Code and source data for "UC-PPO: Universal Cell-wise Proximal Policy Optimization for Passive Cell Balancing in Battery Packs"

This dataset contains Python code, trained reinforcement-learning models, evaluation outputs, and source data supporting the manuscript "UC-PPO for Passive Cell Balancing in Battery Packs with Variable Series Configurations." The archive covers comparisons with no-balancing, rule-based BMS, and pack-specific deep-reinforcement-learning baselines; direct deployment of the same 8S-trained UC-PPO policy to different series configurations; computational-cost evaluation; action-related reward ablation; input-feature masking and mutual-information analysis; integrated-gradients and permutation-sensitivity analysis; voltage-rank retraining ablation; and standardized large-pack baseline comparisons.

A figure-to-file map, checksums, environment information, and reproduction notes are included.