Sizing a commercial battery

There is no standard size for a commercial battery. The right specification falls out of what you are asking it to do. A battery that shaves a half-hour demand peak is sized on power and how long that peak lasts. A battery that shifts cheap overnight units into the expensive day, or stores your own solar, is sized on energy. A battery that keeps a critical process alive through a grid failure is sized on the load it must carry and the hours it must carry it for.

Get the use case wrong and the sizing is wrong, however precise the maths. So the order is fixed: establish the job, read the load, model it, then specify the battery. We do not start from a product and work backwards.

The first split is between sizing on power and sizing on energy, because a battery is rated for both and they are not interchangeable.

Peak-shaving and standby for short demand spikes are power problems. What matters is the kW or kVA the battery can deliver at the moment your site peaks, and for how long that peak holds. A tall, brief spike needs power, not a large store. Arbitrage, time-of-use shifting and solar self-consumption are energy problems. What matters is the kWh you can move from one part of the day to another, so the battery has to hold enough to cover the window you are trying to shift.

Most commercial sites want more than one of these at once. The battery is then sized so the power rating clears the demand job and the usable energy clears the shifting job, with the binding constraint setting the specification.

A commercial battery is quoted as two numbers, and the second is not implied by the first. The power rating in kW or kVA sets how hard the cabinet can push or pull at any instant. The energy in kWh sets how much it holds. Duration is simply the energy divided by the power, and it is the figure that says how long the battery can sustain its full output before it is empty. Buying more power does not buy more hours, and buying more hours does not raise the peak you can shave.

The clearest evidence is a maker who sells one power rating at two durations. These configurations are read from our UK distributor's commercial cabinet listing as at 15 August 2026.

• Commercial Storage System, 50 kW / 102.4 kWh. Holds about 2 hours at full output. The maker's 2-hour designation. The shorter half of a matched pair at one power rating.
• Commercial Storage System, 50 kW / 204 kWh. Holds about 4.1 hours at full output. The maker's 4-hour designation. The same 50 kW of power at roughly twice the energy. The peak it can shave is unchanged; the window it can hold through doubles.
• G-Max, 100 kW / 215 kWh. Holds about 2.2 hours at full output. A higher power rating in an all-in-one cabinet, at a comparable duration to the pair above.
• G-Max Plus, 100 kW / 241 kWh. Holds about 2.4 hours at full output. The same 100 kW again, with more energy behind it. The extra capacity buys a longer discharge, not a taller one.
• G-Max Plus, 125 kW / 241 kWh. Holds about 1.9 hours at full output. The reverse move: the same energy at a higher power rating, so it runs out sooner. A sharp maximum-demand peak wants this shape; a long evening shift does not.

Across that same listing the cabinets run from 50 kW to 125 kW and from 100 kWh to 261 kWh, which is a duration span of roughly 1.9 to 4.1 hours. That listing is one dated view of one catalogue rather than the whole of what we can specify, so a maker or a configuration it does not happen to show is not thereby unavailable, and we re-confirm the part number and the configuration against the current listing and the manufacturer datasheet before either reaches a proposal.

Four things settle where in that span your site should sit, and all four come out of the same model that sets the power rating.

• How deep the peak is, and how long it lasts. The depth sets the kW; the length sets the hours. A sharp half-hour spike is a short-duration job, and paying for a four-hour store to clip it is capital that never discharges. A demand plateau that holds across a whole shift is the opposite case. The mechanics are on peak-shaving.
• How many cycles a day your tariff supports. A tariff with one cheap overnight window and one expensive evening wants a store that covers that evening in a single discharge. A tariff with two separated price spreads can run two shallower cycles a day from a shorter cabinet, which is a different specification at the same annual benefit. Cycles also consume warranted throughput, so the duty pattern belongs in the sizing rather than in a footnote. See time-of-use arbitrage.
• Whether the site is chasing market revenue. The market frameworks specify duration directly. The Capacity Market de-rates duration-limited storage by how long it can sustain output, so a shorter cabinet derates harder than a longer one at the same power, and the frequency-response services each state a response and a sustain requirement the asset has to meet. If those revenues are part of the case, duration stops being a free choice. The mechanisms are on grid services and the Capacity Market.
• The connection limit. Power is the axis your network operator constrains. Where the import or export rating agreed under a G99 application is already tight, adding kW is the move you cannot make, and adding hours at the same kW is the one you can: it raises the energy the site can shift without raising the rating the connection is assessed on.

So the specification we bring you is a power rating and a duration, each traceable to something in your data, and the cabinets we then look at are real part numbers with real configurations rather than a rounded kWh figure. What the duration choice does to the capital is set out on battery costs.

Backup is its own calculation, and it is rarely the whole building. The right question is which loads must stay live through an outage and for how long.

