PhD student (M/F) Frugal diagnostics and fault isolation in reconfigurable multi-source microgrids

Job Description

Thesis Subject

Thesis topic: Frugal diagnostics and fault isolation in reconfigurable multi-source microgrids
- Collaboration and co-supervision : LAAS-CNRS & AMPERE –
- French Program France2030 PEPR FutuRE

Context and challenges:
Electrical grids are evolving towards decentralized microgrids integrating intermittent renewable energy sources (solar, wind), storage systems (batteries, hydrogen), power electronic converters, and controllable loads. These complex cyber-physical architectures are expected to operate in diverse and critical environments such as hospitals, data centers, industrial sites, and isolated power grids.
In this context, microgrids must:
• Operate in grid-connected mode or in islanded mode;
• Guarantee power quality that meets stability, security, and service continuity requirements;
• Maintain their performance despite climatic variations, power disturbances, and component failures;
• Ensure a high level of resilience to incidents affecting equipment, sensors, or communication infrastructure.
However, traditional protection and monitoring methods are reaching their limits in the face of the increasing complexity of distributed energy systems. The simultaneous presence of multiple energy sources, frequent topology changes, interactions between physical and digital layers, and the emergence of weakly observable anomalies make monitoring modern microgrids more difficult.
Technological challenge: Traditional microgrid protection and monitoring methods are gradually reaching their limits due to:
• The detection of weak or precursor anomalies in highly converted architectures;
• The multiplicity of energy sources and components (converters, batteries, hydrogen systems, controllable loads);
• Dynamic reconfigurations related to changes in topology and operating mode (connected/islanded);
• The increasing interactions between physical and digital layers (sensors, communications, monitoring);
• The presence of concurrent defects and abnormal events, intentional or not, that can simultaneously impact physical components, sensors and communication systems.

These limitations complicate the early detection, isolation, and localization of faults in modern microgrids, potentially compromising their availability, reliability, and resilience.

Problem Statement and Scientific Challenges: How can we develop frugal, robust, and interpretable methods for anomaly detection and fault diagnosis in reconfigurable multi-source microgrids, while maintaining high performance in data-constrained and computationally resource-limited environments?
Challenges to Overcome: Several scientific challenges must be addressed to solve this problem.
The first concerns the early detection of weak or precursor anomalies and the extraction of relevant signatures indicative of an emerging fault, before the onset of significant performance degradation or system failure.
A second challenge involves diagnosing multiple faults in a context where several microgrid components can be affected simultaneously and produce signatures that are difficult to distinguish.
The robustness of the developed methods in the face of topology changes and transitions between different operating modes also constitutes a major challenge.
Furthermore, this thesis will focus on the frugality of the proposed approaches, seeking to reduce the requirements for training data, memory, and computing power while maintaining a high level of accuracy.

Particular attention will be paid to the explainability and embedding of the models to facilitate their understanding, validation, and integration into distributed or embedded architectures.
Finally, the developed approaches must be experimentally validated on real-time and HIL/PHIL simulation platforms representative of actual operating conditions.

Scientific Approach :
1. Modeling and Data Generation
• Development of representative models of multi-source microgrids integrating photovoltaic production, electrochemical storage, energy conversion, and controllable loads.
• Generation of normal, disturbed, and degraded operating scenarios taking into account different network configurations.
• Integration of physical faults, sensor drift, and anomalies affecting measurement and communication systems, whether intentional or not.
• Construction of databases dedicated to anomaly detection and fault diagnosis.
2. Anomaly Detection and Fault Diagnosis using Frugal AI
• Selection, feature extraction, and construction of relevant indicators from electrical signals and supervisory data.
• Study of signal processing techniques adapted for the early detection of anomalies (wavelets, symmetric components, time and frequency-domain indicators).
• Development of hybrid approaches combining physical models, signal processing, machine learning, explainable artificial intelligence (XAI), digital twins, and integrated data-model approaches.
• Study of the frugality of the developed approaches by reducing the requirements for training data, memory, and computing resources.

3. Advanced Diagnostics and Robustness
• Real-time detection of multiple and multi-event anomalies.
• Isolation, hierarchical classification, and monitoring of the evolution of faults affecting physical components, sensors, and measurement and communication systems.
• Analysis of intentional and unintentional events that may produce similar signatures.
• Robustness analysis in the face of changes in topology and operating mode.
• Evaluation of the potential contribution of digital twins to improving diagnostics, interpretation of results, and system resilience.

4. Experimental Validation
• Validation in real-time simulation.
• Validation on HIL/PHIL (Hardware-in-the-Loop and Power-Hardware-in-the-Loop) platforms.
• Performance evaluation on scenarios representative of resilient microgrids.
• Experimental demonstration on use cases integrating different energy sources and operating configurations.

Bibliography indicatives referencies
• Hatziargyriou N., Microgrids: Architectures and Control, Wiley, 2014.
• Lasseter R., “Microgrids”, IEEE PES Winter Meeting, 2002.
• Guerrero J.M. et al., “Hierarchical Control of Microgrids”, IEEE Transactions on Industrial Electronics, 2011.
• Karimi H. et al., “Wavelet Based Protection of Microgrids”, IEEE Transactions on Smart Grid, 2019.
• Heidari A. et al., “Deep Learning for Fault Detection in Smart Grids”, Applied Energy, 2021.
• Wen L. et al., “Deep Transfer Learning for Fault Diagnosis”, IEEE Transactions on Industrial Electronics, 2020.
• Zhang C. et al., “Graph Neural Networks for Power Systems”, Electric Power Systems Research, 2023.
• Zhou K. et al., “Artificial Intelligence in Smart Grids”, Renewable and Sustainable Energy Reviews, 2022.
• Farhangi H., “The Path of the Smart Grid”, IEEE Power & Energy Magazine, 2010.
• EH Sepúlveda-Oviedo, L Travé-Massuyès, A Subias, Marko Pavlov, Corinne Alonso « Fault diagnosis of photovoltaic systems using artificial intelligence: A bibliometric approach ». Heliyon, 2023

Desired Profile
• Education: Master's degree in Automation, Electrical Engineering, AI, or Energy.
• Skills: Modeling (Python/Matlab), AI (PyTorch/TensorFlow), signal processing.
• Qualities: A passion for experimentation (HIL/PHIL/Experimental Validation), teamwork (LAAS-CNRS & AMPERE), and project work.

Your Work Environment

This thesis is situated within the context of a PEPR project, in collaboration with eight CNRS and CEA laboratories working on the future of electrical grids.
More specifically, it is linked to WP5, dedicated to cybersecurity and network protection.
The doctoral candidate will be required to work on projects. Some travel and stays at the AMPERE laboratory will be necessary. The duration of these stays will depend on the progress of the work (between one and two weeks for each mission).
A knowledge of various data processing techniques is essential, as is the ability to create reliable databases.
Knowledge of the operation of an electrical grid, a static converter, and renewable energies will be highly beneficial.
Validation of findings will be carried out on technical platforms.

Constraints and risks

Since the applications are related to energy conversion, the doctoral candidate will be required to obtain low-voltage electrical certification during the course of their thesis.
Work will need to be validated progressively on the project platforms.
Knowledge of electronics, microprocessors, and communication technologies will be mandatory. Furthermore, several simulation and programming exercises using Matlab/Simulink and the Python language are planned.
Writing regular reports and presenting the work to the various project partners will be among the expected tasks.

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions