Forecasting solar power
Solar power is playing a critical role in shaping a low-carbon future. Yet, one challenge persists: the sun isn't always shining. To better manage the grid and operate assets efficiently, highly accurate forecasting is essential.
At 3E, we’ve taken a major step forward in tackling this challenge by developping of a machine-learning forecasting framework, now validated through this report.
Validating 3E’s machine-learning forecasting framework
In our latest validation study, we evaluated the performance of 3E’s four forecasting products, all integrated within our SynaptiQ platform, across 38 solar plants in Belgium and the Netherlands. The study compared traditional physics-based forecasting models with our enhanced, machine-learning-driven simulations over a full operational year.
The results are clear:
· The integration of machine learning leads to significantly higher forecast accuracy, with trained simulation achieving an nRMSE of 8.84% compared to 14.62% for plain simulation.
· Bias dropped from 5.60% to -0.41% with a trained simulation, showing our model's ability to self-correct and adapt to site-specific conditions.
· 3E's forecasting framework maintains robust and consistent performance across all geographical locations and plant types.
· The operational forecasts provide reliable predictions for different time horizons (intraday and day-ahead), seasons and weather conditions, making the system a valuable tool for grid integration, energy trading and plant operations.
Download the full report
Want to dive deeper into our methodology, results and how our digital tools can support your solar operations?
👉 Download the full validation report here to explore how 3E's machine-learning approach is setting a new standard in solar power forecasting.
