Alexander Stauch:
Good morning, everyone. It’s two minutes past eleven, and we already see many excited participants in the room, so let’s get started.
Welcome to our webinar, Advanced Weather Intelligence for Grid Stability and Energy Planning. I’m Alexander Stauch, Head of Marketing at Meteomatics, and I’ll be moderating today’s session.
Let me quickly introduce our speakers. From Meteomatics, we have Markus Schwab and Rob Hutchinson from the energy team, and Stefanie Kieferle, Product Owner for our visualization platform MetX. Joining us as guest speakers are Jamie Bright from UK Power Networks and Christoph Glanzer from Swissgrid, who will share real-world examples of how they use Meteomatics weather data in their operations.
Today’s agenda: after a quick poll, Markus will discuss weather in a rapidly changing risk environment and present the Meteomatics vision for Weather Resilience 2.0. We’ll review a case study on Storm Eowyn, then hear from Jamie at UK Power Networks and Christoph at Swissgrid on their applications of our weather data.
The webinar is recorded and you’ll receive the replay by email tomorrow. Please use the Q&A function to ask questions—we’ll address them at the end.
Let’s start with a short interactive poll. Three questions:
-
Which extreme weather events are most challenging for power outages and emergency planning?
-
Which weather events are most challenging for solar power forecasting and control strategies?
-
Which weather events are most challenging for wind power production?
You can select more than one answer for each. We’ll look at the results together in about 20 seconds.
(After responses are collected)
About half of you have answered. For power outages and emergency planning, storms and thunderstorms clearly stand out as the main challenge. For solar forecasting, snow, ice, low stratus clouds, and fog are key difficulties. For wind power production, windstorms, high-speed turbine cutouts, and turbine blade icing are major concerns.
These results frame today’s discussion. We’ll explore how Meteomatics tools—especially MetX and our high-resolution weather models—can help tackle these challenges. Without further delay, I’ll hand over to Markus.
Markus Schwab:
Thank you, Alex. I’m Markus Schwab, Weather Expert and Climatologist at Meteomatics.
I like to start with the image of a red monkey to illustrate adaptability—symbolic of how grids must adjust to a rapidly changing environment shaped by both increased renewable power integration and climate change. Severe and small-scale weather events are becoming more frequent, posing new forecasting challenges.
For solar power, critical small-scale events include convective thunderstorms in summer, fog and Sahara dust in spring and autumn, and snow or icing in winter. For wind power, thunderstorms with gusts, high storm shutdown speeds, and wake effects in offshore wind farms are key risks.
Grid stability faces additional threats: severe summer thunderstorms can trigger overvoltage outages, downbursts, or even tornadoes. Flooding events—like those in Germany, Eastern Europe, or Valencia—can damage substations. Heat waves and droughts increase wildfire risk. In winter, freezing rain and wet snow combined with wind can cause conductor galloping and structural damage.
All of these phenomena require more sophisticated grid operations: control energy, redispatch mechanisms, and sometimes reserve power plants to maintain N-1 grid security. This is increasingly urgent as renewable feed-in grows and climate change drives higher temperatures, more frequent heat waves, and heavier precipitation.
Our vision of a “perfect weather-resilient world” involves three pillars:
-
High-performance computing to run rapid-refresh, high-resolution models that capture small-scale, fast-changing weather events.
-
Real-time data access—what I call a “stargate to the weather world”—so operators can query the data instantly.
-
Flexible visualization for intuitive understanding and actionable decision-making.
Combining these ensures accurate, real-time, and scalable weather intelligence—what we call Weather Experience 2.0.
That’s my introduction. Now I’ll hand over to Rob to dive deeper into how we implement this vision.
Rob Hutchinson:
Thanks, Markus. I’m Rob Hutchinson, leading the Energy & Utilities team at Meteomatics.
Our design philosophy is to give data scientists, integrators, and developers seamless access to the entire universe of weather data through a single standardized Weather API. Built on this API, MetX offers easy, asset-specific visualization, ensuring consistent data whether it’s used for planning, forecasting, or data science.
What sets Meteomatics apart is our proprietary high-resolution weather model, EURO1k, updated hourly with a native resolution of 1 km and downscalable to 90 m. For rapidly evolving scenarios such as convective storms, this level of detail provides a significant forecasting edge. Our forthcoming US1k model will bring the same capabilities to North America.
We ingest data from traditional weather stations, satellites, high-resolution radar, lightning and oceanic measurements, and our own Meteodrones—custom drones reaching up to 6,000 m today and 10,000 m in the next generation—to close the critical data gap in the lower atmosphere.
On the modeling side, we integrate around 30 third-party sources, including the UK Met Office and the ECMWF (the meteorological gold standard), and we are experimenting with next-generation AI models in partnership with NVIDIA.
