Fuel cell technology has been attracting increasing attention during the last two decades. It has been demonstrated that fuel cells can be penetrated in a wide scale of applications thanks to the attractive advantages, such as high-power density, zero on-board emission, less well-to-wheel consumption. Nevertheless, most of the fuel cells in service still suffer from unsatisfactory durability performance, especially for transportation applications. It has been found that the main scientific barrier lies in theĀ system control level. Proactive control dedicated to the enhancement of fuel cell durability is still a fresh topic. Inspired by the results we have obtained in the recent research on prognostics and health management and the recent remarkable development in machine learning, the project DEAL will be dedicated to a multi-level learning-based control strategy for fuel cell systems. In this control strategy, three-level learning will be treated using emerging machine learning methods in a model predictive control framework. The durability is expected to be improved significantly thanks to the control strategy.