Episode 2 - Data Science Mini-Series with Dr. Martin Fengler

Dr. Martin Fengler, a mathematician and CEO of Meteomatics, joins the hosts to explore the data science challenges of weather forecasting for the energy sector. The conversation begins with the fundamental difficulty of knowing the current atmospheric state, requiring a combination of observations and numerical models. Various data sources are discussed, including ground stations, radar, satellites, and innovative unmanned aerial vehicles (drones) developed by Meteomatics that can fly at high altitudes and in harsh conditions. The need to harmonize diverse data formats across different national weather services is emphasized. Meteomatics operates a 1 km resolution pan-European model, significantly finer than global models, and uses a supercomputer with about 40,000 CPU cores. They have built a proprietary in-memory database and a user-friendly Weather API that allows data scientists to access processed data with simple code. Applications in renewable energy forecasting, such as wind and solar power prediction, demand forecasting, and grid operations are described. The episode also touches on algorithmic trading and the importance of forecast accuracy. The expert shares insights into the trade-offs between horizontal and bespoke solutions, and the role of machine learning in refining forecasts. Overall, the episode provides a comprehensive overview of the intersection of weather science and energy data, highlighting both technical and business aspects.