Reducing the impact of uncertainties in the optimization of low-carbon energy system at district level

Energy system optimization models (ESOM) are powerful tools for improving decision making in the transition towards carbon-free energy systems.

The results provided by ESOMs are greatly influenced by data uncertainty since they are considered on a future time horizon. For instance, the possible evolution of energy prices, energy production and demand or the efficiency of technologies must be taken into account. Although many works have started in recent years to study the impact of these uncertainties on the solutions, it has been pointed out that modeling simplifications may induce significant bias in the obtained results.

The work proposed in this new PhD topic aims at studying the response of ESOM along energy system design and transformation steps, and reducing or assessing the impact of uncertainties as early as possible in the process. It will especially aims at limiting the bias related to model simplification, by systematically propagating relevant information from more detailed models towards simplified models used for sensitivity analysis and optimization under uncertainty. To this aim, the currently envisioned path is to leverage techniques such as machine learning, and in particular the constraint learning approach, to extract relevant information from simulation and inject back into the simplified optimization models.

As a result, the work is expected to improved the methods currently in use for designing and improving energy systems at local level, in order to favor energy savings, and limit CO2 emissions as well as other environmental impacts.

Hybrid Generic EMC Filter

In the field of embedded applications, power converter specifications are crucial. They must not only be efficient and compact, but also meet strict electromagnetic compatibility (EMC) standards. Understand that these converters may be susceptible to their own interference (autoimmunity), cause or experience disturbances in their environment, primarily from common mode currents.
Even low-power power converters can generate high-frequency electromagnetic emissions, which can interfere with other nearby equipment or even disrupt radio signals.
Traditionally, to meet EMC requirements, we rely on shielding and passive filtering techniques, which add significant weight, volume and cost to the system. Around 20% of these costs and constraints are attributed to passive EMC filters.
We note the arrival of new converters (based on large gap SiC/GaN components) whose switching frequencies approach, or even encroach on, the frequency ranges of EMC standards. In order to counter this problem, a new alternative is emerging: active EMC filters. The latter offer at least similar performance while considerably reducing bulk and weight.
As part of this thesis, we will explore these active CEM filters through different stages. We will start with a state of the art, followed by the estimation of common mode and differential mode noise of switching components. Then, we will simulate and compare the most relevant solutions, whether active or passive. We will also get hands-on by performing electromagnetic compatibility tests on common filters and converters.
Finally, we will design and test a prototype active filter for a specific converter. To successfully complete this thesis, it is necessary to master both analog and digital electronics, as well as electronic simulation software (LTspice, Pspice or PSIM) and printed circuit design tools (Altium). Additionally, knowledge of embedded programming would be a valuable asset.

Grid-Interface Power Converter with MVAC and MVDC

In this thesis subject, we propose Grid-Interface Power Converter with MVAC and MVDC. GI-PC control strategies to provide system services and facilitate network management and protection will be studied (eg support for the voltage plan, study of resonances, MVRT, etc...). A digital prototype of GI-PC at the MV level will be proposed implementing the control algorithms. The validation of the prototype will include regulation of the MVDC bus according to different scenarios. The GI-PC can contribute for:
• Providing a grid-connected interface for various MVAC systems such as power router
• Providing distribution interface for different levels of DC systems
• Improving power quality of MVAC distribution systems
• Providing a grid-connected interface for high-power DC sources such as electric vehicle charging stations, battery energy storage systems, H2, and PV and wind farms
• Other functionalities: fault support (firewall), imbalance reducing, auto reconfiguration (redundancy), grounding adapting, galvanic isolation …

Topology reconstruction of a ramified network by multisensor reflectometry

Smart Grids aim at monitoring and controlling electric power networks. Many parameters have to be monitored such as production and consumption units, and the integrity of the structure of the interconnection netwotk itself.

Smart grids aim at enhancing the quality of service while protecting people and infrastructures. In this area of research, most algorithms are deployed for taking the human out of the retroaction loops in order to maximize the availability and the reactivity. For that reason, artificial intelligence based algorithms are increasingly incorporated in decision loops.

In that industrial context, we are interested in methods that aim at estimating electrical network topologies. The topology of a network includes the length of the cable lines and their electrical properties, so as the characterictics of the loads that are connected to the networks (production and consumption units), and also potential faults in the network. In the end, the accurate estimation of the topology may be used to monitor the network with more accuracy with the help of a more accurate a priori information.

In order to characterize the topology, we propose to deploy either a single or a distributed set of electric reflectometers. These devices inject signals in the network under test and the study of the reflections gives information back which can be used to reveal the structure of the network. More precisely, every impedance discontinuity along the line wil cause partial reflections of the waves.

