EU industrial electricity tariffs are structured around a concept that most plant engineers understand in theory but underestimate in practice: your electricity bill is not simply a function of how much energy you use. For mid-to-large industrial consumers in Germany, Sweden, the Netherlands, and most other EU member states, a significant fraction of the monthly bill is determined by your peak demand — specifically, the highest 15-minute average load recorded in the billing period. One unmanaged furnace startup, one simultaneous machine initialization at the beginning of a shift, and your capacity charge for the next 30 days is set.
This article explains how capacity charge structures actually work in EU industrial tariffs, why the 15-minute window is the lever that matters, and what load monitoring changes about the ability to manage that lever deliberately.
How the Leistungspreis and Effekttariff Actually Work
In the German grid tariff system, industrial consumers above a certain annual consumption threshold (typically 100,000 kWh/year) pay under a two-part tariff: the Arbeitspreis (energy price, €/kWh) and the Leistungspreis (capacity price, €/kW/month or €/kW/year). The Leistungspreis is calculated from the peak measured demand in 15-minute intervals over the billing period. In practice, for a mid-size manufacturer drawing 3–8 MW average load, the capacity component can represent 35–55% of the total electricity bill.
The Swedish equivalent — the effekttariff — works similarly, though implementation details vary by distribution system operator (DSO). Swedish industrial consumers above certain load thresholds pay a capacity charge based on the average of their three highest 15-minute peak demand measurements in a calendar month. Some Swedish DSOs use the single highest peak; others use an average of the top three. Either way, the principle is identical: the billing period's capacity charge is set by a small number of extreme demand events, not by average consumption.
Finnish and Norwegian industrial tariffs follow analogous structures. The pattern across EU industrial markets is consistent: the highest short-duration demand interval dominates capacity billing in a way that is disproportionate to its contribution to actual energy consumed.
The Furnace Startup Problem
Consider a steel fabrication facility — call it Volmark Steel — operating a 2.4 MW induction furnace and four CNC machining centers, each drawing 45–60 kW at peak spindle load. During normal production, the facility's average demand is roughly 1.8 MW, well within a comfortable capacity band. But at 06:00 each morning, the induction furnace comes online after its overnight cooldown. The startup sequence pulls the furnace from cold to operating temperature over approximately 25 minutes — during which the resistive heating element draws at or near nameplate capacity. If that coincides with the day-shift CNC startup sequence and the facility's compressed air compressors cycling up from night-mode standby, the combined 15-minute interval can reach 3.1–3.4 MW.
That 3.1 MW peak, occurring once on one morning in January, sets the capacity charge for the entire month of January. At a German Leistungspreis of €120–160/kW/month (typical for medium-voltage industrial consumers in 2024–2025), the difference between a 1.8 MW demand baseline and a 3.1 MW peak event is 1.3 MW × €140/month = €182,000/year in additional capacity charges — if that peak pattern repeats monthly. Even as a one-off, the capacity event in January inflates that month's bill by €182,000 × (1/12) ≈ €15,000 for a single poorly timed morning startup.
We are not saying that the furnace should not start at 06:00. Production schedules exist for good operational reasons, and reorganizing them to optimize capacity billing is not always feasible or even desirable. What we are saying is that most facilities have no visibility into which specific 15-minute windows are setting their capacity bills each month, and therefore cannot even evaluate whether a staggered startup sequence would be worth the operational adjustment.
Where Load-Curve Monitoring Changes the Math
At 1-second measurement resolution, a load curve from an induction furnace startup is precisely detailed: the current draw ramp from 0 to full load over several seconds, the initial high-demand zone as the furnace charges from cold, the thermal plateau as the charge approaches target temperature, and the load reduction as the setpoint is maintained. The integration of that curve over any 15-minute window is calculable in advance, not just in retrospect.
This matters because demand management only works proactively. By the time a 15-minute window has closed and the peak has been set, there is nothing to be done. The window for intervention is the 5–10 minutes before a high-demand event is predicted to exceed the current peak threshold.
