Batteries are the most flexible peak reduction measure available. They do not care what the load is, only how large it is and how long it lasts, which is why they can address peaks that no amount of scheduling will touch.
They are also the most capital-intensive, and the sizing exercise is where the economics are decided. Two numbers have to be right, and they are independent.
Power and energy are separate specifications
Power rating (kW) is how much the battery can deliver at any instant. It sets how deep a peak you can shave.
Energy capacity (kWh) is how much it holds. It sets how long you can sustain that depth.
A 500 kW / 500 kWh battery can shave 500 kW for one hour, or 250 kW for two. A 500 kW / 2,000 kWh battery can shave 500 kW for four. They cost very different amounts, and choosing between them requires knowing the shape of the peak, not just its height.
The failure mode is specific and expensive: a battery with the right power rating and insufficient energy capacity runs out before the interval that sets the charge has finished. The peak is then recorded anyway, the demand charge is unchanged, and the capital has been spent.
Read the shape, not the height
The input is the load duration curve: every interval in the year, ranked from largest to smallest. Two questions come off it.
How quickly does the curve fall away from the maximum? If the highest interval is far above the tenth-highest, a small amount of energy capacity captures most of the value. If the top hundred intervals are all within a narrow band, shaving the peak means sustaining the reduction for a long time, and the energy requirement rises sharply.
How long does the load stay above the intended target on the worst day? That duration, not the average, sets the energy capacity. Sizing to a typical day and hoping the worst one is similar is how a battery underperforms in exactly the month that matters.
The calculation
Sizing a battery against a monthly demand target
Worst day in the historical data, not the average day.
- Site peak, unmanaged1,850 kW
- Demand target1,500 kW
- (Required discharge power)350 kW
- Longest continuous period above the target, worst day3.2 hours
- (Energy delivered during that period)1,120 kWh
- Round-trip efficiency88%
- (Energy that must be drawn to store it)1,273 kWh
- Usable depth of discharge90%
- (Nameplate capacity at beginning of life)1,414 kWh
- Capacity retention assumed at end of design life80%
- (Nameplate required so the duty is still met at end of life)1,768 kWh
Specification350 kW / 1,770 kWh
The energy figure grew by 58% between the raw duty and the final specification, entirely through efficiency, depth of discharge and degradation. Omitting those three is the most common sizing error there is. Figures illustrative.
Three multipliers, none of them optional, and every one of them compounds. A specification that quotes 1,120 kWh because that is the energy above the target is describing a battery that will fail to do the job on the day it is bought, let alone in year ten.
Choosing the target
Deeper shaving costs disproportionately more, because the load duration curve steepens as you go down it. Each additional kilowatt of reduction requires more hours of coverage, and hours are energy, and energy is the expensive axis.
The economically correct target is not the deepest achievable one. It is the point where the marginal cost of another kilowatt of reduction exceeds its annual value. Testing three or four candidate targets against the same data, and pricing each, is a half-day exercise that routinely changes the specification and the cost.
Which duty you are sizing for
The three common duties produce very different batteries, and a site facing more than one has to decide which governs.
| Duty | Power driven by | Energy driven by | Frequency |
|---|---|---|---|
| Monthly non-coincident demand | Depth below the monthly peak | Duration above target on the worst day | Continuous readiness, every month |
| Coincident peak or capacity obligation | Reduction needed during called hours | Length of the event window | A handful of events per season |
| Time-of-use energy arbitrage | Load during the expensive period | Whole on-peak window | Daily |
A battery sized for a handful of long summer events is a different machine from one sized for daily readiness against a monthly maximum. Sites in coincident regimes frequently find the second duty far cheaper to serve, because it is a known number of hours in a known season: coincident and non-coincident demand.
Before the specification is issued
- Exhaust the cheaper rungs first. Sizing a battery against a peak that staggered startup would have removed makes the battery larger and the payback longer: staggered startup.
- Size against the worst day in at least twelve months of interval data, not the average day.
- Apply round-trip efficiency, usable depth of discharge and end-of-life retention, in that order.
- Confirm the recharge window. A battery that recharges into a new peak has moved the problem, and the recharge itself is a load that has to fit somewhere.
- Model every demand determinant the tariff bills, separately, against the shaved series: facility, on-peak and billing demand.
- Decide what happens when the battery is unavailable. Under a ratchet, a single uncovered excursion sets a floor and undoes a year of correct operation: ratchet clauses.
- Take the sizing into a full appraisal rather than a payback figure: total cost of ownership for a behind-the-meter battery.
The point about unavailability deserves more weight than it usually receives. A battery that covers 95 percent of events at a site with a ratchet is not delivering 95 percent of the value; the remaining 5 percent sets a floor that persists for months. That argues for conservative targets, for a maintenance regime that keeps availability high, and for a fallback plan on the days it is out of service.
