Battery Thermal Management Digital Twin Using CFD and MATLAB and AI

A student-laptop digital twin of an air-cooled 24-cell 21700 lithium-ion battery module (8S3P). A 2-D conjugate heat-transfer CFD solver written for the project ran 43 cases of the module cross-section (layout, cell gap, air velocity, mesh study) in 69 CPU-minutes. The CFD is reduced to 24 x 24 thermal influence matrices G(U) that drive a transient thermal network coupled with a 2RC equivalent-circuit cell model (Python, mirrored in MATLAB). An AI surrogate of the CFD (test error 2.2 % on the hottest-cell thermal resistance) finds the lowest-fan-power design, and an extended Kalman filter estimates all 24 cell temperatures, the state of charge, unexplained heat and air flow from only 6 thermistors - RMSE 0.32 K on the unmeasured cells in healthy unseen runs. The twin detects internal short circuits and fan blockage with a calibrated CUSUM detector (0/12 false alarms) and drives the fan with a model-predictive controller that matches full-fan peak temperatures with 39 % less fan energy at 2C fast charge. A web application with five database options shows CFD fields, design what-ifs, twin runs, alerts and AI models. The module, its sensors and the drive cycles are emulated/synthetic; the MATLAB files were written but not executed.

Category: Mechanical Engineering / Thermal Engineering / Computational Fluid Dynamics / Electric-Vehicle Batteries / Digital Twins / Machine Learning / MATLAB and Python

Mechanical Engineering Project M08 from ProTechSC Academy, with graphical abstract, video demonstration and project guidance.