Project

The main objective of the Accurate federated Learning with uncertainty quantification for DER forecasting Applied to sMart Grids planning and Operation (ALAMO1) project is to develop technologies to facilitate the management of power grids with very high penetration of Distributed Energy Resources (DERs), while at the same time assuring the privacy of the main stakeholders. In particular, the project will address outstanding research challenges related to developing accurate forecasting algorithms based on Federated Learning (FL) and challenges associated with producing sharp and well-calibrated quantifications of epistemic and aleatoric uncertainty for such forecasting models.

The developed forecasting and uncertainty models will be incorporated into operational planning tools for distribution system operators (DSOs). Such tools are becoming crucial for DSOs due to increased RES connected to the distribution systems. This project will also consider the participation of the Transmission System Operators (TSOs), that are ofen ignored in the planning and operation of power grids.

Tasks

The proposed research plan consists of five tasks, that will be led by teams at MIT in the USA, and at ITI/LARSyS and INESC-ID in Portugal. T1: Dataset Preparation and Grid Modeling (MIT), T2: Development of Federated Learning Forecasting Algorithms (MIT), T3: Uncertainty Quantification and Scoring (ITI/LARSyS), T4: Application of FL Forecasts for Grid Operational Planning, TSO-DSO Coordination, and P2P Trading (INESC-ID), and T5: Demonstration in Real-World Testbed (ITI/LARSyS).