PhD Candidate — Electrical Engineering
I build digital twins that connect batteries, the grid, and AI data centers.
Research Assistant at the Smart Grid Resiliency and Analytics Lab, West Virginia University, linking cell-level electrochemistry to grid-scale control — so utility-scale battery storage and large AI data centers can be integrated into the power system safely, in real time.
01 — Research Profile
My research develops reduced-order, real-time battery models that connect cell-level electrochemical behavior — degradation, thermal response, internal stress — with grid-level control and operational decision-making. The goal is safer integration of large, fast-changing loads, such as AI data centers, into power systems that were never designed to carry them.
Current work focuses on hybrid BESS modeling frameworks that pair real-time equivalent-circuit emulation with degradation-aware electrochemical state estimation at utility scale, physics-informed dynamic load modeling of AI data centers, and real-time grid-impact analysis for mission-critical infrastructure.
Conventional static models fall short under modern grid conditions. AI data centers and other highly dynamic loads introduce fast disturbances that demand physics-informed modeling, real-time validation, and control-aware mitigation — which is what I spend most of my time building toward.
Prof. Anurag Srivastava
Smart Grid Resiliency and Analytics Lab
West Virginia University
Grid-aware modeling and resilient integration of large-scale AI data centers using battery energy storage.
02 — Experience
From EPC consultancy and 120 MW of industrial power distribution in Bangladesh to real-time digital twins of utility-scale storage in the U.S.
03 — Selected Technical Projects
A hybrid architecture pairing fast equivalent-circuit models for HIL-compatible real-time execution with reduced-order thermo-electrochemical models that expose internal states — concentration gradients, SOC deviation, aging — invisible to ECM-only simulations.
Dynamic load models representing fast-changing computational power demand, used to quantify grid stress from load ramps and evaluate how BESS and microgrid support can mitigate the resulting disturbance.
Representative-cell scaling that translates cell-level electrochemical behavior — SOC, SOH, SEI-related loss, lithium plating — into pack-level BESS degradation under real grid duty cycles.
Real-time simulation workflows connecting transmission- and distribution-level grid models, battery models, and converter response for timing-constrained controller validation.
A clustering-based approach to classifying dynamic load behavior in AI data-center operation, built to flag early warning indicators of power-system stress before they produce unacceptable voltage or stability conditions.
04 — Publications & Presentations
Presenting “Powering the AI Boom Without Breaking the Grid” at the WVU Statler College Research Week Poster Symposium, 2026 — 1st place.
05 — Skills & Tools
06 — Education
07 — Journal
Short write-ups from the lab and the bench — new entries land here as the research moves.
ECM-only models track bulk state of charge, but a lot of the physically interesting stress — stoichiometric deviation, SEI-related loss — only shows up in surface-level electrochemical states. Notes on why that gap matters once a pack is doing grid duty.
A summer spent validating cell models against measured behavior instead of simulated ground truth. On what changes when a model has to survive contact with a real cell on a real cycler.
Data centers are usually treated as an IT problem. Once the load ramps get fast enough, they start behaving like a grid disturbance instead — and that changes what a useful model needs to capture.
08 — Contact