The Importance of Accurate Weather Prediction for Power Operations
1 / 2
Fig2-meteodrone-grand-forks-ND-meteomatics
Aaron Larson
Updated 12 min read
Power companies need precise weather forecasts for a variety of reasons. Several companies specialize in providing the type of weather information utilities need to accurately predict resource availability and manage operations. Weather can obviously have a big impact on power systems and plant operations. Therefore, meteorologists play an important role in keeping power companies informed on how load and production could be affected, as well as when and where outages may be expected. Power companies require detailed weather information—more than just a forecast of high and low temperatures for any given day. Energy marketers need up-to-the-minute cloud cover predictions to better understand how solar output will be impacted. They also require accurate wind forecasts to ensure backup resources are available to keep the grid stable around the clock. Knowing when and where a storm will strike can allow repair crews and supplies to be strategically staged so repairs can be completed in the quickest and most effective ways to restore power and return systems to normal. In other words, the benefits of accurate weather prediction for power professionals go well beyond simply knowing whether or not to grab an umbrella as they’re heading out the door.
How Power Companies Use Weather Data
Many large power companies have their own meteorogists on staff. Still, most find value in having outside support that can supplement and/or confirm what their internal experts are predicting. That’s why several companies, such as DTN, Meteomatics, and Vaisala, specialize in delivering the type of insightful weather information power companies need. “Without a doubt, energy is a weather-critical industry. Understanding the impact of weather on electricity generation, on electricity demand, and on transmission and distribution systems is critical for all power companies,” Pascal Storck, head of Renewable Energy with Vaisala, told POWER. In September 2022, Vaisala launched Xweather, a suite of services providing real-time and hyperlocal weather and environmental data to predict and solve challenges from lightning-triggered wildfires to weather-related car accidents. The company says Xweather’s advanced machine learning models and intelligent sensors help a broad range of industries including the energy and power sectors by providing new levels of data accuracy and actionable environmental insight. Said Storck, “On the supply side, the most obvious impact of weather is on renewable energy production, but there are other uses for the information too. Some customers use weather forecasts to schedule downtime for maintenance, avoiding windy or sunny days when they’d rather be producing power. Other customers need warning of extreme weather conditions to protect assets in case of lightning, strong winds, or hail—all of which can negatively impact both wind and solar.” Storck continued: “Power companies use very accurate short-term forecasts of energy output, for example, forecasts at five-minute intervals for the next two hours, to optimize energy storage systems and participate in energy imbalance markets. Participating in day-ahead energy markets is another major use case for weather data and energy forecasts.” “One of the focal points for DTN is working with utility emergency preparedness teams in order to help them better understand and forecast at-risk weather environmental hazards that are going to impact their overhead distribution operations, and understanding and communicating appropriately the outage impact risks,” Nic Wilson, director of product management for weather and climate risk with DTN, told POWER. DTN is a global data, analytics, and technology company with a staff of well-trained meteorologists, climatologists, and data analysts. “Another application is asset inspection,” said Wilson. “After a storm goes through, how does the utility prioritize where it’s going to do inspection along its lines for potential damage?” One way could be using DTN’s tools. Wilson suggested, for example, a company responsible for the operations and maintenance of wind farms could use DTN data to identify turbines that may have experienced blade damage during a weather event. With that insight, the company could proactively inspect for compromises to the fiberglass blades before the damage turned catastrophic. Extreme temperatures can strain the infrastructure and equipment used in fossil fuel and nuclear power plants too. Heatwaves and droughts can impact water supply and cooling, while severe storms (Figure 1) and flooding can damage critical infrastructure and safety systems, increasing the risk of accidents.
1. Hurricane-force winds damaged the rooftop solar panels mounted on this industrial building. Source: Envato Elements
“Weather also is a major driver of electricity demand. Extreme peaks in electricity demand are always due to extreme weather conditions that drive extremes in heating or cooling demand. An accurate forecast helps power companies anticipate these peaks,” Storck explained. “For transmission and distribution, weather can be both a boon and a bane,” Storck noted. “On a cold and windy day, power lines can handle more power due to the cooling effect of the wind. But too much wind may cause trees and vegetation to impact power lines, which is a leading cause of catastrophic wildfires, especially during dry and windy conditions,” he said.