That critical load sets the power rating: the battery and its inverter have to carry the simultaneous demand of everything on the protected circuits, including the surge when motors and compressors restart. The duration you need sets the energy. A few minutes to ride through a dip and hand over to a generator is a very different store from running a cold room or a server room for hours.

Sizing backup honestly means mapping the critical circuits first, then the run-time, then the rating. Protecting the whole site for a long duration is possible, but it is a much larger battery, so the engineering and the cost track the scope you actually need.

Sizing is a modelling exercise, and the inputs are specific to your site. Three things drive it:

• Your half-hourly load data, which shows when you draw power, how high the peaks run, and how long they last across a real year.
• The solar generation profile for the array, modelled against the roof from the in-house insured drone survey, so the surplus the battery would store is a real shape rather than an assumption.
• The use case you are optimising for, which decides whether power or energy is the binding constraint.

We model all three together in PV*SOL before any battery is specified. That is what tells us the power rating, the usable energy and the round-trip behaviour the site needs. It also catches the cases where a smaller, better-matched battery does the job a larger one was assumed to require.

An oversized battery is wasted capital that rarely fills. An undersized one clips the very peaks it was bought to shave, or runs flat before the expensive window closes. Both are sizing failures, and both are avoidable by modelling first.

The whole installation is then engineered around the figure the model produces: the inverter and protection to BS 7671, commissioning and verification to IEC 62446-1, the G99 application to your DNO, and the works run under CDM 2015 on a JCT or NEC contract. The battery is specified to fit the job, not the job stretched to fit a battery.

A battery is not specified by its nameplate kWh. It is specified by the energy it will actually deliver over a warranted life, and that is a smaller number. Two derations sit between the headline figure and the figure the model uses.

The first is depth of discharge. The bankable lithium iron phosphate systems we lead with keep headroom at the top and bottom of the pack to protect cycle life, so the usable energy is a fraction of the nameplate. The second is round-trip efficiency: the energy that comes back out is less than the energy that went in, because the cells, the power-conversion system and the auxiliaries all take a share. We size on usable energy after both, never on the brochure capacity, so the store still clears the shifting window in year ten and not only on the day it is commissioned. The cells, conversion electronics and energy management are tested to the IEC 62619 safety standard for industrial lithium cells and the IEC 62933 series for the system as a whole, and the warranted throughput is read from the datasheet rather than assumed. How chemistry sets that usable depth is covered on how a commercial battery works, and the capital it drives on battery costs.

For a peak-shaving or grid-services battery, the half-hourly meter is the wrong resolution to size the power rating from. Network charges and the metered maximum demand are settled on half-hourly averages, but the current that trips your main breaker or loads your transformer is an instantaneous figure, and a thirty-minute average hides the spikes inside it.

So we read the demand at two resolutions. The half-hourly data, the same data behind your half-hourly metering, sets the energy and the shape of the day. Where the job is to clip a maximum-demand charge or hold a network limit, we look harder at the instantaneous and short-interval peaks, because the power rating has to carry the real spike and its duration, not the averaged one. The same logic decides an export-limitation scheme under G100, where the battery and its controller hold the site inside an agreed limit in real time rather than on average. Getting this resolution right is the difference between a battery that actually shaves the peak and one that is averaged into missing it. The peak job itself is set out on peak-shaving.

When a battery earns from the markets as well as serving the site, the dispatch pattern feeds back into the sizing, because the two jobs compete for the same energy and the same cycles. A pack held full to be ready for a frequency-response call is not free to soak up midday solar, and a pack cycled hard for wholesale arbitrage ages faster than one cycled gently.

We model the stacked use case before specifying, so the rating carries the combined duty rather than one job at a time. The market frameworks are named and modelled by mechanism, never by a promised return: the Capacity Market run by the National Energy System Operator, the Balancing Mechanism, and the ancillary frequency-response services such as Dynamic Containment. Any revenue, payback or arbitrage spread is modelled against your tariff and dispatch and disclosed with its basis; it is never a fixed figure quoted on a page. How those markets pay, and how the duty cycle shapes the specification, is covered on grid services, the Capacity Market and the page on battery costs and return.

The model produces a number, but the site decides whether the number can be installed. Two constraints commonly move the specification after the load is read.

The first is the grid connection. A battery that charges and discharges at the rating the model wants may exceed what your existing connection allows to import or export, which turns the project into a G99 application to your DNO, and can introduce an export or import limit the controller then has to hold. The second is siting and fire safety. A larger pack needs separation distances, ventilation and a fire strategy consistent with BS EN 62933-5-2 and the guidance the HSE and the fire service work to, alongside NFPA 855 as a reference for spacing where it is cited, all of which is set out on battery fire safety. Those requirements can favour a smaller indoor cabinet or push a bigger system to an outdoor enclosure, the choice covered on containerised or indoor housing. We resolve the connection envelope and the siting envelope alongside the load model, so the specified battery is one that can actually be energised on your site, not only one that balances on a spreadsheet.