conpowr | Weather-driven price forecasts for European power trading

Where the forecast breaks down, the market moves.

Beyond day 7, numerical weather prediction loses its grip. Model runs swing from one cycle to the next, ensembles spread out, and a single shift in the guidance can move the power curve. That is where we work.

We don’t chase one deterministic answer. We read the probabilities, track how the models are evolving run by run, and tell you which changes matter for price before the market has priced them in.

German prices are more weather-driven than any other market in Europe, and the neighbours price off Germany. In the weekly and monthly contracts, the risk is not daily detail but the weather regime: a westerly flow means wind and mild weather, a blocking high can mean a week of low wind and cold.

And the weather that moves German prices is not only German. We also follow heat waves in France and the hydro balance in Scandinavia.

Every trading day: DE prices six weeks ahead

A daily DE day-ahead price forecast built from 201 paired weather and price members across EC EPS, EC S2S, MOGREPS and GFS ENS, with the cheap and the expensive scenario spelled out.

conpowr DE day-ahead price forecast, multi-model ensemble, run 11Z 24 June 2026 — Example: DE price forecast, run 11Z 24 June 2026. Click to enlarge.

Price distribution per day

P25 to P75 and the median for every day out to six weeks, with the EC operational run as reference and a flag where the risk of prices above 100 EUR/MWh exceeds 25%.

Alternative scenarios

Members split into low, mid and high renewable output, each with its own price and share. You see how expensive the expensive case is, and how likely.

The weather behind it

Wind and solar output per day, coloured by model, so you can see where EC, GFS and the others disagree and what that does to the price.

The events that move the price, flagged early

On top of the price forecast we track the weather events that matter most for German power, at home and in the neighbouring markets, each with a fixed definition so every call can be checked afterwards.

Dunkelflaute

Days of low wind and little sun. Gas sets the price in many hours, especially peak.

Cold spell

Demand and gas burn rise, and French heating load pulls through the coupled markets.

Heat wave

Nuclear output cut as rivers warm, low Rhine levels, weak wind under the high and an expensive evening peak.

Nuclear curtailment

Reactors close or cut output when cooling water runs too hot or too low, and supply tightens across the coupled markets.

High wind

Low or negative prices, above all at weekends and on public holidays.

Storm cut-out

Turbines shut down above cut-out speed, so output can drop sharply inside a windy scenario.

High solar, low midday prices

Strong sun pushes midday prices low or negative, deepest at weekends, while the evening peak stays expensive.

Negative-price weekend

Sun and wind meet low demand, with deeply negative hours in base and peak.

Beyond Germany

Heat waves in France

Hot rivers force French nuclear plants to cut output just as cooling demand rises. French exports fall and German prices follow.

Hydro balance in Scandinavia

Rain, snow pack and reservoir levels set the Nordic price and how much power flows south to Germany. We track the balance against normal and the precipitation outlook weeks ahead.

W+3 to W+6

Early Warning

Risk well above normal for the season, backed by a physical driver.

W+2 to W+4

Watch

The event scenario is growing in the ensembles.

W+1 to W+2

Warning

The event is the most likely outcome.

Short range

Confirmed

The event is under way or certain.

Early Warning: weeks, not days

Three to six weeks out, the ensembles sit close to normal, so a low probability can still mean a risk two or three times higher than usual. We measure against the season and only raise an Early Warning when a physical driver supports it: a weakening polar vortex, the MJO, a regime signal that holds over several runs, or dry soils ahead of a heat wave.

The curve usually prices normal conditions that far ahead. An Early Warning gives you time to act while protection is still cheap.

Event Watch, as it opens every outlook
Event	W+1	W+2	W+3	W+4
Dunkelflaute	10% →	25% ↑	40% ↑↑	30% ↑
Cold spell	5% →	15% ↑	30% ↑	25% →
High wind	45% ↓	20% ↓	15% →	15% →

Example values. Share of ensemble members meeting the event definition; arrows show the change since the previous run. Each flag comes with the price with and without the event.

Your day with conpowr

The same structure at the same time every trading day, so your desk can act on it in a minute.

06:30

Overnight runs analysed

The new ensembles are in. We compare them with yesterday and with each other.

07:55

Outlook in your inbox

Headline, W+1, W+2 and M+1 prices against the last run and the curve, Event Watch, scenarios and the price chart.

08:15

Live briefing

15 minutes on Teams with Q&A. Attendees cannot see each other, and every briefing is recorded.

08:30Summer time

Updated outlook

An update with the signals from this morning’s EC run.

