In this dissertation, the state of art in battery management is reviewed and its limitations are exposed. It is shown that the main reason of the inability for today's system to increase the precision of their estimation and the reliability of their prediction is deeply rooted in the approach adopted. Indeed, modern systems are all relying on behavioral models, either explicitly or implicitly which excel to predict the battery next state as long as the battery is not aged, and as long as the temperature does not vary too much, etc. In other words; today's system excel in predicting the battery behavior as long as the battery behaves as expected.
The reason of this situation comes from the lack of actual tool to diagnose the battery material state in the embedded system field. The field of chemistry engineering posses numerous measurement and diagnosis tools, but these methods are either destructive or non-embeddable. This work investigates a new comer in the field of battery diagnosis, thermodynamics, as the measurements involved are nondestructive and cover every possible evolution of the internal state of the battery. In a first time, a real-time embedded tracking algorithm is proposed, implemented and tested. Then, in a second time, different use-cases of thermodynamics in battery management are proposed. An analysis of the aging characteristics of the different phases of lithium is presented and its implications for battery and algorithms design are discussed. Finally, a interpretation of the entropy evolution is proposed in the case of determining which electrode is responsible for the entropy evolution during a situation of galvano-static cycling at high State Of Charge (SOC) under high temperature conditions.