6 min read

Machine Learning Unlocks More Grid Capacity—Without New Hardware

Power grids are under pressure. Electricity demand is rising, yet infrastructure expansion is slow and costly. The good news? We don’t always need more power lines—we just need to use the ones we have more efficiently. Dynamic Line Rating (DLR) makes this possible by adjusting transmission capacity in real-time. But for DLR to work, we need highly accurate wind forecasts at the exact location of power lines. That’s where Gridraven’s machine learning (ML) model comes in.

Eneli Toodu
Eneli Toodu
Machine Learning Unlocks More Grid Capacity—Without New Hardware

Why Wind Forecasts Matter for Grid Capacity

Transmission lines heat up when they carry electricity. The wind cools them down, allowing them to safely transport more power. The challenge is that traditional weather models—like those used in IEEE 738 and CIGRE 207 calculations—aren’t designed to capture how wind behaves at the exact height of power lines. Standard Numerical Weather Predictions (NWPs) focus on large-scale atmospheric patterns and don’t account for the trees, hills, and buildings that shape wind speeds near the ground. As a result, utilities either underestimate grid capacity or take on unnecessary risks.

A Smarter Approach: Machine Learning for Wind Forecasting

To fix this, we built an ML model that enhances wind predictions by combining:

* Numerical Weather Predictions (NWP) – Industry-standard weather forecasts from AROME, ICON, HRRR, IFS, and GFS.

* High-Resolution Geospatial Data – LiDAR and satellite mapping to capture terrain, trees, and buildings.

* 30,000+ Weather Stations – Real-world data to train and validate our model.

This allows us to improve wind speed predictions by up to 50% compared to other numerical weather predictions, particularly in complex environments where NWPs struggle—like forests, valleys, and urban areas.

* Lower equals better

Real-World Validation: More Accurate Forecasts, Safer Grid Operations

We put our model to the test on a 25 km, 110 kV transmission line in Estonia, running through dense forest. The result? 58.7% better accuracy compared to the MetNo NWP.

‍

We’ve also validated our approach in California, using data from utility weather stations. Our model consistently delivers precise wind speed estimates, with confidence intervals that quantify uncertainty—critical for making safe, real-time grid decisions.

What This Means for the Grid

By improving wind forecasts at the span level, we enable utilities to dynamically adjust power line capacity without installing new weather sensors. This unlocks:

More transmission capacity—without expensive infrastructure upgrades.
Better integration of renewables—by reducing congestion bottlenecks.
Higher grid reliability—with data-driven certainty about safe operating limits.

Instead of just predicting the weather, we’re making it work for the grid—smarter, safer, and more efficient.

‍