15:0014:00 in winter

Afternoon outlook

The signals from the 06 UTC runs of GFS and AIFS and from the EC46 extended range, ahead of the next trading day.

Any time

Alerts

Only when an event changes level, straight into your Teams or Slack channel and your inbox.

Model comparison in plain words. You see the EC and GFS maps on free sites and hear the chatter on the desk. Every outlook says where EC, AIFS, GFS and MOGREPS disagree, which has verified better in the current pattern, and what the difference is worth in EUR/MWh.

For trading companies

Algo trading on the curve: weekly and monthly power futures

A system for your company’s own algorithmic trading of weekly and monthly contracts in Germany and the Nordics. Day-ahead and intraday are already automated. The curve is next. We build the algo with you, or bring the weather into the models your quants already run, from ensemble data to orders in the market.

Weekly futuresMonthly futuresQuarterly futuresGermany (EEX)Nordics

3 yearsHeading a desk trading weekly and monthly power futures on EEX and the Nordic market

13 yearsWeather and energy forecasts in 24/7 production at ConWX, with SLAs, monitoring and failover

201Paired weather and price members in the conpowr ensemble price model

The algo we build, from ensemble to order

Data

EC EPS, EC46, MOGREPS, GFS ENS, AIFS and IRIE, plus gas, carbon and the curve itself.

Weather features

Regimes, cluster shares, run-to-run shifts, wind and solar indices, hydro balance.

Fair value

A price distribution for every contract from W+1 to Q+1, not a single number.

Trading signal

Buy or sell per weekly and monthly contract, from the edge against the curve, weighted by confidence and how fast the models are moving.

Algo execution and risk

Orders sized and placed automatically within your limits, with stops and P&L attribution back to the weather.

Operations

Monitoring, failover, audit log and model governance your risk team can sign off.

Runs automatically after every model run, every trading day, with alerts when data is late or a model drifts

Desk AdvisoryWe bring stages 1 to 3 into the models your quants already run

Desk BuildWe build and run the whole chain with you, then hand it over

Fair value for every weekly and monthly contract, every run

For every weekly, monthly and quarterly contract we turn the ensembles into a price distribution and set it against the curve. The gap is the edge. The width of the distribution tells you how much to trust it, and the run-to-run change tells you whether the market is about to catch up.

Because the same chain also attributes P&L back to the weather, your desk can see which signals made money and which did not, contract by contract.

Illustration, not real data.

Alongside your quants

Desk Advisory

Your quants own the models. We bring the meteorology and the curve experience, and work with your team every month so the weather is used the right way.

Everything in Early Warning, plus the full data feed by CSV or API
Weather features built for W+1 to Q+1: regimes, cluster shares, run-to-run changes and inter-forcing spread from IRIE
Review of how your models use ensembles, biases and model disagreement
Working days with your desk every month, on site or remote

Annual agreement

Two engagements a year

Desk Build

For trading companies ready to take weekly and monthly trading from discretionary to systematic. We build your own algo for weekly and monthly contracts with you, from the ground up, and hand it over running in production.

The whole chain: data, model and signal through to execution, risk and governance
Fair-value distributions per contract from ensembles, renewables, gas and carbon
Run in production from day one, with monitoring, failover and SLAs
Handover to your team, then continued as Desk Advisory

Typically six to nine months. We take on two builds a year.

Intro call Scoping session with your desk Desk Advisory or Desk Build

Your strategy and your signals stay yours. We license the conpowr models and data that feed them, so nothing you build with us ends up with another desk.

Four ways to work with us

Priced per company, with a cap on named users. Live Briefing and Early Warning start with a one-week free trial, followed by a 90-day paid pilot. Desk Advisory and Desk Build start with a scoping session with your desk.

	Live BriefingDaily outlook and the live call	Early WarningEverything, plus alerts	Desk AdvisoryWith your quants	Desk BuildYour own curve algo
Daily outlook mail with the price chart	Included	Included	Included	Included
Live briefing every trading day, with recordings	Included	Included	Included	Included
Event Watch and model comparison	Included	Included	Included	Included
Monday week-ahead edition and monthly seasonal note	Included	Included	Included	Included
Early Warning and alerts on every level change	Not included	Included	Included	Included
Chat with the forecaster around model runs	Not included	Included	Included	Included
Quarterly hedge review against your book	Not included	Included	Included	Included
Data feed by CSV or API: scenarios, event probabilities, IRIE	Add-on	Add-on	Included	Included
Weather features built for your own curve models	Not included	Not included	Included	Included
Working days with your quants every month	Not included	Not included	Included	Included
Your own algo for weekly and monthly contracts, built and handed over in production	Not included	Not included	Not included	Included
How it starts	Free week, then 90-day pilot	Free week, then 90-day pilot	Scoping session	Scoping session
Term	Annual	Annual	Annual	6 to 9 months, then Advisory

Data for your own models

Scenario prices, event probabilities and the price with and without each event as CSV or API. Included in Desk Advisory and Desk Build, and available as an add-on to every other plan. Seasonal and hedging packages for origination and PPA desks. Neighbouring markets as spreads to Germany.

