Most commercial building BESS projects are lost in the same place: nobody decides which of the three jobs is actually paying for the battery. Demand management wants power. Solar self-use wants energy. Backup wants a battery that never cycles. Put all three on one set of cabinets and the nameplate grows faster than the savings.
Three jobs, one battery, three incompatible answers
A building's peak, its roof, and its emergency loads each write a different specification. The conflict is not commercial. It is arithmetic.
Condition for the table above: low-voltage commercial connection with an interval meter, one tariff contract, one rooftop PV array. Source: control duties follow published C&I dispatch practice; the cycle counts here are design assumptions and should not be read as warranty terms.
Read the fifth row against the fourth and the problem appears. Demand management and solar self-use both want the battery to work every day. Backup wants it full, uncycled, and waiting. Every kilowatt-hour you lock for backup is a kilowatt-hour that cannot earn anything on a Tuesday afternoon.
The building imposes a fourth requirement that rarely appears in a tariff study, which is where the cabinets can physically go. A plant room with no crane access, a boundary that the neighbours can hear, and a footprint already occupied by the chiller set constrain the format long before the chemistry matters. Air-cooled outdoor cabinets are easier to place and easier to service; liquid-cooled units hold more energy per square metre and a tighter cell temperature spread, which matters more the harder the asset is cycled. Format first, then the electrical design.
What the building has to give you before any sizing starts
The interval meter is the first deliverable, not the battery.
Source column note: these are the inputs a commercial BESS proposal cannot be built from a monthly bill. A single monthly kWh figure conceals both the peak and its duration, and those two numbers price the battery.
A worked example
Worked example. The building below is hypothetical; its load figures are illustrative assumptions chosen to make the arithmetic traceable. The equipment figures in the last table are from published product data.
The building. 5,000 m2 mixed office and ground-floor retail, one 800 kVA transformer, 240 kWp rooftop PV, annual consumption 1,850,000 kWh, night base load 78 kW. Metered peak of 617 kW, reached in 45 consecutive intervals in August. Winter peak 548 kW. Demand charge 13.40 per kW per month. Day energy 0.22 per kWh, night 0.09, export 0.055.
Job 1, demand management. Cap the peak at 480 kW.
- Shave required: 617 - 480 = 137 kW
- Energy at the meter: 137 x 0.75 h = 102.75 kWh
- Battery-side energy: 102.75 / 0.94 = 109.3 kWh
- PCS rating: about 150 kW, covering a 137 kW duty plus control margin
- Months exceeding the cap: 7 of 12, from the load data
- Annual avoided charge: 137 x 13.40 x 7 = 12,850.60
137 kW of shaving needs about 110 kWh of battery. That is roughly 3% of what a four-hour system of the same power would carry. Most commercial buildings are sold the four-hour system.
Job 2, solar self-use. At midday the array produces 165 kW against a 130 kW building load, so 35 kW leaves the site at 0.055 per kWh.
- Surplus absorbed: 35 kW x 5 h = 175 kWh/day
- Margin gained: 175 x (0.22 - 0.055) = 28.88 per day
- Annual working days: 260 assumed, giving 7,508.80
Job 3, backup. Critical loads total 100 kW for 2 hours, which is 200 kWh delivered.
- Nameplate required: 200 / (0.90 DoD x 0.94 discharge path) = 236.4 kWh
- Locked reserve: 236.4 kWh, unavailable for daily cycling
Combined. The working block has to cover whichever daily duty is larger. The two duties do not stack, because the peak usually arrives outside the solar window:
- Locked for backup: 236.4 kWh
- Working block: max(109.3 ; 175) = 175.0 kWh
- Minimum nameplate: 236.4 + 175.0 = 411.4 kWh
Two published configurations bracket that number.
Figures are quoted from the published product specifications; confirm against the current datasheet before specifying.
Reading the numbers without overselling them
Three things follow from the arithmetic above.
The demand charge is usually the largest single line, and the smallest energy requirement. On a ratchet tariff, one missed peak sets the floor for the following eleven months, which raises the value of the function and the cost of getting the control loop wrong.
Solar self-use is priced by the tariff spread. The array only decides how much energy is available to shift. Where export is compensated near retail, storing surplus adds nothing and Job 2 deletes itself. Where export is capped or paid at a wholesale rate, the same 175 kWh carries real value.