Creating and Enhancing Weather Models
Meteomatics, meanwhile, is a company founded by Martin Fengler, who holds a doctorate in applied mathematics and used his expertise to develop several numerical weather prediction codes. Meteomatics opened in 2012 and is headquartered in St. Gallen, Switzerland. It has about 700 customers around the world. In April this year, Meteomatics launched a new North American division and appointed Paul Walsh, a former executive at The Weather Company, as CEO of the division. “We have a presence here on the ground in the U.S., and we are in the process of scaling up right now,” Walsh told POWER during an interview in late September. Meteomatics specializes in high-resolution commercial weather forecasting; power output forecasting for wind, solar, and hydro; weather data gathering from the lower atmosphere using Meteodrones; and weather data delivery via the Weather API. “One of the unique things that we bring to the table is a weather data platform that effectively enables our customers—many of whom are meteorologists—to very quickly access every global weather model available today. They’re able to access historical weather data and even climate-scale weather data via our platform,” Walsh explained. “One of the big advantages is that they can easily and efficiently download data via our technology. Inside the platform, in addition to historical weather, they can access short-range, medium-range, and long-range weather data. And they can do it in a way that is very simple, using something we’ve developed called medium-cache technology, which enables people to very quickly ingest the type of data that they need to create their own models that are specific to the work that they’re doing,” said Walsh. In global weather models, a mesh size of 10 kilometers (km) to 50 km is common. The American Global Forecast System (GFS), for example, has a resolution of about 25 km, and the forecast model of the European Centre for Medium-Range Weather Forecasts (ECMWF) has a resolution of about 10 km. However, for very precise forecasts, granular weather data is highly important. Notably, Meteomatics has developed an extremely high-resolution weather model for all of Europe called EURO1k. EURO1k is calculated by the company’s team of experts to a resolution of 1 km, which is unique for a weather model. “It’s very, very accurate,” said Walsh. While the company does not have that level of granularity in the U.S. at this time, Walsh said it is a goal. “We’ll be bringing that 1-km model to the U.S. probably over the next 18 months,” he said.
Weather Balloons Are Still a Thing (For Now)
Since the late 1930s, the National Oceanic and Atmospheric Administration (NOAA) National Weather Service (NWS) has taken upper air observations with radiosondes. A radiosonde is a small, expendable instrument package that is suspended below a large balloon inflated with hydrogen or helium gas. As the radiosonde rises at about 300 meters/minute, sensors on the radiosonde transmit pressure, temperature, relative humidity, and GPS position data each second. These sensors are linked to a radio transmitter that sends the sensor measurements to a sensitive ground tracking antenna. Wind speed and direction aloft are also obtained by tracking the position of the radiosonde in flight using GPS or a radio direction finding antenna. The data collected by radiosondes is used to create weather forecasts and climate models, as well as to study the Earth’s atmosphere. Worldwide, there are more than 800 upper-air observation stations. All observations are taken daily at the same time. When severe weather is expected additional soundings may be taken at a select number of stations. While radiosondes are important tools for collecting data, there are disadvantages to their use—mainly cost and environmental impact. With so many launches, the cost for radiosondes and balloons adds up. Furthermore, the balloon and radiosonde materials, which could fall back to the ground almost anywhere, can be hazardous to wildlife. Therefore, finding possible alternatives to weather balloons makes sense. Meteomatics believes drones—unmanned aerial vehicles—offer a viable solution. The company says its Meteodrones offer a more flexible and precise way to collect weather data. “A Meteodrone is effectively a weather balloon, only it’s a drone,” Walsh explained. “It’s autonomous. It’s engineered to go straight up. It measures the wind, temperature, humidity, and all the things you get from a weather balloon, only it’s a drone.” Accompanying the Meteodrone is a Meteobase (Figure 2). The Meteobase provides local support for the operation of the drone. It functions as a communication hub, facilitating seamless interaction between the pilot and the drone while also serving as a control center for autonomous flights. It includes a charging station for the Meteodrone and offers real-time visual oversight of the drone’s immediate surroundings thanks to strategically positioned cameras. The base is weatherproof and has an internal climate control system, with heating and air conditioning, to ensure optimal environmental conditions for the Meteodrone, its electrical components, and batteries.
