Understanding the behavior of metal anodes in batteries and accurately predicting their performance is a challenge due to the methodological gap between theoretical models and experimental observations. In particular, we focus on two key phenomena: unstable material growth [1] and Galvanostatic Discharge-Charge [2,3] that are crucial for the advancement of next-generation battery technologies. In this talk, based on [4,5], we propose to apply DeepLearning as a new approach for parameter estimation in Partial Differential Equation (PDE) models that describe these phenomena. In particular, we tackle two problems in battery modeling: 1) parameter identification for a Reaction-Diffusion PDE model [4] of electrochemical pattern formation, where a Convolutional Neural Network (CNN) trained on Turing patterns that are the numerical solutions of the PDE is used, 2) parameter fitting for a Metal battery Anode Cycling PDE model [5] using a CNN-Long Short-Term Memory (CNN-LSTM) trained on simulated charge-discharge voltage profiles. In both cases, our AI algorithms were successfully applied to estimate parameters also for true experimental data: degradation images of cathodes and discharge-charge time series. These results highlight the robustness of our approach, which allows to bridge the gap between theory and experiments.