Energy Storage
Electrochemical Energy Storage
✕
Electrochemical Energy Storage
We address battery lifetime prediction and operational reliability through integrated approaches that
combine physics-based modeling with careful feature engineering and machine learning transfer techniques.
This enables models to respect physical laws while adapting across different battery chemistries, formats,
and usage patterns — tackling the heterogeneous, incomplete, and event-driven nature of field data to
improve uncertainty quantification in remaining-useful-life estimates. Our work produces non-invasive
diagnostics and digital representations revealing thermal dynamics, aging precursors, and optimal charging
approaches within practical constraints.
Data Management — build standardized, cross-chemistry datasets and robust health indicators from noisy field logs.Advanced Modeling — embed electrochemical constraints and transfer models across cells, packs, and use cases.Predictive Analytics — quantify prediction intervals for RUL/SOH and propagate to risk-aware maintenance.System Optimization — link diagnostics to charging, warranty, and asset-management policies.
Associated members: Sangjun Jeon , Yooil Son
Hydrogen & Energy Systems
Hydrogen & Sector-Coupled Energy Systems
✕
Hydrogen & Sector-Coupled Energy Systems
Hydrogen production from renewable sources faces obstacles including high costs, infrastructure
limitations, and significant uncertainty from weather patterns and market volatility. Our approach
integrates process-level physics with economic investment and operational models, examining how
grid-scale batteries can improve performance. Using scenario-based and stochastic optimization, we
develop strategies that maintain cost-effectiveness despite variability while revealing where policy
support provides maximum benefit. Our broader vision extends to coordinating electricity, heating, and
transportation systems for a sustainable energy transition.
Techno-economic Assessment — compare renewable H₂ pathways and levelized costs under weather/price variability.Infrastructure & Siting — optimize pipelines, storage, and ports with spatial constraints and permitting limits.Electrolyzer–Battery Dispatch — co-optimize sizing and market participation for reliability and cost.Sector Coupling & Planning — integrate electricity–heat–transport for robust decarbonization roadmaps.
Associated members: Sunwoo Kim , Hana Kim , Hae Yong Jung , Byoungyoun Lee
Separation
Membrane & Separation Processes
✕
Membrane & Separation Processes
We address challenges in membrane processes used for water treatment and carbon capture by combining
computational fluid dynamics with physics-informed neural networks to better understand flow dynamics
in spacer-filled channels. This approach enables optimization of spacer geometry and operating
conditions while balancing flux, energy consumption, and fouling resistance. Our models support
digital-twin development that translates insights from channel-scale studies into practical guidance
for module and plant-level operations.
CFD → PINN Workflow — generate channel-level datasets and train PINNs that respect continuity/momentum constraints.Geometry & Operation Optimization — co-design spacer patterns and operating windows for flux, pressure drop, and fouling risk.Risk & Sustainability — predict fouling/cleaning trade-offs and connect to TEA/LCA for process-level decisions.Digital Twin & Scale-up — translate channel models to module/system design and real-time monitoring.
Associated members: Sungjin Bae , Jongwoo Kim
AI & Process Design
Process Systems Synthesis & AI-Driven Design
✕
Process Systems Synthesis & AI-Driven Design
We integrate design and operational considerations rather than optimizing them sequentially. We construct
comprehensive superstructures that enumerate possible units, pathways, and connections, then solve
mixed-integer nonlinear programs to identify reconfigurable networks capable of operating across diverse
conditions. Our methodology employs inverse-design and Bayesian-optimization techniques to guide
discovery while minimizing computational requirements; decomposition and convexification maintain
mathematical tractability. Planning and scheduling components ensure strategic designs accommodate
real-world constraints including maintenance and logistics — producing flowsheets that are optimal
on paper, resilient in practice, and defensible economically and environmentally under uncertainty.
Superstructure Enumeration — encode alternatives for units, connections, and recycle structures.MINLP/MILP & Decomposition — exploit convexification, cuts, and hierarchical strategies for tractable solves.Bayesian Inverse Design — learn surrogate objectives/constraints and search flowsheets efficiently.Planning & Scheduling — couple strategic designs to production/maintenance under uncertainty.
Associated members: Naeun Choi , Sunwook Kim , Hyunji Kwon
Design, Control & AI
Graph-Theoretic Co-Design and AI-Augmented Optimization
✕
Graph-Theoretic Co-Design and AI-Augmented Optimization
This research develops automated, graph-theoretic methods for simultaneous process design and control
structure selection, addressing the operational risks of narrow safety margins in sequentially optimized
designs. It also develops an AI-augmented stochastic optimization framework for large-scale energy
systems—multi-microgrids, electricity markets, hydrogen systems—where AI supports
decomposition, tuning, and initialization while stochastic optimization remains the core decision-making
backbone, improving computational efficiency and scalability.
Graph-Based Co-Design — variable-equation graphs and I/O path analysis to jointly determine process design and control configurations.Hierarchical Validation — clustering controllers and subnetworks, validated through plant-wide case studies.Uncertainty-Aware Decomposition — using variable-constraint coupling under uncertainty to divide large stochastic problems.AI-Guided Verification — AI-recommended decomposition and tuning, accepted only after feasibility checks, with fallback to baseline methods.
Associated members: Ji Hee Kim , Yoojung Yoon
