The Energy Systems Integration Group (ESIG) has released a new report, Forecasting for Large Loads: Current Practices and Recommendations, reviewing utilities’ and system operators’ approaches to forecasting large loads -- including from data centers, oil and gas, emerging industries, and manufacturing -- and providing recommendations for improving forecast accuracy.
“After two decades of relatively flat demand, utilities are now forecasting that the U.S. electricity system will grow at its fastest pace since the 1960s. Data centers are expected to account for roughly half of this growth and other emerging large loads accounting for most of the rest,” ESIG said.
“However, the pace and scale of this change are uncertain, making it difficult for utilities and system operators to accurately forecast these loads to better inform resource and transmission planning,” it said.
“Forecasting the extent to which new loads will emerge and grow is extremely challenging because of the complex characteristics of new loads,” said Trieu Mai, visiting fellow at ESIG. “The sheer number of requests, rapid advancements in AI technologies, and evolving business practices are examples of challenges utilities need to consider.”
Inaccuracies in load forecasts have important consequences -- for consumers’ costs as well as for grid reliability, ESIG said.
Load forecasts that underestimate actual demand can lead to resource adequacy risks and a risk of underserving customers due to insufficient infrastructure, it noted.
“Conversely, load forecasts that over-estimate demand could result in higher electricity prices to utility customers because infrastructure could be built but not ultimately needed. Utilities need accurate forecasts in order to efficiently plan and construct the necessary generation, transmission, and other facilities needed to serve new large loads together with other loads,” ESIG said.
”Large load forecasting is rapidly developing,” said John D. Wilson, of Grid Strategies LLC and project team lead. “The study and recommendations benefited from the active engagement of about two dozen utilities and many other experts. Everyone is eager to learn and improve.”
The Load Forecasting Project Team for ESIG’s Large Loads Task Force offered the following nine findings:
• Finding 1: Large load forecast methods lack transparency and consistency.
• Finding 2: Customer-supplied data and historical data are currently insufficient for accurate forecasting.
• Finding 3. Load forecasts are using weighting methods and thresholds to evaluate prospective load.
• Finding 4. Some large load forecast practices insufficiently differentiate between types of large loads.
• Finding 5. Data centers have distinctive characteristics that increase uncertainty compared to other large loads.
• Finding 6: Data center developers do not generally share alternative site locations and plans with utilities.
• Finding 7: Most large load forecasts do not have much geographical detail for all proposed projects, only for contracted projects.
• Finding 8: Utility rate tariff reforms are helping to reduce load forecast uncertainty.
• Finding 9: Few large load forecasts include any meaningful consideration of load flexibility.
The report recommends that the industry adopt clear definitions for five core metrics that large load forecasts can use to characterize and evaluate new loads: (1) project realization (the rate at which projects included in the load forecast are placed in service), (2) energization date (the schedule for when a load will be placed into commercial operation), (3) load realization (the forecast peak load once new load projects fully ramp up to maximum operation), (4) load ramping (the monthly or annual forecast of demand from initial project energization to full forecast peak load), and (5) load factor (actual energy use as a proportion of peak demand) and load shape (more detailed energy use information such as an hourly schedule).
“Together, these describe whether, when, and how completely large load projects materialize and how they use electricity over time,” ESIG said.
The Large Loads Task Force made the following 10 recommendations on load forecasting (the report offers additional details on the 10 recommendations):
• Recommendation 1: Use all five large load metrics to create a large load forecast.
• Recommendation 2: Develop a consistent framework to differentiate among types of large loads.
• Recommendation 3: Account for uncertainty
• Recommendation 4: Increase certainty through large load financial requirements.
• Recommendation 5: Reduce uncertainty in regional large load forecast practices
• Recommendation 6: Improve geographical detail.
• Recommendation 7: Seek continuous improvement through forecast validation.
• Recommendation 8: Collect large load forecast data in a shared database.
• Recommendation 9: Apply consistent loadweighting and modeling practices.
• Recommendation 10: Adopt forecast standards for load flexibility.
The ESIG Large Loads Task Force was formed to assist the power industry in addressing new challenges introduced by the rapid proliferation of large electronic loads such as data centers, as well as other large loads including manufacturing, electric vehicle fleets, and hydrogen production.
