The central topic of this seminar is modeling approaches to facilitate resource conservation and a just energy transition. Potential subtopics are an emerging technology’s potential for scaling, life-cycle assessment for measuring social and environmental impacts, uncertainty quantification, and economic modeling for the energy transition. Our goal is to create an intimate, collaborative space for students, postdocs, scientists, and PIs within the Stanford techno-economic modeling and systems modeling community. These seminars will provide an opportunity to disseminate insights from your studies, connect with fellow researchers, and strengthen bonds across the community.

This week's speaker is:

WIlliam Xu, Ph.D. Candidate, Materials Science and Engineering, Stanford University
"AI-Driven Optimization under Uncertainty for Mineral Processing"

Abstract:

The clean energy transition demands a significant expansion of mineral processing capacity. Yet, conventional deterministic optimization in mineral processing often ignores key uncertainties—such as feedstock variability and process dynamics—leading to inefficiencies, excessive waste, and environmental harm. This work introduces a sequential decision making under uncertainty (DMU) approach, modeling mineral processing as a partially observable Markov decision process (POMDP). This framework accounts for uncertainty and helps identify what data to collect, where, how often, and at what accuracy to optimize process design and operation.

As a case study, we focus on phosphate mining in Morocco, which holds 70% of global reserves and is a major phosphate producer. Phosphate is vital for fertilizer and increasingly used in lithium iron phosphate (LFP) batteries. However, Morocco discharges phosphogypsum (PG)—a waste byproduct with residual phosphate—into the Atlantic Ocean, posing environmental risks and missing value opportunities.

We apply the POMDP framework to froth flotation, a key phosphate processing step, capturing uncertainties in feedstock and internal process dynamics. This DMU-based optimization improves both economic value and sustainability, offering a scalable approach for redesigning the integrated phosphate value chain with circularity in mind.

Bio:

William Xu (he/him) is a PhD candidate in Materials Science and Engineering. He works on making sourcing critical minerals for the clean energy transition more secure, sustainable, efficient, and equitable as a member of Mineral-X. Specifically, he is developing a general computational framework for optimizing mineral processing under uncertainty, and is applying this framework to improve the efficiency and reduce waste of phosphate mining in Morocco.

As a RAISE fellow, William is working with communities in the Salton Sea region of southern California to model scenarios that inform state policymakers about the potential impacts of the planned Lithium Valley industrial hub. The project aims to ensure that plans for a lithium extraction and battery manufacturing hub are aligned with community needs and interests. On campus, William serves as the Vice President of the Stanford Science Policy Group and The Dish on Science, a science communication club. He is also a participant in the Woods Institute’s Rising Environmental Leaders Program (RELP).