Tag: forecasting

Weather-Smart Fundamentals Modeling Gives More Meaningful Power Price Forecasts

Jessica Tomaszewski, author of Weather-Smart Fundamentals Modeling Gives More Meaningful Power Price Forecasts
Author: Jessica Tomaszewski
Senior Research Scientist

Forecasting of power prices is a part of everyday life for the renewables industry. Accurate power price forecasts are necessary to support the green energy transition by empowering investment, procurement, and financial-planning workflows of buyers, sellers, and investors of clean energy.

There are two primary types of models used to forecast power prices: Statistical Models and Fundamentals Models. Statistical models learn relationships between prices and other variables based on previously observed outcomes and apply these relationships to make predictions about the real world. Such models are useful in short-term, high-frequency workflows like near-term trading, where market behavior is not expected to deviate significantly from recent history. However, statistical models struggle in longer-term applications where evolving grid mixes, generation technologies, and market designs mean that the future will look different than the past. Fundamentals models can capture the impact of these changes, which is key for making long-term predictions of power prices.

Fundamentals models offer users the benefits of extendability, flexibility, and transparency in their predictions. However, the heavy computational burden of these models has traditionally required price forecasters to make simplifying assumptions about the weather. Much of the industry uses a “Weather-Normal” or “Typical Meteorological Year” to represent forward-looking weather conditions in their price models. But “atypical” meteorological conditions like heat waves, cold snaps, and severe storms can all cause dramatic surges in electricity demand, alter wind and solar supply, and affect prices. A price model that only uses a typical meteorological year will miss the extreme prices that come with extreme weather, resulting in a dramatically different modeled outcome from the true range of expected possibilities.

REsurety has taken a different, Weather-Smart approach: Our price model captures the hourly signal of 40 representative weather years in a computationally efficient way, unlocking several key benefits to customers of our fundamentals-based price models.

Key Benefits of Weather Variability in Price Models

A More Representative Mean

Weather variability is an important driver of the economics of any type of clean energy project, but storage is particularly sensitive to atypical weather conditions. For storage projects engaging in energy arbitrage, profitability relies on buying energy when prices are low and discharging it when prices are high. Energy arbitrage becomes more lucrative as the spread between high and low prices increases. Forecasting price using a weather-normal year results in inaccurate forecasts of storage value because it misses the extremes of price that drive value for storage resources. Simply put, weather-normal modeling materially undervalues the energy arbitrage revenue opportunity of storage projects.

When adequate weather variability is represented in a price model, the predicted mean value of a storage project will better reflect these high price events, as illustrated in the schematic below. Two cases are considered. The blue line represents the case with a distribution of daily forecasted storage project values created using prices produced by a Weather-Smart model fed with the signal of 40 weather years. The second case represents in green the distribution of daily storage values based on prices produced by modeling a single weather-normal year of data. While both distributions of storage project value report a similar median, the Weather-Smart distribution produces a mean value that is significantly higher than its median. In this case, the mean can serve as a simple, single quantity that distills the potential profitability of a storage project in a way that acknowledges the extreme events contributing to its value.

The benefit of Weather-Smart distributions: Including adequate weather variability produces a mean value that acknowledges extreme weather and price events.

Visibility into Range

REsurety’s fundamentals price model gives users unique insight into their price forecasts, and by modeling hourly weather variability representative of 40 years, we elevate this insight to give visibility into a broad range of outcomes. For example, a forecast driven by a weather-normal year of input data will give a forecasted wind capture rate for a typical summer, perhaps at 80%. But how can a clean energy buyer set expectations with their financial planning team in the event of a warmer-than-average summer with lower-than-average wind speeds that yields a capture rate of only 60%? A wide distribution of potential outcomes exists depending on the weather conditions, and visibility into this range of outcomes is important for making financial decisions and planning for downside scenarios. This approach is outlined in greater depth in another REsurety blog post.

Better-Informed Portfolio Optimization

Visibility into range provided by Weather-Smart price forecasts lends itself to better portfolio optimization as well. Portfolio risk mitigation is possible through understanding the tails of distributions of individual assets, which price models with awareness of adequate weather variability can provide. By optimizing portfolios to include assets that provide value in countering scenarios, the overall portfolio risk can be narrowed. For example, a forecasted summer month that has low wind speeds and high temperatures will likely be profitable for solar projects, which can help offset low wind value or Fixed Volume Swap losses. A Weather-Smart price model that is aware of such anomalous weather conditions allows for this kind of portfolio optimization.

About the author
Jessica Tomaszewski is an atmospheric scientist with experience in boundary layer meteorology, numerical weather prediction, and wind resource assessment. Prior to joining REsurety, Jessica completed a National Science Foundation Graduate Research Fellowship with a focus on simulating interactions between wind farms and the lower atmosphere, as well as two summer internships at NextEra Analytics investigating improvements to the wind farm wake modeling process. As a research scientist at REsurety, she builds and investigates new techniques for analyzing renewable resources and mitigating their financial risk.