Demand charges are billed on the 15-minute average rather than on the instantaneous spike, and that distinction decides where the energy goes. A control loop that reacts in ten seconds still lets a fourteen-minute overload land inside the interval that sets the bill. To hold the cap for the whole window, the 137 kW has to be sustained, so the margin belongs in the PCS rating and in the energy block, not only in the control logic.
Backup has no billable value unless someone has written down the cost of an outage hour. If the number does not exist on paper, the reserve should not exist in the battery.
Capacity fade reaches the reserve before it reaches the savings. A cabinet that starts with 236 kWh locked for backup does not still hold 236 kWh in its tenth year, and the end-of-warranty figure is the one the backup function actually depends on. Size the locked block from the Year-10 number and the working block from the Year-1 number. Sizing both from the nameplate is the most common arithmetic error in this configuration.
And the sentence that decides most specifications: if the backup reserve is more than half the nameplate, the battery is a generator, not an energy-saving asset.
The four assumptions that move the answer most
Two of the four changes shrink the battery. That is the point of running them before the quotation, not after.
One caveat I cannot resolve from the load data: whether the 45-minute peak arrives in the same month every year, or whether it followed a specific tenant load that has since changed. On one year of interval data I would size the cap but not promise it. Two years would settle it.
Where this sizing stops working
The arithmetic above holds for a low-voltage commercial connection below 1 MVA with an interval meter and a single tariff. Above that, the protection study and the tariff structure change, and the demand charge is rarely the binding constraint.
It also stops working in four common cases.
If the peak lasts longer than an hour, the demand charge is no longer the driver and the energy bill takes over. If the transformer is already at its limit, the constraint is the connection, not the battery. If the landlord does not pay the tenants' electricity, the demand saving accrues to someone else and the case has to be rebuilt on resilience or on rent. And if nobody is contracted to answer the site at 2 a.m., the backup function is theoretical, which is a maintenance question rather than a sizing question.
The saving also has to be provable. Fix the baseline before commissioning, from the same meter and the same interval length, and keep every cap event in the EMS log with a timestamp. A demand reduction that cannot be shown month over month will be credited to the tariff rather than to the battery, and if the tariff structure is renegotiated mid-year the baseline has to be restated before any comparison means anything.
Two similar sites were commissioned in Europe with very different modularity. A 600 kWh commercial installation in Estonia, completed in January 2026, was built as three 200 kWh cabinets on an on/off-grid arrangement ( project record ); a 112 kWh outdoor system in Poland, completed in July 2026, pairs a single cabinet with a 50 kW inverter, the array mounted on a bifacial fence beside the enclosure ( project record ). Both were delivered by Ruibit Energy as an authorized distributor of Dawnice Energy.
Two items on this example are still open. Who owns the peak-control loop, the EMS or the building management system, and which of the two meters the control loop reads. And the tariff contract renewal date, because a demand ratchet that resets changes the value of every month in the table above.
FAQs
1. How much battery capacity does demand-charge shaving actually need?
Far less than most quotations assume. Shaving a 617 kW peak to 480 kW needs about 110 kWh of battery-side energy over a 45-minute peak, roughly 3% of a four-hour system of the same power.
2. Can one commercial building battery handle demand management, solar self-use, and backup?
Only if the backup reserve is locked first, because it cannot cycle. The working block must then cover the larger daily duty: a 236 kWh reserve plus a 175 kWh daily duty gives a 411 kWh minimum nameplate.
3. Why is a monthly electricity bill not enough to size a building battery?
A monthly kWh figure hides both the peak and its duration, and those two numbers price the battery. Sizing needs 12 months of 15-minute interval data plus the demand charge, its ratchet clause, and the tariff periods.
4. Does storing solar surplus always save money?
Not always. Self-use value depends on the tariff spread. Where exported power is paid near retail, storing surplus adds nothing; the battery earns only where export is capped or paid at a wholesale rate.
5. What happens to backup capacity over ten years?
It shrinks. A cabinet holding 236 kWh for backup will not hold that figure at end of warranty. Size the locked block from the Year-10 number and the working block from the Year-1 number.
6. When does demand management stop being the main driver of the design?
When the peak lasts longer than about an hour. Past that point the energy bill takes over and the demand charge stops driving the sizing.
