Association of control theory and data analysis techniques to implement planned perennity in software defined power converters – DataPower

Power electronics is a key technology to the energy transition. However, power converter design techniques are function-defined, resulting in a long and costly development process yielding non-reusable solutions. A software-defined approach, focused on repairing, reprogramming and refurbishing is possible. This approach is called planned perennity. The central problem to perennial assets is aging. In the case of power electronics, it is well known that different control algorithms will impose different thermal stresses in active and passive components. It is thus essential to find a mean of keeping track of a perennial product usage history and integrate this history in its control. The DataPower project proposes the creation of an online data platform that will gather the converter history, create its digital twin by combining model-driven control algorithms with data-driven machine learning techniques and study the best approaches to compensate this aging history for software defined power converters. This open online platform will provide the means to study several use cases using the same standardized open-source power electronics hardware. Its main contribution will be on the field of machine learning for power electronics, specifically on their preventive maintenance. The datasets acquired during the project will be used to create an open database of power electronics and systems applications. Collected datasets will constitute a database kernel that is expected to grow even beyond the project completion, creating a diverse community with both industrial and researchers members.