building is cooled exactly as before. The electricity that cools it has moved.
What moves and what does not
The cooling energy delivered is unchanged. The electrical energy consumed usually rises a little, because storage is not free: a chiller making ice runs at a lower evaporating temperature and therefore at lower efficiency, and a tank has standing losses.
So the honest description is: thermal storage trades a small increase in kilowatt-hours for a large reduction in peak kilowatts. Whether that trade is worth making depends entirely on the ratio between your demand rate and your energy rate — which is why the tariff has to be read before the equipment is specified. The distinction is the one set out in load shifting is not energy efficiency.
Full storage against partial storage
Full storage produces the entire day's cooling overnight and runs no chiller at all during the peak period. It gives the largest demand reduction and needs the largest tank and the largest night-time chiller capacity.
Partial storage is the common retrofit. The tank is sized to shave the peak rather than to eliminate daytime chiller operation entirely. Chillers run during the day at a reduced, level output while the store covers the difference. It costs far less, uses a smaller tank, and captures most of the demand benefit because the demand charge prices only the height of the curve.
For a site whose objective is the demand charge rather than time-of-use arbitrage, partial storage is usually the better economic answer.
The sizing question
Sizing follows from the load shape, not from the cooling load. The relevant quantity is how many kilowatts have to be removed during how many hours to bring the peak down to the target.
Sizing partial ice storage against a demand target
A site shaving an afternoon cooling peak.
- Site peak, unmanaged1,480 kW
- Demand target1,180 kW
- (Reduction required)300 kW
- Chiller electrical input contributing to the peak420 kW
- Chiller performance, cooling delivered per kW input3.2 kW cooling per kW
- (Cooling that must come from the store: 300 × 3.2)960 kW cooling
- Duration the peak must be covered5 hours
- (Storage capacity needed: 960 × 5)4,800 kWh cooling
Store size, expressed in ton-hours1,365 ton-hours
The number that matters is the duration, not just the reduction. A store sized for the right kilowatts and the wrong number of hours runs out before the peak window closes and the peak reappears. Figures illustrative.
The duration point is the one that catches retrofits. A store that covers four hours of a five-hour window leaves an hour in which the chillers restart and the demand determinant is set anyway. The saving is then close to zero despite the equipment working exactly as specified.
Where it fits and where it does not
Good fit: a large, predictable cooling peak; a defined on-peak window; a demand-heavy tariff, or a time-of-use tariff with a wide differential; space for a tank; night-time hours where load is genuinely low.
Poor fit: a site whose peak is not cooling-driven; a peak that occurs at unpredictable times; a facility that already runs its chillers at night for process reasons, leaving no night-time headroom; a tariff that measures facility demand across all hours, where moving the load to the night may simply relocate the peak rather than remove it. That last case is worth checking carefully before committing capital: facility, on-peak and billing demand.
The comparison with a battery
Both store energy to shave a peak. They differ in a way that decides most projects.
| Thermal storage | Battery | |
|---|---|---|
| What it can shave | Only the cooling portion of the peak | Any load |
| Round-trip cost | Lower chiller efficiency, standing losses | Round-trip electrical losses |
| Degradation | Very slow; a tank is a tank | Capacity fades with cycles and age |
| Footprint | Large tank | Comparatively compact |
| Responds to a short notice event | Slowly, needs planning | Instantly |
| Suits a coincident peak program | Poorly | Well |
Thermal storage is cheaper per unit of peak shaved where the peak is cooling. A battery is more flexible and better suited to short, unpredictable, called events — which is why sites facing a coincident peak obligation usually end up looking at batteries instead: sizing a battery for peak shaving.
They are not mutually exclusive. A site with a large cooling peak and a coincident obligation may reasonably do both, thermal storage for the routine monthly shave and a smaller battery for called events.
Before commissioning a design
- Establish from interval data how much of the peak is cooling, not how much of the annual energy is.
- Establish the duration that must be covered, using the widest peak window in the data rather than the average.
- Confirm which demand determinants the tariff bills, and model the shifted load against each.
- Check that night-time load has headroom for the charging period — a store that recharges into a new night-time peak has moved the problem.
- Include the efficiency penalty and standing losses in the energy line of the model, not as a footnote.
- Check whether a utility incentive applies. Load management measures are frequently funded differently from efficiency measures: utility incentive programs.
- Confirm the design day. A partial store sized to the average summer day will be short on the day that sets the ratchet.
The last item is worth capital. Under a ratchet, the day the store runs out is the day that prices the following eleven months, and the incremental cost of an extra hour of capacity is small compared with what a set floor costs: ratchet clauses.