Previous works were conducted by our team of researchers on that topology reconstruction topic, by exploiting optimization algorithms coupled to a simulator. We would like to extend these works in two directions. First, we would like to explore a machine learning regressor-based approach in a mono sensor version. Second, we would like to estimate the topology by combining the measurements from multiple sensors, either with already available optimization-based approachs, or by the new machine learning-based approach.

Nanocrystalline Soft Magnetic Composites: Powder morphology and design for controlling their magnetic properties for high frequency applications

Context: Achieving carbon neutrality by 2050 will require massive electrification of the power production systems. Power electronics (PE) is a key-enabler that will this transformation possible (renewables, integration of energy micro-grids, development of electric mobility)
Problem: Current developments in PE converters aim at increasing the switching frequencies of large bandgap switches (SiC or GaN). At low frequencies, magnetic components remain bulky, occupying up to 40% of the total footprint. At high frequencies (HF > 100 kHz), very significant gains are expected, but only if the losses generated by these components remain under control. Today, the main class of magnetic materials applied to HF is MnZn or NiZn ferrites, due to their low cost and convenient electrical resistivity (?elec > 1 O.m). The main drawbacks of these materials are their low saturation induction (Bsat < 0.4 T), which limits their size reduction, and their mechanical fragility. Nanocrystallines materials have better Bsat (1.3 T), but their ?elec is about 1.5 µO.m (6 times less resistive than ferrites), which generates significant induced current losses at HF.
Thesis objective: To develop magnetic composites by grinding nanocrystalline ribbons, electrically insulating the powders (coating fabricated by sol-gel), compacting of the powder at high pressure (1000-2000 MPa) for the core shaping and finally by applying an annealing treatment to relax the thermal constraints.

Environmental monitoring of pollutant dispersion: optimal placement of mobile sensors

Do you feel concerned by environmental pollution issues? This research will enable optimal deployment of mobile sensors for monitoring air quality in urban environments. Complex urban geometries [1] and dynamic pollutant dispersion scenarios are the scientific challenges to be met in order to better estimate local air pollution, identify sources and anticipate exposure peaks.
Our previous research has focused on the identification of pollution sources, neglecting the critical aspect of sensor placement [2]. For partial differential equation models, promising approaches based on the structural property of observability of dynamical systems have been proposed [3]. A generic two-stage approach will be studied in the thesis: the development of an infinite-dimensional variational approach for an advection-diffusion model, followed by the finite-dimensional implementation. The results of this thesis will include new sensor placement strategies, quantitative evaluation results in simulation under realistic conditions on a city district in Grenoble and/or Paris, and an in-depth understanding of how Physics-Informed Machine Learning (PIML) [4] can improve air quality monitoring in urban areas, both in 2D and 3D.
CEA-Grenoble ( employs over 2,500 researchers and technicians on a 64-hectare campus in the foothills. The activities of our lab focus on sensor signal fusion through studies in signal and information processing, artificial intelligence, and embedded algorithms, and brings together some twenty experienced research engineers and students from Master 2 to post-doctorate. To join our team, we are looking for a candidate with an applied mathematics profile, a taste for physical models and numerical methods, and good writing skills. You will be co-supervised by Prof. Didier Georges of GIPSA-Lab at Grenoble- Alps University (UGA)( You will also have access to scientific databases, a computing cluster with GPUs and will be trained in the use of a state-of-the-art atmospheric dispersion simulator: Parallel Micro-Swift-Spray co-developed at CEA. Remuneration will be around €2400 (gross) per month during the three years of the thesis. Join us in a unique research environment dedicated to ambitious projects that address today's major societal challenges.
[1] M. Mendil, S. Leirens, P. Armand, C. Duchenne, “Hazardous atmospheric dispersion in urban areas: A Deep Learning approach for emergency pollution forecast”, Environmental Modelling & Software, Volume 152, 2022
[2] R. Lopez-Ferber, D. Georges, S. Leirens, “Fast Estimation of Pollution Sources in Urban Areas Using a 3D LS-RBF-FD Approach”, submitted to the European Control Conference 2024
[3] D. Georges, “Optimal Location of Mobile Sensors for Environmental Monitoring”, European Control Conference (ECC), July 17-19, 2013, Zürich, Switzerland
[4] M. Raissi, P. Perdikaris, and G. E. Karniadakis. Physics-informed neural networks: A deep learning
framework for solving forward and inverse problems involving nonlinear partial differential equations. Journal of Computational physics, 378 :686–707, 2019.