Precise Energy Load Forecasting: A Critical Need for Data Center Growth and AI Expansion

The growing demand for clean energy to power data centers, support industrial onshoring, and electrify transportation is projected to drive rapid, localized electric load growth across the United States. This increase in large point loads is challenging electric companies to keep pace.

For an electric sector undergoing transformative change, improved load forecasting can drive more efficient investment decisions, better grid performance, and reduced operating costs. Better projections regarding large loads and more accurate decision-making techniques—to build or not build long lead-time infrastructure, for example—are becoming increasingly important.

The Data Center Surge

A recent EPRI report analyzing data center energy consumption highlights that data centers represent one of the fastest-growing industries. Between 2017 and 2021, electricity consumption by major cloud service providers like Meta, Amazon, Microsoft, and Google more than doubled.

While AI applications were estimated to use only 10%–20% of data center electricity in 2023, that percentage is growing rapidly. As AI and other digital technologies (also growing rapidly) integrate more fully into every aspect of business and daily life, the importance of data centers becomes even more crucial.

AI models are typically much more energy-intensive than the data retrieval, streaming, and communications applications that drove data center growth over the past two decades. At 2.9 watt-hours per ChatGPT request, AI queries are estimated to require 10x the electricity of traditional Google queries, which use about 0.3 watt-hours each. Emerging, computation-intensive capabilities such as image, audio, and video generation have no precedent.

Forecasting Future Data Center Loads

The rapid development of data centers in certain regions can create local and regional supply challenges, necessitating swift grid investment and buildout to provide adequate network capacity. Coordination between data center developers and electric companies regarding power needs, timing, and delivery constraints is vital, as are efficiency improvements and advanced modeling tools.

Importance of Load Forecasting

Enhanced load forecasting is essential for efficient investment decisions, optimal grid performance, and reduced operating costs—especially as the electric sector undergoes transformative changes. Accurate load projections and improved decision-making techniques are critical as new load locations emerge and long-term infrastructure projects are considered.

EPRI’s Load Forecasting Initiative

EPRI’s Load Forecasting Initiative, launched late in 2023, addresses the critical need for improved load forecasting amidst drivers like electrification, extreme weather, and evolving customer behaviors. This initiative, divided into three workstreams, aims to reduce forecast uncertainty and enhance grid planning and operations:

1. Industry Coordination: A load forecasting interest group that is free for utilities to join, holds periodic webcasts, discussions, and surveys to share best practices and determine optimal approaches.

2. Long-term Forecasting for Planning: This workstream focuses on developing methodologies and guidance to incorporate new load drivers, such as data centers, and identifying the data and models needed to integrate large loads into forecasts better.

3. Short-term Forecasting for Operations: This segment aims to develop strategies to improve short-term forecast accuracy, especially during extreme weather events.

The initiative will also address other forecasting variables, such as the impact of distributed energy resources, electrification, and emerging technologies.

In summary, EPRI’s Load Forecasting Initiative aims to address the growing energy demands driven by data centers, AI, electrification, and extreme weather. Collaboration with data partners, including energy companies, regional transmission organizations, and independent system operators, will validate the ongoing research on operating systems. The Initiative will also serve as an important roadmap to help meet the evolving needs of the energy sector.

For more information, visit EPRI Load Forecasting Initiative.