Fletcher is a Ph.D. student in Environmental Engineering. Prior to coming to Stanford, he received a B.S. in Environmental Engineering from Cornell University, then worked as both a software and environmental engineer. He is interested in applying software best practices to water resources and energy issues. His research currently focuses on the optimization of wastewater treatment facilities' design and operation for demand response.
Team Members
Overview
Most pathways for decarbonizing the electricity sector by 2050 include substantial increases in intermittent renewable generation paired with several TWh of grid-scale energy storage. Grid scale storage includes the ability to shift or schedule large amounts of power over time and can be deployed via a diverse range of technologies, like batteries or pumped hydropower. This project aims to understand the degree to which we can use existing infrastructure as virtual storage, by scheduling the operations and consequently the power consumption of electrified industrial processes (e.g., water treatment plants, chemical manufacturing, data centers, etc.). Our group develops methods rooted in numerical optimization to design, operate, and evaluate energy flexibility strategies across a range of industrial sectors and geographic regions. By modeling these processes from the ground up, we aim to understand the cost and feasibility requirements to unlock more flexibility from existing infrastructure and strategies to design and deploy flexibility more effectively for next generation infrastructure.
Team Members
Overview
Digital twins and other digital solutions are transforming the planning, design, operation, and maintenance of water assets. Implementing these solutions is often slowed by data management activities including cleaning, storage, and querying. We identify three limitations of existing data management platforms: data inaccessibility, inadequate integration of data and metadata, and the absence of embedded data analysis capabilities. We introduce Python for Process Engineering Schema (PyPES), an object-oriented, open-source schema for water data management, to address these shortcomings.
Links
Supported By
Team Members
Overview
Recent advances in machine learning and artificial intelligence show great promise for process automation. The data-driven nature of many state-of-the-art solutions presents challenges along the data pipeline from collection to validation, such as unavailability, misidentification, fragmentation, and low quality of data. We propose a water treatment ontology, WaTr, coupled with a data management platform, Acquirium, to address these challenges. A formal ontology (e.g., WaTr) can automatically enforce metadata rules and enable querying by logical relationships instead of esoteric tag names to increase data availability and avoid misidentification. Acquirium leverages the power of a formal ontology with a more user-friendly experience, providing built-in data reconciliation and validation tools to reconcile fragmented data and improve data quality. WaTr and Acquirium can help to accelerate the adoption of cutting-edge digital water solutions by alleviating data management headaches.
Links
Supported By
Team Members
Overview
Data centers are among the fastest-growing electricity consumers in the U.S. and also require substantial cooling water, imposing significant economic and environmental impacts on water and energy systems. Because they can respond to changes in electricity prices, emissions, water availability, and utility incentive programs, data centers are strong candidates for flexible, dynamic operation—yet operators and policymakers lack the datasets needed to decide where to site facilities and how to operate them. This project develops a national, geospatial dataset spanning retail electricity and water costs, Scope 2 emissions, grid reliability, water supply sustainability risk, time-of-use tariffs, and demand response programs to inform data center siting and flexible operation. Preliminary analysis across the eight most common U.S. data center regions quantifies the tradeoffs between candidate locations—for example, low electricity costs in Phoenix are accompanied by high Scope 2 emissions and the most at-risk water supply—informing site-specific permitting, utility incentive design, and operational decisions.
Links
Supported By
Team Members
Overview
The U.S. power system is facing unprecedented load growth driven largely by data centers for AI and large-scale computing, carrying significant greenhouse gas and water-consumption footprints that depend heavily on where facilities are sited. Existing analyses predominantly rely on static, short-run emissions assessments that fail to capture the bi-directional interaction between large new loads and the grid's long-term evolution. This project couples data center load-growth scenarios—spanning energy efficiency, flexible operation, and siting—with a power grid capacity expansion model to determine how siting and operating decisions shape, and are shaped by, future investment in generation, storage, and transmission. By comparing uniformly distributed load, clustering in today's major markets, and DOE-recommended sites, we quantify the system-wide cost, emissions, and water-consumption implications of different deployment pathways, helping policymakers and utilities mitigate the economic and environmental costs of AI growth while enabling data centers to come online in reasonable timeframes.
Supported By
Team Members
Overview
Data on wastewater treatment plant’s (WWTP’s) installed processes enables regional and national emissions inventories and infrastructure planning, yet existing datasets like EPA's Clean Watershed Needs Survey (CWNS) suffer from data sparseness and update frequency. This project develops and validates a methodology combining large language models (LLMs) with formal ontologies to extract WWTP unit process configurations from unstructured regulatory permit documents. Applied to 609 California facilities, our ontology-structured LLM approach achieves an F1 score above 0.9 and reduces unit-process category-level error from 49% (CWNS baseline) to 13%. The methodology enables repeatable, automated extraction as permits are renewed and can be extended to other states and infrastructure types.
Links
Supported By
