Direct answer: Demand charges are billed on your site's highest power draw in a billing period — typically the highest 15-minute average — so a single brief surge sets the charge for the whole month, no matter how efficiently the site runs afterwards. A battery reduces that charge by discharging behind the meter whenever site load approaches a defined grid-demand cap, so the operator's meter never records the peak. Sizing is therefore a two-part calculation: how many kW you intend to shave, and for how long you must hold that reduction.
It is also the class of value that solar alone cannot deliver. A PV array reduces energy charges — kilowatt-hours. It does not reduce demand charges — kilowatts — because the site's peak draw usually happens when the array is producing little or nothing.
Most commercial bills mix two structures that behave nothing alike:
Line item | Billed on | Reduced by |
Energy charge | Total kWh consumed in the period | PV generation, load shifting, efficiency measures |
Demand charge | The peak kW reached in the period | A battery discharging at the right moment — and little else |
That distinction is where most misleading payback estimates originate. Industry analyses commonly cite NREL research putting demand charges at roughly 30% to 70% of a commercial electricity bill, with the share rising for facilities whose load is spiky rather than steady — cold storage, food processing, water treatment, plastics, machining, and buildings with large HVAC or chiller loads.
The mechanics are worth stating plainly: because the charge is set by the highest interval reading in the period, one bad 15 minutes defines the month. A chiller starting, a production line ramping up, or a compressor kicking in simultaneously with an HVAC call can lock in a charge that survives every efficient day that follows.
Peak shaving is sized from a load profile, not from a rule of thumb. The five steps:
Step 1 — Get interval data. You need the site's load at 15-minute (or finer) resolution for at least a full year. The electricity supplier or the operator's meter data portal is the normal source. Monthly bills alone are not sufficient; they hide the peaks you are designing against.
Step 2 — Find the target peaks, not the single maximum. Rank the interval readings and look at how often the site exceeds a candidate cap. The economic question is not "what is the single worst 15 minutes" but "what cap captures most of the charge, cheaply".
Step 3 — Convert the cap into required energy.
Required battery energy (kWh) = shaved power (kW) × hold duration (h) ÷ (DoD × round-trip efficiency)
The DoD (depth of discharge) and efficiency terms matter. A 100 kWh nameplate pack cycled to 90% DoD with 90% round-trip efficiency delivers roughly 81 kWh at the meter, not 100.
Step 4 — Check the C-rate. The battery must be able to deliver the shaved power continuously, and the inverter's battery-side current rating must support it. A pack sized for energy but not for power will fail exactly when it is needed.
Step 5 — Iterate against the tariff. Demand charges differ by market, by season and by contract. Some tariffs charge on a monthly maximum, some on a seasonal one, some on a rolling 12-month maximum — and the last of these punishes a single event for a year. The target cap should be re-optimised against the actual tariff structure, not a generic one.
Illustrative arithmetic. Suppose a site peaks at 400 kW and the operator's tariff makes a 300 kW cap attractive. Shaving 100 kW for 2 hours requires:
100 kW × 2 h ÷ (0.90 × 0.90) ≈ 247 kWh of usable battery capacity
and a discharge capability of at least 100 kW — roughly a 0.4C rate on a 247 kWh pack, which is comfortably inside the capability of a lithium iron phosphate (LFP) system. Which is the point: peak shaving is a power requirement, and LFP's excellent cycle life at moderate C-rates is exactly why it dominates this application. (Values are illustrative; substitute the site's own interval data and tariff.)
For a C&I peak-shaving system the battery and the power conversion stage are specified against separate windows, and both need checking:
Battery voltage window and energy.
● DMUDA's high-voltage rack-mounted LiFePO4 range spans 143–215 kWh
At 512–768 V and 280 Ah, with an integrated master control and combiner box and a two-level or three-level BMS architecture. Charging is specified from 0 °C to 55 °C
and discharging from −15 °C to 60 °C— note the asymmetric charge window, which is the LFP cold-charging limit stated explicitly and the reason an unconditioned plant room in Northern Europe needs a heating provision in the design.
Power conversion.