Our API offers over 1,800 parameters, including industry-specific indices such as line-icing risk, conductor galloping, soil moisture, drought and fire-weather indices. Data is accessible in multiple formats (JSON, XML, CSV, PNG, HTML map, vector tiles) and via connectors for Python, ArcGIS, and more—about 20 integrations in total.
Queries can target points, polylines (e.g., transmission lines for dynamic line rating), grids (down to 90 m), or polygons (useful for hydrological catchments or operational areas).
Validation shows that EURO1k consistently reduces wind forecast errors by 10–15% compared to ECMWF.
Let me illustrate with Storm Eowyn, which struck Ireland and Northern Ireland with wind speeds up to 114 mph on January 24. About one-third of Ireland lost power, prompting a Europe-wide restoration effort. Using MetX, we provided forecasts with 140 km/h gust warnings five days in advance and operationally visualized impacts on high-voltage power lines.
This case demonstrates how MetX and EURO1k support both real-time operations and probabilistic risk planning, improving resilience and reducing economic and environmental costs.
That concludes my part. I’ll hand over to Stefanie for a look at upcoming MetX developments.
Stefanie Kieferle:
Thank you, Rob. I’ll give a brief outlook on what’s next for MetX.
First, we’re introducing a time-navigation slider, complementing the existing time-step buttons (from 5 minutes to 1 day). The slider allows faster, more customizable browsing of high-resolution data.
Second, we’re enabling self-service GeoJSON upload and management. Users will be able to upload their own polygons, lines, or areas directly through the UI and access public GeoJSON layers organized by category, which can then be visualized in dashboards.
Additional upcoming features include:
-
Integration of natural event layers (floods, earthquakes) combined with warning systems.
-
Enhanced animation controls with multiple loop options, pre-set date/time ranges, and custom color maps.
-
Expanded parameter customization for richer, user-defined visualizations.
With these enhancements, MetX will provide even greater flexibility for weather-dependent planning and operations.
Alexander Stauch:
Thank you, Stefanie. This brings us to the customer case studies. We’re very grateful to UK Power Networks and Swissgrid for sharing how they leverage Meteomatics weather data. I’ll now hand over to Jamie Bright from UK Power Networks.
Jamie Bright:
Thanks, Alex. I’m Jamie Bright, DSO Data Science and Development Manager at UK Power Networks. A big part of my role is building software and capabilities to operate our network more efficiently.
My background is in weather data and energy meteorology, including a PhD in synthetic solar irradiance for forecasting. I built Singapore Power’s first solar PV forecasting system and know firsthand how difficult convective clouds and thunderstorms are to predict.
At UK Power Networks, we subscribe to Meteomatics’ Weather API, especially the ensemble offerings, to improve risk management with probabilistic forecasts. We update forecasts every 30 minutes, looking up to seven days ahead, training individual models for each monitored asset—solar farms, wind farms, gas turbines, circuits, or substations.
Our system tests multiple models and retrains regularly to adapt to changes such as new PV installations. Because network operations are fundamentally about risk management, we rely heavily on probabilistic forecasting.
A key application is dynamic outage management. Traditionally, planned outages are scheduled a year ahead using worst-case assumptions, but predicting the sun a year in advance is impossible. Instead, we now re-evaluate outages one week or even one day ahead, incorporating real-time weather data to refine scenarios and reduce uncertainty.
In our first implementation, we combined Meteomatics forecasts with operational data to dynamically restore capacity. For example, during one outage in August, we instructed seven wind and solar farms to temporarily resume production. This avoided roughly £4 million in lost generation and actually enabled about £2 million in additional revenue while offsetting an estimated 4,500 tons of CO₂.
Since that pilot, we’ve applied this approach more than 30 times, making it a standard part of operations. The result: higher customer satisfaction, better renewable integration, and more informed, flexible decision-making powered by Meteomatics weather data.
That’s the overview. I’m always happy to answer more questions about how we do this. Thank you for listening to that case study—I hope it raised some questions for discussion.
Rob Hutchinson:
Thank you very much, Jamie. That was a fascinating insight into your work at UK Power Networks. Over now to Christoph Glanzer at Swissgrid. As I understand it, this is a different kind of use case, but again focused on renewable power production forecasting.
Christoph Glanzer:
Exactly. Thank you, Rob. Let me quickly share my slides—just made some last-minute changes. Can you see only my slides? Perfect.
I’m Christoph, or Chris, and I work at Swissgrid, Switzerland’s transmission system operator (TSO). We run and own the country’s highest-voltage power lines. Our core mission isn’t profit but stable operation of the grid for everyone in and around Switzerland.
I’m a mathematician by training but work as a data scientist. One of my main projects is our solar energy (PV) forecast, which I’ll present now.
To start, here’s a video showing PV installations in Switzerland over time. Yellow dots indicate new systems; dot size reflects installed capacity. Installations accelerate dramatically around the 2000s, primarily in the valleys. This map is already a year old: the video shows 233,000 registered sites, but today it’s about 280,000.
