My journey
At university, I studied Earth Sciences—a field that combines the wonders of meteorology, oceanography, and geology. It was during my thesis that I took my first steps into coding, thanks to my supervisor, who encouraged me to dive into Python. At first, coding felt like solving puzzles with pieces that didn’t quite fit, but it became one of my favorite tools over time. If you’re just starting, trust me—it gets fun, I promise! And in a field where data is endless, learning to code is a game-changer.
My first job at Planet OS set the foundation for my passion for geospatial data. I worked on managing datasets in Datahub, where I developed a deep appreciation for how diverse and fascinating geospatial datasets can be. But it wasn’t until I joined Gridraven that I discovered just HOW exciting this work could become!
So, what do I do at Gridraven?
At Gridraven, I work with a wide range of geospatial data—weather observations, digital elevation models (DEM), and numerical weather prediction (NWP) models. My job is finding, cleaning, and preparing these datasets to train machine learning models. The end goal? Precise wind forecasts that can help optimize the performance of high-voltage power lines.
Now, you might wonder: Why wind forecasts? Here’s the key—wind speed directly impacts how efficiently electric lines can carry power. Accurate forecasts allow for Dynamic Line Rating (DLR), which improves grid efficiency and reduces the risk of blackouts. Numerical weather models tend to heavily overestimate wind speeds in sheltered areas.
The Challenge and the Thrill
Working with geospatial data is like solving a giant, ever-changing puzzle. Every organization has its own way of formatting and sharing data, so no two datasets are alike. The preparation process—converting formats, dealing with gaps, and ensuring compatibility—can be daunting but also incredibly rewarding.
Once the data is ready, training the model is where the magic happens. You pour hours into preparing everything, and then you wait to see what the model will reveal. For me, the most exciting part is testing the model on real-world data. Imagine seeing how your predictions perform at the actual location of a power line—it’s a moment where all the hard work pays off. It can be very rewarding but sometimes also disappointing.
When I joined Gridraven just three months ago, I began working on a pilot project in Germany. At first, the results weren’t promising—I was working with data from just 25 observation stations and a relatively small area, and the predictions struggled to match reality. But scaling up to over 1,000 stations made all the difference. Now, the model isn’t just performing well; it’s outperforming standard numerical weather predictions by 35%. Seeing these results and knowing the effort behind them is incredibly satisfying.
And while our wind predictions are a crucial piece of the puzzle, the work doesn’t stop there. My talented colleagues take these forecasts and use them to calculate Dynamic Line Ratings (DLR), optimizing grid performance and pushing energy efficiency to the next level. But I won’t steal their thunder—stay tuned for a blog post from them where they’ll dive deeper into how these calculations work!
Why This Work Matters
What excites me about my work is its potential impact. Every improvement in wind forecasting translates to better grid efficiency, greener energy, and fewer interruptions in power supply. It’s not just about crunching numbers—it’s about creating solutions that make a tangible difference.
A Note to Anyone Starting Out
If you’re new to geospatial data or machine learning, remember: it’s okay to feel overwhelmed. Working with large, messy datasets can be frustrating, and coding doesn’t always come naturally. But when you see the results of your efforts—when your model performs better than expected or when your data helps solve a real-world problem—it’s all worth it.