| Mots-clés | Digitalisation de l'énergie, data science, intelligence distribuée, optimisation, edge analytics, AI on the edge Energy digitalisation, data science, distributed intelligence, optimization, edge analytics, AI on the edge |
| Résumé | Context and challenges:
In the vertically integrated energy systems of the past, power system management was carried out centrally by the transmission and distribution grid operators (TSOs, DSOs). In the frame of energy transition, the paradigm changes towards a decentralized one. This is due to the high number of emerging new actors with different interests (aggregators, microgrid operators, energy community managers, self-consumption etc.), and due to the very high population of assets connected to the grid like renewable energy (RES) plants, storage devices, electric vehicles (EVs), smart-homes/prosumers with IoT devices, electric heating/cooling systems etc. As a function of the involved actors and the related business models, physical or virtual groupings of assets (“cells”) may emerge, i.e. virtual power plants (VPP), energy communities, microgrids, a.o. In the power systems of the future most likely all these “cell” variants will coexist. The operation of each such cell is optimized based on the specific interests of the involved actors. For example, a VPP operator aggregates hundreds to thousands of assets to achieve a critical mass of flexibility and valorize it on electricity markets. Optimization functions (“distributed intelligence”) are necessary at these lower levels down to the grid edge (cell, feeders, assets/devices…). These functions cannot though be performed independently from the grid. The grid operators may have to send signals for grid-aware control downwards to influence the local optimization processes.
Main objective of the thesis:
The overarching objective of this research project is to develop distributed optimization methods for systems with a very large number (tens/hundreds of of thousands) of connected devices. The aim is to account for involved uncertainties, the organization of the assets/devices into different typologies of virtual/physical cells, the fact that individual assets may have limited computation and communication capabilities, and are subject to environmental disturbances, QoS constraints, and privacy issues. Distributed optimization at such a very large scale requires a design of appropriate signals for thousands of devices to be grid-aware so that the aggregates provide a predictable response even though it is the result of the response of different assets.
Methodology and expected results:
The first step of the research project is a bibliographic research and familiarization with tools developed at our group. The initial use case of focus will be the predictive management of the assets (the latter for time frames in the order of a few minutes to a few days ahead). The developed approaches should be scalable to a very high number of assets to cover use cases like distribution grids or/and Virtual Power Plants (VPPs) that integrate EVs, RES plants, storage devices, prosumers etc. Each type of asset may have a high inherent uncertainty in its production or consumption profile. Approaches will integrate predictive models that reduce the complexity associated to multiple uncertainties, implementing statistical or machine-learning methods for high-dimension problems (e.g. sparse models, functional data analysis, edge ML). Given that very large amounts of input data are considered, strategies to reduce the computational cost via distributed optimization will be examined. These strategies will investigate decomposition approaches and optimal decision trees able to adapt dynamically to the incoming information. The high-level signals will be global or locational to account for the local grid conditions (i.e. prediction of congestions/overloads). It will be evaluated how these signals are used in the different cells of assets as a function of the cell optimization objectives. ... |
| Profil candidat | Please send the following elements by email:
• Curriculum vitae (CV).
• Motivational letter for the application (cover letter).
• Contact details of two individuals that can provide a letter of reference (and eventually available already letters of reference).
• Copy of grade transcripts and last diploma (in English or French).
To Prof. George Kariniotakis (georges.kariniotakis@minesparis.psl.eu) AND to Dr. Simon Camal (simon.camal@minesparis.psl.eu)
Use in the title of email the acronym of this PhD topic “PHD-2023-DistrOptim”
Deadline for applications: 15/03/2023.
Do not hesitate to email to the above addresses for an early expression of interest and for further information on the position. Profile: Engineer and / or Master of Science degree (candidates may apply prior to obtaining their master's degree. The PhD will start though after the degree is succesfully obtained).
Good level of general and scientific culture. Good analytical, synthesis, innovation and communication skills. Qualities of adaptability and creativity. Motivation for research activity. Coherent professional project. Skills in programming (eg R, Python, Julia,…). A succesful candidate will have a solid background in three or more of the following competencies:
• applied mathematics, statistics and probabilities
• data science, machine learning, artificial intelligence
• optimization
• energy forecasting
• power system management, integration of renewables
Expected level in french : Good level
Expected level in english : Proficiency |
