Mots-clés | Digitalisation de l'énergie, data science, edge analytics, systèmes électriques intelligents, optimisation, intelligence distribuée Energy digitalization, data science, edge analytics, smart grids, distributed intelligence, optimization |
Résumé | ... Context and challenges:
In the vertically integrated electrical energy systems of the past, power system management was carried out centrally by the transmission and distribution system operators (TSOs, DSOs). In the frame of the energy transition, emerging new actors (aggregators, microgrid operators, energy community managers, self-consumption etc.) and the proliferation of assets connected to the grid, such as renewable energy (RES) plants, storage devices, electric vehicles (EVs), smart-homes/prosumers with IoT devices, electric heating/cooling systems, etc., urge for a paradigm shift towards decentralization. New business models are likely to be based on physical or virtual groupings of assets (“cellsâ€), instantiated as virtual power plants (VPPs), energy communities, microgrids, and others. In the power systems of the future, all these “cell†variants will probably coexist, and their operation should be optimized accounting for 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 in electricity markets. Optimization functions (“distributed intelligenceâ€) are necessary at these lower levels down to the grid edge (cell, feeders, assets/devices…) and also need to be aligned with the grid operation.
Main objective of the thesis:
The overarching objective of this research project is to develop distributed optimization methods for grids with a very large number (tens/hundreds of thousands to millions) of connected devices. The aim is to account for the involved uncertainties, the classification of assets/devices into different typologies of virtual/physical cells, their computational/communication capabilities/limitations, environmental disturbances, QoS/grid constraints, and privacy concerns. Large scale distributed optimization requires the design of appropriate grid-aware signals that affect the local optimization processes for a multitude of devices, so that their aggregation provides a predictable response (even though it is the result of the response of different assets), while ensuring an optimal use of grid infrastructure.
Methodology and expected results:
The first step of the research project is a bibliographic research and familiarization with the methods and tools developed at our Group. The initial use case of focus will be the predictive management (scheduling) of the assets (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, with inherent uncertainties in their production/consumption profiles, to cover use cases such as distribution grids and/or VPPs that integrate EVs, RES plants, storage devices, prosumers and the like. The research project will integrate predictive models that reduce the complexity associated to multiple uncertainties, employ machine learning and/or statistical methods for high dimensional problems (e.g., sparse models, functional data analysis, edge ML), and explore distributed optimization strategies to cope with the very large problem sizes. Optimization strategies will involve decomposition methods, blended with machine learning developments (e.g., optimal decision trees able to adapt dynamically to the vast amount of incoming information), and produce signals that account for the local grid conditions (i.e., congestions/overloads) to which the different cells of assets respond based on their capabilities and individual objectives. |