Accurate Federated Learning with Uncertainty Quantification for DER Forecasting Applied to Smart Grids Planning and Operation
The ALAMO project aims to address the challenges posed by the increasing integration of Distributed Energy Resources (DERs) into power grids while ensuring consumers’ privacy and accounting for uncertainties in forecasting models. The primary objectives include the development of accurate forecasting algorithms using Federated Learning, quantifying epistemic and aleatoric uncertainty in these models, and integrating them into operational planning tools for Distribution System Operators (DSOs).
Current planning tools lack privacy considerations and do not adequately account for uncertainties, rendering them obsolete in the context of growing renewable energy sources (RES) connected to distribution systems. This project also emphasizes the importance of coordination between Transmission System Operators (TSOs) and DSOs, especially in services like frequency regulation and congestion management. Additionally, it addresses the incorporation of uncertainties in Peer-to-Peer (P2P) energy markets, which are becoming increasingly prevalent.
The research plan comprises five critical tasks involving teams from MIT in the USA and ITI/LARSyS and INESC-ID in Portugal. These tasks encompass dataset preparation, Federated Learning algorithm development, uncertainty quantification, application of forecasting in grid operational planning, TSO-DSO coordination, and P2P trading. The project aims to demonstrate the effectiveness of the developed models in both virtual and physical testbeds, even with limited data availability due to privacy constraints.
01/12/2023 - 30/11/2024
Fundação para a Ciência e Tecnologia
Associação do Instituto Superior Técnico para a Investigação e o Desenvolvimento (PT); INESC-ID - Instituto de Engenharia de Sistemas e Computadores, Investigação e Desenvolvimento em Lisboa (PT).
Critical Computing; Privacy and Security; Specific Applications Areas; Sustainability, and Social Justice