● DMUDA's high-voltage three-phase hybrid inverter platform (DF-HVHHxxK3P-EU) runs to 60 kW per unit with a 160–700 V battery window, up to 10 units in parallel
, diesel-generator direct input, and six programmable charge/discharge time zones — the last of which is what a time-of-use or demand-charge strategy actually runs on. It is certified to IEC 61727 / IEC 62116 and EN 50549-1 under grid regulation, and IEC/EN 62109-1 / 62109-2 under safety.
Monitoring and control.
● Demand-charge control is an EMS function, not an inverter setting. Whatever platform you specify must be able to read site load and act on it; a system that can only follow a fixed time schedule will not hold a demand cap against a variable load.
C&I batteries in Europe now carry obligations that most payback models ignore:
Carbon footprint declaration
● applies to rechargeable industrial batteries with a capacity greater than 2 kWh from 18 February 2026.
● A battery passport, accessible via QR code, becomes mandatory for LMT, EV, and industrial batteries over 2 kWh from 18 February 2027 — and most C&I storage falls into the industrial category.
Recycled-content minima
● apply from 18 August 2031 (16% cobalt, 85% lead, 6% lithium, 6% nickel) and rise again in 2036.
The European Commission's Batteries regulation page carries the current timeline, and both dates are implemented through delegated acts — confirm them against EUR-Lex before writing them into a contract. The important structural point for a C&I buyer is that LFP contains no cobalt and no nickel, so the cobalt and nickel recycled-content obligations have no target material in an LFP pack. On a chemistry-neutral law, that is a chemistry-dependent advantage.
● Site interval load data at 15-minute resolution, minimum 12 months
● The tariff structure: what sets the demand charge, and over what window
● Target demand cap (kW), agreed with the site's energy manager, not assumed
● Required hold duration (h) for the peak event
● Usable energy required (kWh) after DoD and round-trip efficiency — not nameplate
● Continuous discharge power required (kW) and the resulting C-rate
● Battery voltage window, energy per rack, BMS architecture, and charge/discharge temperature windows stated separately
● Inverter battery-side current rating and the parallel-unit limit
● Certifications actually listed: EN 50549-1 plus national addendum, IEC/EN 62109-1/-2, IEC 62619 for the battery
● Carbon footprint declaration and battery passport readiness for industrial batteries >2 kWh
● Declared cobalt and nickel content, or confirmation of absence
● Monitoring platform capable of closed-loop demand control, not a fixed schedule
Can solar panels reduce demand charges on their own? Generally no, and this is the most common modelling error in commercial proposals. PV reduces energy charges, but a site's peak demand frequently occurs outside PV production hours — early morning, evening, or during a weather-driven load spike. Without a battery able to discharge at the peak moment, the demand charge is largely untouched.
How do I know what size battery a site needs? From interval data, not from the annual kWh figure. You need the kW you intend to shave and the hours you must hold it. Two sites with identical annual consumption can need batteries of very different sizes depending on how spiky their loads are.
Is a monthly maximum the same as a rolling 12-month maximum? No, and the difference is expensive. A rolling 12-month demand charge means one bad 15-minute event follows the site for a year rather than a month. Identify which structure applies before optimising the cap — the same battery can be an excellent investment under one and marginal under the other.
Does peak shaving wear the battery out? It is a moderate-C-rate, shallow-cycling duty, which is the operating regime LFP handles best. That is why LFP rather than a higher-energy-density chemistry dominates stationary peak-shaving installations. Cycle life is always quoted at a stated depth of discharge and temperature though — insist on both in the datasheet.
Can the same battery do peak shaving and solar self-consumption? Yes, and doing both is usually where the economics work. The constraint is that both functions compete for the same stored energy and the same inverter power, so the control priority between them must be an explicit design decision rather than something the site discovers in operation.
What is the smallest shave worth doing? It depends on the tariff, not on a fixed percentage. Calculate the annual demand-charge saving for a candidate cap from the interval data, divide by the installed cost, and compare against the site's hurdle rate. The answer moves with the tariff — which is why the tariff structure belongs in the feasibility study, not in the final quote.