Jessica holds a PhD and Master’s degree in Atmospheric and Oceanic Sciences from the University of Colorado. She also holds a Bachelor’s degree in Meteorology from the University of Oklahoma. Learn more about Jessica here.

Return to the blog post main menu.

REsurety’s Weather-Smart Fundamentals Power Price Forecasts are Now Available

Solar Wind Energy
Adam Reeve, author of REsurety's Weather-Smart Fundamentals Power Price Forecasts are Now Available
Adam Reeve
SVP of Software Solutions

Here at REsurety, we know how important accurate forecasting is. We know that our customers need credible, explainable predictions of the expected value and upside/downside risks of the value of clean energy in order to make long-term investment or procurement decisions. And for that reason, for the better part of the last ten years, our team has been developing and improving upon long-term power price and renewable generation forecasting models. Up until now, however, those forecasts have relied on machine-learning methods and have only been available on a limited basis to customers via our Advisory services. We’re excited to announce that as of today, we have released a new fundamentals forecasting model, and are making it available across all of our product and service offerings including our SaaS platform, REmap.

REsurety developed our latest fundamentals forecasts in order to give our customers unprecedented ease of access and confidence about the future value of their clean energy projects. With these newly released forecasts, you can:

  • Develop an optimal portfolio: simulate portfolio performance under a range of outcomes to develop/manage your portfolio.
  • Calculate project-specific forecasts: all of our forecasts are natively calculated using project-specific hourly generation, so you can calculate the expected performance of your project with one click in REmap.
  • Stress test: gain visibility into downside risk driven by weather variability or changes in market dynamics, e.g. increased storage penetration, a hot summer/winter, or the impact of high renewables build out.

The Importance of Weather Variability

The most distinguishing characteristic of REsurety’s forecasts is that we don’t just model a single weather-normal year (e.g., an 8760), because we know that models based on 8760s will likely overestimate value for renewables, underestimate value for storage, and underestimate variability across all projects. Instead, we simulate ~40 years of representative hourly weather – and the impact that has on every project and load center on the grid – to develop a thorough distribution of possible weather outcomes. Importantly, this means that hourly project-specific generation is an input into our model, as opposed to being calculated after the fact.

This extremely data-intensive and compute-hungry approach is designed to give customers the answers that they need about the future. Users can: run sophisticated portfolio simulations across projects and markets using realistic and consistent weather inputs; confidently calculate the value of storage, where profitability is highest during periods of extreme weather and market volatility; and calculate the expected value and downside risk in their PPAs for accurate budgeting.

Unlike traditional forecast providers, REsurety’s fundamentals-based forecasts realistically take into account a range of possible weather conditions and the impact that they have on each project in order to solve for power prices in each hour. The plot below shows the value of the approach: for each of the five market scenarios, 40 representative weather-years (represented by thin lines) are simulated in the model. We’re calling this realistic approach to weather variability “Weather-Smart.”

The value of full weather distributions: weather and various market scenarios drive variability in the capture rate for solar generators.
The value of full weather distributions: weather and various market scenarios drive variability in the capture rate for solar generators.

“We’re excited to bring together our strengths in Atmospheric Science and Power Market Analytics in this model release,” said Adam Reeve, SVP of Software Solutions at REsurety. “Traditionally, those two fields have been separate in the industry, limiting the ability for customers to apply forecasts to their clean energy projects or portfolios. This Weather-Smart approach gives users a much more robust way of forecasting the value of clean energy.”

Why A Fundamentals Model?

REsurety’s newest forecasts leverage an hourly production cost model that accurately represents the operational and market design complexities of the power markets. It takes into account the physical power flows, hourly generation from each renewable plant, hourly load, and future market conditions inputs to solve for hourly power prices. As an example, this means that, in each hour, we model the generation at every renewable plant on the grid (based on localized wind speeds / solar irradiance, turbine / panel type, etc.) as well as production costs for dispatchable generators. We also model load in each hour, as well as the transmission limitations of the grid and other market-specific characteristics. Given these inputs and constraints, we then solve for power prices in the same way that a system operator (such as ERCOT) would.

After years of creating advanced models, we’ve learned that such a rigorous approach has a number of advantages over machine-learning (ML) models. Specifically, ML models struggle to make accurate predictions about a future that may look very different from the recorded history – such as predicting price formation in a market with a rapidly changing installed base of grid-scale storage. Results from ML models are less interpretable, making it harder for customers to understand why a certain price was produced – and by extension whether it is reasonable or not. Lastly, ML approaches are less capable of accurately simulating how changes to market rules or regulatory policies will impact prices. For these reasons, REsurety has invested in the latest fundamentals-based model that we’re excited to release today.

REsurety’s Weather-Smart fundamentals power price forecasts are currently available in ERCOT, with CAISO available later this year and full market coverage by mid 2023.

To learn more, please visit http://resurety.com/remap or email [email protected].

Return to the blog post main menu.