GridCast: forecasting for power markets
GridCast forecasts the global grid: production mix, load and cross-border flows. Price forecasts emerge from this picture.
Bidding zones
Cross-border flows
New forecast
Horizon
Granularity
How GridCast works
Power forecasts built on the grid
Validated grid data. Price and fundamental forecasts across interconnected zones. Specialized models for every market, signal and horizon, continuously evaluated.
Fundamentals: Global
Prices: Europe, extending
Up to 72 h ahead
Available via API
Performance by zone
Quality metrics
A single quality metric can’t tell you how useful a forecast is in every scenario. We’re showing you three of the more commonly requested ones here, by zone, across our releases since June.
MAE
Average absolute gap between our forecast and the cleared day-ahead price, in €/MWh (£/MWh for GB).
RMSE
Root mean squared gap between forecast and cleared price, in €/MWh (£/MWh for GB). Weighs large misses, such as spikes, more heavily than MAE.
Absolute bias
Average signed gap, shown as an absolute value. It tells you how far off the forecast ran on average, not whether it ran above or below the cleared price.
Period
1 October 2025 to 31 July 2026, identical for every version and zone.
Spikes
Spike detection
A lot of the value in power markets is in price spikes. GridCast 1.5, now in testing, gets much closer to them than GridCast 1.4.
One backtest,
for us and for you
Our backtests are out of sample by design. We use a single methodology for our internal model development, and to issue backtests for our partners.
1
All delivery times included
Our backtests score every 15-minute MTU at every lead time, from 72 hours ahead to just before delivery, across every hourly run. Score us at the lead time you trade on, and see what each new run adds.
2
No data leakage
Each model is scored only on days it never trained on, with the inputs available at the time, exactly as it runs live. We know what every release improves, and you know what to expect at any reference time.
3
Any window you choose
Pick the periods that mattered most to your desk and we backtest those. Every new model faces the same windows as the last, so releases compare like for like.
Backtest
Contact us for a backtest
We backtest GridCast against the forecast you use today, on your choice of zones and time interval.