Best paper award at the 3rd DSBFI Conference 2025

The research paper of the IAD research group won the award: Best paper award at the 3rd DSBFI Conference 2025

Title: Explainable Lightweight Federated Learning for Load Forecasting with Smart Meter data

         Le Vu Hoang Duc, Ngo Van Uc, Kim Duc Tran, Ludovic Koehl, Kim Phuc Tran

Abstract:

The purpose of forecasting electricity demand is to achieve a balance between supply and demand between providers and customers, which is a crucial factor in the energy field. This study aims to employ an approach using XAI and FL in a DNN model, combined with datadriven techniques to deliver accurate predictions based on data from current smart meter systems in smart buildings. The paper continues the work on federated learning to ensure the privacy of individual meter data, allowing several parties, commonly referred to as clients, to keep their data distributed and not centralized in the training process. Additionally, the paper highlights the benefits of the XAI approach in providing clear and reasonable explanations for the model’s predictions. This not only supports energy providers in analysis, decisionmaking, and optimizing operations but also enhances the trust and understanding of stakeholders regarding the forecasting model, thereby optimizing the efficiency of energy distribution and usage.