Peak shaving

A commercial or industrial grid connection is billed for energy and for power. The power part depends on the highest load the meter records in a billing interval. Peak shaving keeps that recorded peak down. A battery discharges, or flexible load such as heat pumps, compressors or EV chargers is reduced, during the intervals when consumption would cross a threshold. The battery recharges when demand is low. The threshold and the interval length come from the network tariff, so the tariff rules decide what the control software has to do.

Two models are common in Europe. In the first, the site has a contracted capacity and pays a fee whenever it goes above it. In Poland, under § 53 of the tariff regulation of September 3, 2026, the network operator works out each hour's overrun from power averaged over 15-minute periods and charges the fixed component of the network rate multiplied by the sum of the ten largest overruns in the month. Where the meter cannot record 15-minute or hourly values, the fee is ten times the largest overrun in the billing period.

In the second model, the site has to keep its load low only inside time windows. In Germany, § 19(2) of the StromNEV gives a consumer whose peak predictably falls outside the annual network peak an individual network charge, which may not be lower than 20% of the published charge. Network operators publish high-load time windows (Hochlastzeitfenster, HLZF) per voltage level and season, following the Bundesnetzagentur decision BK4-13-739 of December 11, 2013. The windows apply on working days only. For 2026 one operator, N-ERGIE Netz, sets these conditions: the maximum load inside the windows must be at least 20% below the annual peak at medium voltage (10% at high voltage, 30% at low voltage), the difference must be at least 100 kW, and the yearly saving must reach at least 500 EUR.

The tariff rules turn into these requirements for the specification:

• The controller works on the same averaging interval as the bill. If the tariff uses 15-minute averages, the EMS tracks the running average of the current quarter-hour and how much energy is left before it crosses the limit. Reacting only to instantaneous power wastes battery capacity on harmless spikes.
• The battery has to be full when the window opens. That needs a load forecast and a reserved state of charge, so the EMS charges ahead of the window and does not sell that energy.
• Windows, seasons, public holidays and contracted capacity are configuration, because operators update them every year. A code release for a new tariff year is a design fault.
• Peak shaving and trading compete for the same kilowatt-hours. The priority between them, and what happens when the reserve runs out mid-window, must be written down.
• Battery ramp rates and conversion losses between battery and grid change how much power actually reaches the meter, so the controller needs both.
• The EMS keeps its own 15-minute record of grid power and compares it with the operator's metering data. Without that record you cannot check an overrun invoice.
• A missed window is expensive, so alerts must fire before the window: battery unavailable, state of charge too low, communication lost.

Peak shaving is one of the operating modes of the EMS we built for Skyfri, which has run fully offline on 7 plants since June 2024. As the project documentation recorded in October 2024, the first requirements for the system came from industrial sites in Germany and included peak shaving against HLZF windows, trading through an external partner and export curtailment on grid operator request. By then the main control loop and a maximum site load limiter were done. Work then in progress included live HLZF tests in November 2024, peak shaving window management, battery ramp rate management and a calculation of losses between battery and grid. An algorithm that prepares the battery before each high-load window, and further variants for other factory layouts, were next on the plan.