Original Article Writen by Emmanuelle Delsol, published 11th of April 2024 for the LeMondeInformatique.fr.
The following is an English translation of the original article, which discusses the DI-Hydro project in which SATRAI is participating for IMT.
The equipment with sensors, digital twins, and AI could optimize the operation of hydroelectric power plants and limit their environmental impact. A research project involving Télécom SudParis is working on this, as detailed by I’MTech, a publication of the Institut Mines-Télécom.
The production of renewable energy is paradoxically not free from environmental impact. Hydroelectric power plants are no exception; their operation can block fish migration or increase water temperature, among other harmful effects. To mitigate these negative impacts, the European Union is financing the Di-Hydro project (Digital maintenance for sustainable and flexible operation of hydropower plants) with 4 million euros for the digitization of these facilities, as mentioned by the I’MTech publication of the Institut Mines-Télécoms.
Started in October 2023 for a duration of 3 years, Di-Hydro relies on an IoT architecture for the plants. The two main objectives are to optimize the operation of these installations by improving their maintenance and to monitor the evolution of water quality and ecosystem biodiversity. Local energy providers and research teams such as those from Télécom SudParis will supply or develop suitable sensors, with installation and commissioning expected to take 6 months.
Consolidation of Local Data Models
Di-Hydro researchers are also developing a shared data analysis and exploitation tool. The agenda includes AI and digital twins. Initially requested to ensure the system’s cybersecurity, the research teams from Télécom SudParis (Georgios Bouloukabis, associate professor, and Joaquim Garcia Alfaro, professor) have also taken on the integration of the IoT platform into the plant ecosystem. Digital twins of these plants will be designed from static descriptive site data and dynamic data generated by sensors. The Di-Hydro teams had to work on defining a common standard for this data from different sources and formats, often incompatible with each other.
To exploit these digital twins, the research teams are developing a distributed decision-making algorithm based on several AI models trained initially on local data. They plan to work on selective sharing and federated software architectures. This AI model would then rely on consolidating several models trained on local data for multi-level decision-making and completing missing data.