Predictive peak alerts work from two data sources: the historical load curve of the asset in question (showing typical peak duration and magnitude for a furnace-from-cold startup) and the current demand baseline (what is already running and contributing to the 15-minute average). If the current baseline is 1.9 MW and the furnace is scheduled to start in 8 minutes and historically adds a 45-minute, 1.8 MW load profile during initial charge-up, the predicted 15-minute peak is approximately 3.0–3.2 MW. That prediction, delivered 8 minutes before the event, gives the plant engineer four practical options: delay the furnace start by 20 minutes (until the shift's CNC startup demand has stabilized), reduce the initial furnace heating rate if the temperature setpoint permits a slower ramp, briefly curtail another deferrable load, or accept the peak and log the decision.
The key word is deliberate. The same peak that costs €15,000 in one billing period becomes a known, planned event — or avoidable — once the underlying load data is visible.
The Compressor Contribution to Demand Peaks
Induction furnaces and large resistive heating elements are the obvious demand peak contributors. Less obvious is the compressor. A 75 kW screw compressor starting from a complete standstill draws startup current equivalent to 4–6× its running current for 2–4 seconds. That initial transient rarely affects 15-minute averages in a meaningful way. But a compressor in an aggressive duty cycle — driven by high leak losses, as discussed in the previous article — can contribute to demand peaks through frequent load-unload cycling.
In a load-unload compressor control mode, the machine draws full power during the loaded phase and near-zero power during the unloaded phase. If the loaded phase duration happens to align with another peak demand event on the plant — a furnace startup, a large welding robot initializing, a packaging line cycling up — the 15-minute window that captures both events sets a peak that neither event would have set alone. Compressor staging optimization (covered separately) addresses this through better duty cycle management, but the prerequisite is having the load data to see the coincidence pattern.
Baseline Establishment and Ongoing Monitoring
Effective demand management starts with a 30-day load baseline that maps every recurring demand event to its typical 15-minute peak contribution. This baseline is the foundation for both alert thresholds and operational rule changes.
The baseline analysis typically surfaces three categories of demand events: predictable scheduled events (furnace starts, shift changeovers), semi-predictable variable events (large batch processes with variable timing), and unpredictable reactive events (alarms, equipment recovery sequences). Predictable events are the most tractable — scheduling adjustments can eliminate or flatten them. Semi-predictable events can be managed with soft alerts. Unpredictable events require either demand response capability (the ability to curtail a non-critical load on short notice) or acceptance that occasional peaks are unavoidable.
For most mid-size EU manufacturers, the 30-day baseline exercise alone — just reviewing the load-curve data for the top five demand events per month and understanding what caused each one — typically identifies 3–5 operational changes that reduce the monthly peak by 12–22% without touching production schedules. The compressor cycling into the morning furnace startup, the packaging line initializing simultaneously with the shift-change HVAC ramp-up, the welding station bank powering on together rather than in sequence: these are scheduling inefficiencies that compound into capacity charges, and they are only visible when the underlying data is present.
What the Data Does Not Fix
Load monitoring creates visibility; it does not automatically create the operational discipline to act on that visibility. The gap between knowing which 15-minute window is setting the monthly peak and reorganizing production sequences to avoid it involves production planning, maintenance scheduling, and sometimes shift supervisor buy-in. Facilities that treat demand management as a one-time configuration exercise — set alerts, generate a report, forget it — tend to see initial improvement followed by gradual reversion as operational pressures reassert themselves.
The most effective demand management programs we have observed run load-curve reviews as a monthly ritual: a 30-minute session where the energy manager or plant engineer reviews the previous month's peak events, confirms whether they were anticipated or unexpected, and updates the alert thresholds for the following month based on any operational changes. The review does not need to be elaborate. What it needs to be is consistent.
The 15-minute window that inflates your capacity bill is not a natural law. It is an operational artifact — a consequence of equipment starting at the same time because nothing has ever explicitly scheduled them to start at different times. That is the problem the data makes tractable.