Judge us on how early we were

Every run, every event level and every alert is logged with a time stamp. We publish how far ahead of the market we were, and when we got it wrong.

Two meteorologists, 40+ years between them

conpowr is run by two equal partners with more than 40 years of combined experience from national weather services, private forecasting, energy analytics and the power market.

Jesper Thiesen

Co-founder, meteorologist with trading experience

Trained at DMI, with 20+ years bridging meteorology and energy markets. Founded the forecasting company ConWX in 2008 and sold it in 2021. Three years heading a power trading desk, trading weekly and monthly futures on EEX and the Nordic market.

2008Founded ConWX

2021Sold ConWX

2024Launched conpowr

3 yearsHead of a power trading desk, EEX and Nordic

Tijana Radovic

Co-founder, meteorologist and metocean specialist

20+ years of meteorological experience from national weather services, private forecasting companies and energy analytics. Responsible for long-term metocean assessments using hindcast and reanalysis datasets for offshore wind projects.

2002Started in meteorology

2009National Met Service, Serbia

2025Joined conpowr

Also from conpowr

The same expertise, available as project work alongside the trading service.

Offshore metocean studies

Long-term metocean assessments for offshore wind, from site screening to design basis and installation planning. We have many years of experience from the Baltic Sea and the North Sea, two very different environments: fetch-limited seas, stable stratification and low-level jets in the Baltic, and high waves, swell and storm surges in the North Sea.

Every study is built on long hindcast and reanalysis records, typically 30 years or more, and calibrated against site measurements from buoys, LiDAR or met masts where available. Extremes are estimated with established methods such as Gumbel and GEV fits to annual maxima and peak-over-threshold analysis, by direction and season, with the uncertainty stated alongside each value.

Example: Gumbel fit to annual maxima of omnidirectional 10 m wind, return levels out to more than 100 years — Example: Gumbel fit to annual maxima of omnidirectional 10 m wind, with return levels out to more than 100 years. Click to enlarge.

What a study typically delivers

Wind and wave climate at the site, including wind shear and wind-wave misalignment
1-, 10-, 50- and 100-year return values for wind, significant and maximum wave height, current and water level
Joint probability of wave height and period for load calculations
Weather windows, persistence and expected downtime for installation and O&M
A documented report with data sources, validation and uncertainty, aligned with IEC 61400-3-1 and DNV-RP-C205

NWP systems

Cloud-hosted numerical weather prediction at 1 to 10 km resolution with ensemble uncertainty, custom onshore and offshore domains and delivery by API. We run the high-resolution regional model IRIE with three boundary forcings, EC, GFS and AIFS, alongside WRF, and also deliver historical forecast runs for back-testing and training your own models. Data is available in hourly resolution out to 360 hours for any weather parameter.

IRIE example maps: precipitation and mean sea-level pressure out to T+360 hours, and 2 m temperature from the AIFS-forced run — IRIE example output: precipitation and pressure out to T+360 hours, and 2 m temperature from the AIFS-forced run. Maps: IRIE by ILMETEO. Click to enlarge.

Why three forcings through one model matter beyond correlation

Running EC, GFS and AIFS through the same regional physics turns the three runs into a controlled experiment. Because resolution, physics and post-processing are identical, the differences between them isolate the large-scale forcing uncertainty, which is impossible when comparing raw global models. For a desk this gives one consistent feature set and one post-processing chain across all three sources, more frequent updates that are directly comparable from run to run, and a high-resolution inter-forcing spread that can be used for position sizing and ramp risk. It also allows regime-dependent weighting of the forcings, early signals of where the EC-based market consensus is likely to be revised, and physically consistent local fields from AIFS, whose large-scale skill is otherwise hard to use at energy-relevant scales.

Climate risk advisory and talks

Climate impact analysis and risk scenarios for your sector, board presentations, and half-day or full-day sessions for leadership teams in Copenhagen, Hamburg or on site.

We see the weather shift before the power market does