Mahdi Abid

My research focuses on the development of advanced time series algorithms based on LSTM, Transformer, and diffusion model architectures for predicting key health indicators of Lithium-ion batteries, such as SOC, SOH, and RUL (State of Charge, State of Health, Remaining Useful Life). The approach combines the processing of electrochemical cycling data, feature extraction related to degradation mechanisms, and the optimization of attention-based architectures to improve long-term prediction accuracy. Particular attention is given to the robustness of the models on experimental data, whether real or synthetic, collected from both laboratory cycling tests and real-world usage in electric vehicles.
This work contributes to the application of research in artificial intelligence for predictive battery maintenance. The AI algorithms developed and pre-trained are designed to be integrated into a full-stack software structure dedicated to battery data real-time forecasting. I focus on the design, training, and evaluation of predictive models, in line with the industrial challenges addressed by PreDeeption, a deeptech startup emerging from the CNRS.