The installed cost of a battery system is a single, visible, negotiable number, and it dominates the conversation for exactly that reason.
It is not what determines whether the project works. A fifteen-year case is decided by things that happen slowly: capacity fading, availability slipping, an augmentation event nobody budgeted for, and a disposal obligation that arrives after the spreadsheet ends.
The lines a complete model contains
Capital
- The battery modules, the power conversion system and the enclosure.
- Electrical works: switchgear, protection, interconnection.
- Civil works: foundation, containment, access.
- Fire detection and suppression, and whatever the local code requires.
- Engineering, permitting and commissioning.
- Controls integration with the site's existing systems.
Operating
- Maintenance and the service contract.
- Thermal management, which consumes energy continuously.
- Round-trip losses: the energy drawn to store is more than the energy returned.
- Monitoring, and the human time to review it.
- Insurance.
Mid-life
- Augmentation or module replacement to restore capacity.
- Power conversion equipment replacement, which frequently has a shorter life than the battery.
End of life
- Decommissioning, removal and transport.
- Recycling or disposal.
- Any residual value, which should be treated conservatively.
Offsets
- Tax treatment of the capital: the investment tax credit and depreciation on energy equipment.
- Utility incentives: utility incentive programs.
- Program revenue where the asset can be sold into demand response: demand response programs.
Degradation is the line that decides it
A battery does not deliver its nameplate capacity for its whole life. Capacity fades with cycling and with calendar age, and the rate depends on chemistry, depth of discharge, frequency of cycling and temperature.
For a peak shaving duty this has a specific and awkward consequence. The battery must deliver a given reduction for a given duration on the worst day of every month, including the last month of its life. So the requirement is set at end of life, and everything before that is surplus.
Two ways to handle it, and a model should say explicitly which it assumes:
Oversize at the start. Buy enough capacity that end-of-life retention still meets the duty. Simple, and it means paying up front for capacity that does nothing for a decade.
Augment mid-life. Buy closer to the requirement and add capacity later. Lower initial capital, at the cost of a mid-life capital event that has to be budgeted, scheduled and physically accommodated.
A model that assumes constant capacity for fifteen years with no augmentation line is doing neither, and is wrong.
Availability, and why it is not a percentage
A battery out of service on the day of the monthly peak saves nothing that month. If a ratchet applies, it does considerably worse than nothing: the peak it failed to cover sets a floor that persists for the rest of the look-back period.
So availability does not scale linearly with value. A system available 95 percent of the time does not deliver 95 percent of the benefit; it delivers the benefit in the months where its unavailability happened to miss the peak, and much less in the ones where it did not.
The practical response is threefold: a maintenance regime designed around the demand cycle rather than around convenience, a conservative target that leaves headroom, and a written fallback for the days the system is out. It is also an argument against sizing the battery to exactly cover the historical worst day: sizing a battery for peak shaving.
A comparison worth making
The instructive exercise is not battery against nothing. It is battery against the alternatives that produce the same avoided kilowatt.
| Measure | Capital intensity | Ongoing cost | Life | Degrades? | Also useful for |
|---|---|---|---|---|---|
| Startup sequencing | Negligible | Negligible | Indefinite | No | Equipment stress |
| Demand limiting controls | Low | Low | 10–15 years | No | Visibility of load |
| Thermal storage | Moderate to high | Low | Long | Barely | Nothing else |
| Battery | High | Moderate | 10–15 years | Yes | Backup, response, arbitrage |
The final column is the honest argument for storage. A battery is expensive per avoided kilowatt and it is the only one of the four that can also provide backup, respond to a called event within seconds, and arbitrage a time-of-use spread. Where several of those are worth something, the stacked value is what makes the case; where only peak shaving is needed, cheaper measures usually win.
The caution on stacking is that the benefits are not automatically additive. A battery discharged for a called event may be unavailable for the site's own peak the same afternoon, and contract terms govern whether one reduction can be claimed twice. Model the stack with the conflicts in it.
Sensitivities that matter
Test at least these, because each has changed enough in recent years to matter:
- The demand rate falls in a rate case. The entire benefit is a function of a regulated rate.
- Degradation is faster than warranted under the actual duty cycle.
- Availability is lower than assumed, with the ratchet interaction applied.
- The load profile changes, removing the peak the battery was bought to shave.
- The tax or incentive treatment changes between approval and commissioning.
A case that survives all five is a case worth funding. One that depends on all five going well is a case that should say so.
What to insist on from a supplier
The warranted capacity retention for your duty cycle, not a generic curve. What operating conditions void it. What augmentation would cost and whether the design accommodates it. The efficiency figure and whether it is measured at the battery terminals or at the point of connection — the difference includes the conversion losses and is not trivial. And the assumptions behind any savings projection they provide, so you can substitute your own avoided cost: what a kilowatt of avoided peak is actually worth.
Then take the whole thing into an appraisal rather than a payback figure: building the business case for demand reduction.