This rapid growth brings new challenges not only for us but for the entire industry. In 2023 alone, Switzerland installed 1.6 GW of solar capacity—roughly equivalent to one and a half nuclear power plants. One third of all existing PV panels were installed in 2023.
Traditionally, solar forecasting is done by balance groups—large energy firms responsible for matching production and consumption and trading to maintain a zero balance. But as PV capacity surges, the TSO must also forecast solar output because it now significantly impacts grid operations.
Forecast errors create imbalances. To improve visibility and stabilize the grid, we launched a project to produce high-resolution PV forecasts. We don’t build models from scratch; we rely on a strong partner and integrate the data with our internal use cases.
Switzerland has a unique advantage: an open-source database called Pronovo covering roughly 80 % of all PV plants. Combined with Meteomatics’ Weather API, we generate a forecast for each registered PV plant—about 280,000 hourly forecasts. High resolution lets us aggregate flexibly and validate forecasts against available measurements at system level.
Currently we run hourly forecasts and also backcasts—querying yesterday’s data through the Meteomatics API—because we lack legal access to measurement data from every PV plant. Backcasts help us analyze past events and refine forecasts. We also forecast up to two weeks ahead.
Forecasts feed directly into our Databricks data platform. We track how forecast values change over time to spot high-uncertainty days and alert control room colleagues to potential imbalances.
Ultimately, PV forecasting and backcasting not only improve operations but also support policy discussions on controlling PV output for grid stability. It’s like driving a car: you need to see ahead, verify where you’ve been, and be ready to steer differently if conditions change.
That concludes my presentation. Questions are welcome at the end. I’ll hand back to Rob or Markus.
Markus Schwab:
Thank you, Christoph. As mentioned, we’ll now move to the Q&A. We have a few minutes left, so let’s start with the first question.
Alexander Stauch:
The first question comes from Michael regarding the API: Is it possible to query forecast data from a specific past time—not a hindcast but the actual forecast that was issued for the next day at that past time?
Rob Hutchinson:
Yes. We archive forecasts from selected key models. Because of data volume we can’t store every run of every model globally, but we do retain important model runs in forecast mode. We’d be happy to share details after the webinar.
Alexander Stauch:
Thank you, Rob. Next question: In a polyline query, is it possible to choose different heights for different points?
Rob Hutchinson:
Yes. We interpolate all models in 1 m increments from ground level up to 20 km, so you can specify different heights for each point.
Alexander Stauch:
Great. The next question is for Christoph from Swissgrid: What type of models do you use for forecasting?
Christoph Glanzer:
We let Meteomatics decide which model is optimal using their Mix model approach. Currently we rely on EURO1k, a 1 km × 1 km grid model. For longer horizons the system automatically switches to the best available model. It’s essentially “fire and forget.”
Alexander Stauch:
Thank you. Another participant asks two related questions: What is the optimum latency and lead-time requirement for forecasts used in load balancing and outage planning? And what impact does space weather have on power grids—are current forecasts sufficient?
Jamie Bright:
I’ll take the first. It depends on the use case. For our half-hourly updating forecasts we aim for less than 30 minutes latency; anything longer can create issues. We used to work with 6-hour latency data, but moving to 1 hour with Rob’s help improved short-term accuracy significantly.
For day-ahead flexibility markets, we can spend more time producing a forecast as long as it’s ready for the daily 10 a.m. market instruction. Outage planning is similar—we generally plan at least a day ahead. Control methods vary across our 600 generators: some can be switched automatically with second-by-second instructions, others require a manual breaker operation. So optimum latency always depends on the specific operational need.
As for geomagnetic storms, I’m not a space-weather specialist, but severe storms can affect power grids. Modern grids have more protection today. Space-weather indices are available via the Meteomatics API and could be incorporated into models if needed.
Alexander Stauch:
Thank you, Jamie. The next question: How do you predict and manage real-time data in highly variable environments like India, and how do you measure accuracy?
Rob Hutchinson:
Two aspects: verification and data provision. We continuously verify model performance with in-house tools. Meteomatics ingests global data and the API returns the best possible model mix for any location and parameter at any time. India is challenging during monsoon season, but the mixed-model approach accounts for regional differences and maintains accuracy.
Alexander Stauch:
Final question: Is it possible to receive results from multiple models for comparison when making a forecast request, or is only the best model provided?
Markus Schwab:
Both options are available. You can use the mixed-model mode, which selects the best models automatically, or you can request forecasts from all available models to compare them yourself. Meteomatics gives you full flexibility.
Alexander Stauch:
That brings us to the end of today’s webinar. Many thanks to all speakers—especially Christoph from Swissgrid—for their contributions. We’ll send out the recording this afternoon or tomorrow morning. You can also book an expert call directly from the follow-up email or via our website.
Thank you all for joining and have a great afternoon.
