Author(s): Georgios Bouloukakis, Ingrid Colleau – IMT Télécom SudParis

The Di-Hydro project involves digitising hydropower plants, to operate them as sustainably as possible and better plan maintenance. Di-Hydro partner IMT, in particular Télécom SudParis, is contributing to the development of a federated platform designed to facilitate secure data-exchange and decision-making, both at the level of each individual plant and collectively.

Initially, the Télécom SudParis team was engaged to address the platform’s cybersecurity needs, but they quickly recognised a significant requirement for interoperability— the ability to exchange and share data effectively. This necessity became apparent when considering the vast array of data needed to train AI models and the diverse range of stakeholders involved in the Di-Hydro project, both in their activities and geographic locations.

Georgios Bouloukakis, involved from the start due to his extensive experience in building IoT systems, shared the data exchange requirements to the project coordinator. Consequently, he was appointed to oversee this aspect of the project. His responsibilities intersect with those of his colleague Joaquim Garcia Alfaro, a cybersecurity researcher and co-PI of Di-Hydro. Together, they are developing a federated platform within a secure IoT context to facilitate decision-making and experience-sharing among various stakeholders, whether within a single country or across different countries, while ensuring the security and confidentiality of each entity’s data and models.

Di-Hydro Platform Digital Twins and Data Standards

The first step in developing the Di-Hydro platform involved integrating “Digital Twins” of hydropower plants. Each physical site is digitally represented based on static data (e.g., rooms, hallways) and dynamic data (e.g., a person entering a room), supplied by IoT devices. The IMT team noted the challenge that arises because each sensor provides data in its own structure, and each provider supplies data in different formats. Therefore, establishing a standard, common format that all partners must follow is crucial to enable collaboration.

This standard will apply to all data, whether from existing sensors or new ones to be integrated into the system. Partners developing new sensors must ensure that the data collected adheres to the project-defined format. In addition, IMT mentioned that some partners might encounter challenges to comply with the chosen format, in which case, they will receive assistance in creating connectors to integrate non-standard data.

Machine Learning to Consolidate Models

The platform aims to offer new solutions through a distributed, non-centralised, decision-making algorithm. Initially, multiple AI models will be trained independently using local data. To consolidate these advancements, IMT scientists will apply their expertise in selective sharing mechanisms and federated software architectures.

This AI approach might involve creating a global model that consolidates several models trained on local data from various plants. These federated models can then be simulated for decision-making at multiple levels.

Standardising and sharing this data could also help fill in missing information. In fact, IMT anticipates variability in data across different plants, with some providing certain data while others cannot. The research and tools being developed aim to address these gaps. The ultimate goal is to continually enhance the understanding of these systems to ensure well-functioning, optimal, and sustainable hydropower plants.