Battery-module thermal identifiability benchmark: code, results and figures

This research resource focuses on the methodology and benchmark framework for parameter identifiability and optimal sensor placement in distributed battery thermal models. In battery thermal management systems, estimating internal states and parameters from limited surface measurements is a major challenge. The source paper presents a benchmark code repository containing a forward simulation engine, profile-likelihood and Fisher-information matrix (FIM) identifiability analyses, and a three-dimensional battery pack solver. To address the computational cost of high-fidelity 3D simulations, the methodology utilizes a neural surrogate model combining Proper Orthogonal Decomposition (POD) with a Multilayer Perceptron (MLP). This surrogate enables rapid, identifiability-gated sensor placement studies and multi-condition experiment designs. The benchmark code, structured across Python

modules and MATLAB files, reproduces the thermal behavior, sensitivity analyses, and sensor optimization routines. For M.Tech researchers, this framework provides valuable guidance on structuring reduced-order thermal models, evaluating parameter sensitivity, and implementing machine learning surrogates for real-time battery monitoring and state-of-charge/state-of-health estimation. Note that this resource serves as an implementation and structuring guide based on the published benchmark code; actual physical validation is limited to the datasets provided in the source repository.

Message: Hi FE, I need help with “Battery-module thermal identifiability benchmark: code, results and figures” in “Mechanical Engineering”