2. GrandSKY, TruWeather Solutions, and Meteomatics installed a Meteobase equipped with a Meteodrone at Grand Forks Air Force Base in North Dakota. The base is shown here with the hangar open. Courtesy: Meteomatics
“We’ve got four of these set up in Switzerland,” Walsh explained. “We have a pilot who can sit in his house at a computer. He’ll check with air traffic control and basically hit a button. These things will fly up to about six kilometers and then come right back down. It takes about 20 minutes for the entire journey. It lands itself within the Meteobase. The Meteobase closes and then the data is immediately input into our 1-km forecast model. It’s the only commercially available drone of its type today.” Walsh said the company is currently working on the next iteration of the Meteodrone with the goal of reaching heights of more than 9 km.
Evolving Forecasts: AI and Machine Learning Gain Traction
Storck noted that many forecasting techniques have changed since he’s been with Vaisala. He said global weather forecasting datasets, such as those provided by ECMWF and NOAA, have become more accurate and much richer overall—higher resolution, vastly more information, and more frequent updates. Meanwhile, artificial intelligence (AI) and machine learning, which were once fringe technologies, are becoming mainstream. As such, probabilistic information and forecasts are also now more widely accepted to improve decision-making under uncertain conditions.
“AI and machine learning play a key role in our forecasting methodology. They are a core part of our strategy of being the best in the industry,” Storck said. “We combine machine learning and model training/retraining, in conjunction with human intervention to manually make model adjustments and to QC [quality control] input data. This is a continuous process that happens every day and requires a higher human touchpoint, all toward better results.”
Storck suggested competition with other solutions providers has benefited Vaisala, forcing it to improve products and services by adopting new technology. The company also sees collaboration with end-users as an important aspect in meeting customers’ needs.
“A typical collaboration for us with a utility begins with understanding their particular use case and challenge,” Storck explained. “Many conversations start with: ‘We need a renewable energy production forecast,’ but then through discussion we begin to understand that, for example, the utility really needs a short-term solar production forecast that is optimized for each five-minute interval starting from 60 minutes though 120 minutes in the future. Then, we get all the observations from the customer that matter—not just where the project is, but the actual measurements of wind/sun and energy production from their project. At that point, the magic happens. We use our archives of weather data and forecasts to train prediction models using machine learning to the observations the customer cares about to solve their particular use case,” he said.
“Our customers tell us that price—cost of ownership—and forecast accuracy matter. We are not the most expensive option in the market, nor are we the cheapest. However, we strive to be the best full-solution provider with the highest accuracy at a competitive price point for small organizations and scale for the growth of large organizations with large portfolios. We desire to be our customers’ vendor of choice, plain and simple,” said Storck.
Yet, even with the latest advances, models are not done evolving. “Unfortunately, the current accuracy level of longer-term forecasts is probably an order of magnitude less than what we can predict a few days out, but improvements are being made all the time,” Storck noted. “A few decades ago, we could barely predict tomorrow’s weather. Now, we get that same accuracy a few days in advance. With continued research and development, we very well might be able to accurately predict weather anomalies a few months in advance.”
Furthermore, forecasting advancements are unlikely to stop there. Researchers are also focused on getting a better understanding of global weather patterns and how climate change is and will affect regions of the world in the future. It’s clearly an exciting time for the industry with new and improved tools already well along in the development queue, and power companies stand to benefit from all of the work.
