Ahead of every front.

Frontwatt turns over 100 weather scenarios into probabilistic power forecasts for wind, solar and hydro plants, then picks the day-ahead schedule with the lowest expected imbalance cost.

Book a demoSee how it works

A cold front passes a 120 MW wind farm

Illustrative example with synthetic data

  • Median forecast (P50)
  • 50% range (P25–P75)
  • 80% range (P10–P90)
  • Actual output
  • Day-ahead schedule
A cold front passes a 120 MW wind farmDay-ahead wind power forecast for one day. The median falls from about 100 MW in the morning to about 30 MW after a cold front passes in the early afternoon. The 80% range is narrow in the morning and very wide during the arrival window from 11:00 to 16:00. Actual output dropped about an hour earlier than the median but stayed inside the range. The schedule stayed below the median through the arrival window.Cold front arrival window0306090120MW00:0003:0006:0009:0012:0015:0018:0021:0024:00
A cold front passes a 120 MW wind farmDay-ahead wind power forecast for one day. The median falls from about 100 MW in the morning to about 30 MW after a cold front passes in the early afternoon. The 80% range is narrow in the morning and very wide during the arrival window from 11:00 to 16:00. Actual output dropped about an hour earlier than the median but stayed inside the range. The schedule stayed below the median through the arrival window.Cold front arrival window0306090120MW00:0006:0012:0018:0024:00
A cold front passes a 120 MW wind farmDay-ahead wind power forecast for one day. The median falls from about 100 MW in the morning to about 30 MW after a cold front passes in the early afternoon. The 80% range is narrow in the morning and very wide during the arrival window from 11:00 to 16:00. Actual output dropped about an hour earlier than the median but stayed inside the range. The schedule stayed below the median through the arrival window.Cold front arrival window060120MW00:0006:0012:0018:0024:00

The range widens where the timing of the front is uncertain. The schedule leans low there, because on this day promising energy that might not arrive is the costlier mistake.

Hover over the chart or focus it and use the arrow keys to read each hour.

Show the numbers
Hourly values in MW, synthetic data
HourP10P25P50P75P90ActualSchedule
00:0088949910410810197
01:00909610110611010499
02:009197103108112106101
03:009298104109113103102
04:009097103108112105101
05:008895102107111100100
06:00859310010611010298
07:008291991051109796
08:007889971041099994
09:007285951031099491
10:006279921011088885
11:00487087981067476
12:00335679941045264
13:0022406686993448
14:0016285074922736
15:0014223760822428
16:0015213046662626
17:0018243141553129
18:0022293644553834
19:0027344149604439
20:0031384654654744
21:0034425058695348
22:0036445261725150
23:0037455362735551

Single numbers hide the risky hours

When a front moves through, a wind farm can lose most of its output in a few hours. A point forecast puts that drop at one hour and says nothing about how sure it is.

If the front comes early or late, the gap between your schedule and your meter is settled at imbalance prices, and in some markets a deviation penalty comes on top.

With Frontwatt you see

  • The likely range of output for every hour, not just a midpoint.
  • Where that range widens, so you know which hours carry the risk.
  • A schedule chosen for what being long or short will cost you.

How Frontwatt works

The same five steps run for every plant, each time a new weather model run is published.

  1. Sample the weather

    We read the latest runs of AI-based and physics-based global weather models at your plant and at every turbine. Together they describe over 100 equally plausible versions of tomorrow.

  2. Turn weather into power

    Each scenario runs through a model of your plant: turbine power curves at hub height with air-density correction, a full photovoltaic chain for solar, flow and reservoir behaviour for hydro. A machine-learning layer then calibrates the result on your own metered history.

  3. Keep the uncertainty

    You get P10 to P90 ranges for every hour. Portfolio totals are added up scenario by scenario, so the range reflects how your plants actually move together.

    What P10 to P90 means

  4. Choose the schedule

    For each hour, Frontwatt picks the schedule with the lowest expected cost under your market's imbalance prices and penalty bands, and exports it in the format your market expects.

    How the schedule is chosen

  5. Check the result

    After delivery, we compare forecast, schedule and metered output for every plant and calculate what the imbalance cost.

    How we measure accuracy

What’s included

For every technology

  • Wind

    Turbine-level modelling with each turbine’s own power curve and hub height.

  • Solar

    Irradiance, panel geometry, temperature and inverter clipping.

  • Hydro

    Recent output, rainfall and announced availability for run-of-river and reservoir plants.

  • Geothermal

    Base load corrected for ambient temperature.

For your team

  • Maintenance and curtailment

    Enter planned limits once. Every forecast and schedule respects them.

  • Exports and API

    Excel, CSV and ENTSO-E XML files, or pull everything through the REST API.

  • Data import

    Upload production history, outages and plant inventories as CSV or Excel files.

  • Daily scorecards

    Error against metered output and against your submitted schedule, plant by plant.

Measured, not marketed

Forecast providers rarely publish how they score themselves. Frontwatt stores every forecast with its timestamp and scores it against what your meters recorded, with the same rules for every plant.

Read the full methodology

Scored at your deadline
We score the forecast that existed when you had to decide, not a later revision.
Normalised to capacity
nMAE is the average hourly error divided by installed capacity, the yardstick the industry uses.
Gaps stay gaps
Hours without metered data are reported as missing. They are never filled with zeros.
Next to your schedule
Where your submitted schedule is available, its error appears next to ours.
New models earn their place
A model replaces the current one only after it wins on locked test periods that cover two seasons.

Built global, deployed local

The forecasting core works anywhere. What differs between markets is the schedule format, the price data and the penalty rules, so each market gets its own module.

MarketStatusWhat’s included
TürkiyeLiveKGÜP day-ahead schedules ready for submission, EPİAŞ prices and metered output, imbalance and KÜPST cost tracking.
EuropePlannedENTSO-E schedule documents and 15-minute resolution.
Other marketsOn requestProbabilistic forecasts for wind, solar and hydro plants. Tell us where your plants are.

How we handle your data

Your production and outage data is used to forecast your plants and for nothing else. It is never shared with other customers.

Each portfolio is kept separate from every other and has its own access keys.

See tomorrow’s range before you bid

Tell us about your plants. We’ll walk you through a live portfolio and show what a pilot with your data would look like.

Book a demo