Energy demand forecasting

Plan tomorrow's loadwith more confidence

  • When demand will peak
  • How much demand is needed for tomorrow
  • How uncertain the forecast is
Talk to us
Site demand (MW)0123456Today 12:0018:0000:00Tomorrow 06:0012:0018:0024:00NowObservedForecastExpected peak4.8 MW · 14:3080% prediction interval
24-hour demand forecast

Values are illustrative.

Why it matters

A single demand forecast can affect cost, capacity, and operations

Peak demand follows a daily pattern. The main drivers:

  • Weather
  • Site activity
  • Operating conditions
Demand (MW)012345600:0006:0012:0018:0024:00Annual peak5.0 MW · 08:304.1 MW · 18:00
Day-ahead market closes 12:00 the day beforeBoth profiles are committed before the day starts.
Typical dayEarly-peak day
Profile shapes are two real German weekday load curves (SMARD, 12 Feb and 22 Jan 2026), scaled to a 5 MW site. Your own profile will differ.

A bad forecast can affect your entire business

Market & cost

01
  • Higher procurement cost
  • Greater imbalance exposure
  • Higher demand charges

Capacity

02
  • Contracted limits can be exceeded
  • Operating headroom is reduced
  • Capacity upgrades may be needed earlier

Operations

03
  • Less time to react
  • Workloads may need rescheduling
  • Storage may need to be activated
  • Consumption may need to be reduced
  • Flexibility may need to be procured at short notice

Business case

Estimate your savings

See what better day-ahead forecasting could be worth based on your annual electricity consumption

GWh
1 GWh100 GWh

Move the slider or type a value to see your potential savings.

Estimated annual savings

€30,500/ year

ConservativeExpectedStrong
€20,000€41,000

Compared with a simple day-ahead strategy, using historical day-ahead and intraday prices.

How we calculate this

Peak charges. Your peak is implied at 5,000 full-load hours, a representative German industrial profile. We assume 4% of that peak is avoidable with better forecasts, valued at the medium-voltage Leistungspreis (€130 / kW / year in 2026).

Market costs. A 1.5 percentage point cut in day-ahead forecast error, valued at €12 / MWh: the measured mean absolute gap between the intraday ID3 price and the day-ahead auction price. Because this is an absolute gap, the market component is an upper-bound scenario, not a guaranteed directional saving.

The range. Conservative and strong vary the two inputs we have measured spreads for: the network charge (€90 to €160) and the intraday gap (€5.54 to €24.82, the observed spread of daily averages).

Gap measured over 288 hours, Sep 2025 to Aug 2026, from Energy-Charts (Fraunhofer ISE). Day-ahead cross-checked against SMARD.

Estimates are based on historical price spreads and representative load assumptions.

Forecast workflow

How sktime helps

sktime helps teams work with multiple models, complex covariates, and probabilistic outputs, so forecasts are easier to test, trust, and use in practice.

01

Inputs

  • Demand history
  • Weather
  • Metadata
  • Time resolution
02

Workflow

  • Prepare data
  • Compare models
  • Backtest
  • Predict intervals
  • Diagnose pipeline
03

Outputs

  • Forecast curve
  • Peak timing
  • Peak magnitude
  • Risk range
  • Rule trigger

Why sktime?

  • Open source and extensible
  • No vendor lock-in
  • Sovereign data stack
  • Rich model and estimator set
  • Easy benchmarking and backtesting on your data

Get started

Get started in 3 easy steps

  1. 01

    Scoping call

    We map your series, your drivers, and where the load forecast hurts most today.

  2. 02

    Pilot on your own data

    We forecast a real slice of your demand, with ranges, and you see the lift first-hand.

  3. 03

    Run in operations

    Roll it into your dispatch and planning cadence, with the maintainers a message away.

Open source meets enterprise

Bring honest demand forecasts into your operations.

sktime is free and open-source. When you want this running on your own load data, the people behind it can help.

Talk to us