Quantifying Energy Use across U.S. Water Sectors: A Scalable AI/ML-Driven Framework for Data Integration and Resilience Planning

This white paper explores opportunities to quantify and characterize energy use across major U.S. water sectors, including irrigation, water supply, and water treatment, with a focus on energy for water dimension. A scalable AI/ML-driven framework will be developed to harmonize fragmented datasets to support benchmarking, energy efficiency assessment, and energy-water (EW) resilience planning, with a pilot application of the framework in the western U.S.

Energy use data for water services in the conterminous United States (CONUS) are currently fragmented across multiple agencies, lack spatial, temporal, and sectoral consistency with limited availability of centralized reporting systems. Existing physics-based and statistical methods are limited by data availability, computational demand, and coarse resolution. A unified approach is needed to integrate diverse datasets and enhance the accuracy, granularity, and accessibility of energy-use estimates across sectors that can guide operational decision-making.

The proposed framework will harmonize existing and open data sources (e.g., EIA, USGS, USDA) and implement AI/ML and hybrid modeling approaches to fill data gaps and improve reliability of the estimates. Advances in AI architecture such as Graph Neural Networks and transfer learning?enable integration of multi-fidelity datasets, including existing estimates of energy use to generate scalable, high-resolution metrics. Leveraging ORNL?s high-performance computing capabilities, the framework will produce open-access datasets and decision-support tools aimed at enhancing EW resilience. A pilot project in a hydroelectricity-dominant, irrigation-heavy regions in the Pacific western US will demonstrate the framework?s scalability and practical utility.

Project success will be evaluated through four specific measurable outcomes:
(a) Development of an open-access, high-resolution energy-use dataset for U.S. water sectors
(b) Creation of a large training dataset for community-driven AI/ML enhancement
(c) Demonstration of a validated, scalable decision-support system, and
(d) Establishment of a reproducible, extensible data-integration and modeling workflow that can support national-scale EW resilience and planning strategies.