EPRI and Salt River Project (SRP) combine historic weather data and climate models to support resource adequacy planning
Phoenix, Arizona, has always been hot. But the frequency and relative severity of extreme heat events are increasing. According to the National Weather Service (NWS), the state’s capital city experienced its earliest heat warning this past March, when temperatures soared above 100 degrees Fahrenheit for four consecutive days. As of early September, Phoenix recorded 44 heat warning days, the most on record and significantly above the 20.3-day average for the first eight months of the year.
Salt River Project (SRP), the largest electricity provider in the Phoenix metropolitan area, whose service area includes large swaths of central Arizona, is keenly aware that the exceptionally hot temperatures of 2026 are likely to become the norm in the future. The latest climate models project that Maricopa County, which includes Phoenix, could average more than 89 days above 105 degrees Fahrenheit by midcentury, an increase of roughly 30 percent compared to today’s average.
Reliably meeting electricity demand on an increasing number of hot days is one of the more obvious challenges faced by SRP planners. Also daunting is demand driven by data centers, which adds to the already robust load growth from the area’s expanding population and economy.
As the load SRP must serve grows, its generation portfolio is also evolving in ways that necessitate new approaches to resource adequacy planning. Indeed, renewable energy is playing an increasingly vital role in meeting the rapidly growing electricity demand of SRP customers today and into the future. In its most recent Integrated System Plan, the utility projected that it would need to more than double its resource capacity by 2035 to meet growing demand and sustainability goals.
SRP already has more than 1.6 gigawatts of solar and nearly 700 megawatts of wind serving its customers. Recently, it has secured several more gigawatts of renewable generation through procurements and agreements, including a development partnership expected to bring 3 gigawatts of solar online by 2034.
The Need for a Different Kind of Planning Data
SRP understood its planners needed a different set of data than they had traditionally relied on. Planners use software tools that simulate how a utility’s generation portfolio meets electricity demand under various conditions and scenarios. Understandably, those planning tools need hourly data to anticipate how temperatures, humidity, wind conditions, and other factors change through the course of the day. This type of granular detail is especially important to resource planners with large portfolios of renewable generation. That’s because it helps answer questions like how much energy storage will be needed to meet demand in the early evening, when the sun is setting and the air-conditioning load rises.
That type of hourly data exists. But it is based on decades of historical records, not what future conditions in a warming world may bring. One way to fill that gap is with climate models, which project how the planet will warm. The problem with feeding climate model data into planning tools is that most climate model outputs are in the form of daily values, such as high, low, and average temperatures, as well as a single metric for solar irradiance and wind speed. Those daily values are of less use to planners than hourly data.
“Planners don’t just want to know the peak load. They want to know how hourly weather will unfold and how temperature covaries with wind and solar, which is so important to SRP’s system,” said Delavane Diaz, EPRI program manager for Energy Systems and Climate Analysis. “There is a need to have a comprehensive set of temperature, humidity, wind, and solar data on an hourly basis that is forward-looking, not backward-looking.”
Combining the Past and the Future
EPRI and SRP worked together to develop a granular, forward-looking data set to help the utility’s resource adequacy and capacity planners. The basic approach is to take full advantage of existing historical weather data while adjusting it to reflect expected climate change.
“There’s a lot of really important information contained in that historical time series data,” Smith said. For example, past weather observations capture nuances that are hard to simulate, such as how a heat wave builds over several days, how temperature, wind, and sunlight interact, and how temperatures in summer afternoons unfold.
To get the data that planning models require, EPRI combined decades of historical weather records with climate models’ projections of how much warmer a given month is expected to become. “Our approach takes that historical hourly information and applies trends from a climate model while preserving the hourly chronology in the weather,” Diaz said. “The data is shifted based on the warming in a climate model.”
A Versatile, Flexible Approach
The result is what’s referred to as a synthetic weather year that maintains the nuances of an actual day—including its temperature swings and the relationships among heat, sunlight, and wind—while reflecting a hotter future. Using 74 years of historical weather data, EPRI produced these adjusted datasets across four climate scenarios, giving planners a range of plausible futures rather than a single forecast.
“The warming trend is, by and large, the most important component,” Smith said. Capturing that trend accurately, while preserving everything else from the historical record, was the core accomplishment of the project. The combination of historic data with climate models ensured temperature, humidity, wind, and solar interacted in realistic ways. For example, a hot, sunny, calm afternoon looks very different from a hot, cloudy, windy one. Maintaining that distinction is important to any utility trying to know how much solar and wind power will be available when demand peaks.
There are other ways to produce hourly future weather data. The most rigorous alternative is known as dynamical downscaling, which runs a full physics-based atmospheric simulation at high resolution. But its practical value is limited by computing costs and the time required to run the simulation. “It takes years to get a dynamical downscaling right,” Smith said.
That is because the simulation outputs typically need to be tuned by iteratively adjusting model parameters, and ultimately statistically bias-adjusted to produce the final outputs. “You’re doing this for months and years to try to get the simulation right, and then at the end, you still have bias that has to be corrected.” In addition, the heavy time and computational costs mean that dynamical downscaling can usually be run for only one or two climate scenarios at a time.
A Range of Possible Futures
Because EPRI’s statistical approach doesn’t require simulating the physical atmosphere the way dynamical downscaling does, it can be run quickly with dozens of climate models whenever robust historical weather records exist. Speed and low computing cost also make the method replicable. The same underlying method remains; only the utility’s historical weather and climate model changes.
EPRI and SRP worked together to run the newly created datasets through detailed validation checks. This involved comparing the weather adjusted based on climate models against the historical record to confirm it behaved the way real weather should. This was all part of a highly collaborative process between EPRI and SRP in which results and findings were presented regularly, and care was taken to ensure the data could be used in the utility’s planning tools. “Together, we were coordinating the development of data sets that could be used in the tools they use in-house,” Diaz said. That collaboration has continued since the project finished. Because SRP quickly implemented and began using the datasets, the utility has shared a few minor technical bugs that planners encountered, which EPRI has resolved.
One principle reinforced throughout the project is the value of a shared meteorological dataset across utility functional teams. “This is something we frequently emphasize in our conversations with companies,” Diaz said. “When your load forecasting, capacity expansion, and resource adequacy teams work from the same internally consistent weather and climate data, it improves coordination and supports more robust planning outcomes.”
EPRI’s collaboration with SRP isn’t the only initiative leveraging this new method. For instance, it has also been incorporated into EPRI’s Regional Economy, Greenhouse Gas, and Energy (REGEN) model to help researchers better understand how the evolving energy system performs in a changing climate. EPRI has also partnered with the Argonne National Laboratory to apply the approach to some of the lab’s new projects, including a resilience analysis focused on the Gulf Coast of Texas.
Other new tools are gaining momentum, most notably the use of artificial intelligence (AI) to produce hourly weather projections based on climate model outputs. Smith doesn’t see AI rendering EPRI’s approach obsolete. “It’s becoming more and more clear that this will have a place going forward, regardless of other products like dynamical downscaling and AI,” Smith said. The reason: speed, price, and a grounding in historical weather are features planners need more than ever as they prepare for a changing grid and warming world.
EPRI Technical Experts:
Erik Smith and Delavane Diaz
For more information, contact techexpert@eprijournal.com.
