Peak Shaving with Battery Storage: How Demand Charges Actually Work

The one fact that matters

For sites with interval metering, the demand portion of the grid fee is billed on the highest average power drawn in any 15-minute window of the billing year, multiplied by a price per kilowatt. One quarter-hour sets the number; you pay it for twelve months.

This is why two factories with identical annual consumption can pay very different grid fees. The one whose machines start in a staggered sequence has a flat profile and a low peak. The one that fires everything up at 6:00 on Monday pays for that habit all year.

How the 15-minute measurement works

The meter records average power per quarter-hour, the load profile. At year end (or month end, depending on the tariff), the single highest value becomes your billed peak. Not the average of the top ten, not a typical day: the maximum.

Two properties of this rule do the damage. It is unforgiving, a peak that lasts one quarter-hour costs the same as one that recurs daily. And it is invisible until too late, by the time a person sees the spike forming on a dashboard, most of the window has already passed.

A worked example

The price per kilowatt is set by your grid operator's published tariff sheet, and the spread between operators is large. On 2026 German medium-voltage price sheets for sites above 2,500 utilization hours, the annual demand charge runs from around €60 to over €165 per kilowatt.

A battery that had discharged 200 kW during those fifteen minutes would have erased the entire line item. This is the core peak-shaving proposition: the battery doesn't need to run often to pay for itself; it needs to run at exactly the right quarter-hours.

What the battery actually does

Peak shaving is a forecasting problem wearing a hardware costume. The battery caps the peak by discharging while site load would otherwise exceed the target ceiling. Everything hangs on knowing which quarter-hours those will be: discharge too early and the battery is empty when the real peak arrives; hold too long and the peak is already billed.

This is why WEM treats the load forecast as the center of the system. The cloud plans a ceiling and a charge schedule against the forecast; the on-site edge watches the live meter and reacts in under a second when reality runs hotter than the plan, including when the internet is down, because the meter and the battery are both on site.

Why manual peak management fails

The traditional alternative is load shedding on alarm: a person or a relay switches consumers off when draw crosses a threshold. In a 15-minute averaging window, a reaction that takes five minutes has already conceded a third of the damage. And switching off production to save on grid fees is usually a bad trade, the product not made costs more than the peak.

A battery changes the economics because nothing has to stop. The site keeps running; the grid just doesn't see it.

The German specifics

Germany adds a layer that makes peak behavior even more consequential: reduced grid fees for predictable consumption patterns.

In outline, the legal basis is § 19 StromNEV: atypical grid usage rewards sites that avoid drawing power during the grid's own peak windows, and energy-intensive sites with very high utilization hours qualify for separate reductions. A battery interacts with both, it can shift a site's grid draw out of the critical windows, but it also changes the utilization calculation. Whether that helps or hurts depends on the individual tariff situation, which is a feasibility question, not a blog-post answer. The full mechanics are in the dedicated guide: atypical grid usage, explained.

What peak shaving stacks with

The same battery can shave peaks, arbitrage day-ahead prices, and bid balancing markets, but not with the same kilowatt-hour at the same moment. Capacity reserved to catch a forecasted peak is capacity not sold into the evening price spike. The honest version of "stacked revenue" is a daily optimization that prices these trade-offs against each other, which is precisely the job of the EMS layer.

Related: C&I battery storage, explained, the full picture including arbitrage and grid services · EMS vs BMS vs SCADA, which layer makes these decisions.