Edge-Computing FogFlow Framework For Solar Generation Prediction Exploiting Federated Learning
Thodoris Samaras, Asimina Dimara, Petros Tzallas, Alexios Papaioannou, Napoleon Bezas, Stelios Krinidis, Christos‐Nikolaos Anagnostopoulos, Dimosthenis Ioannidis, Dimitrios K. Tzovaras · 2023
The integration of renewable energy sources into residential power systems, accompanied by the growth of edge computing, presents a dynamic landscape for energy management. In this paper, an innovative solution for solar generation forecasting is introduced, effectively addressing issues of data privacy, operational efficiency, scalability, and network optimization. A Federated Learning Gradient Boosting (FedXGB) framework integrated with edge computing is applied, significantly amplifying energy efficiency and sustainability in residential environments. Powered by FogFlow, the orchestration and optimization of IoT ecosystems, particularly within smart homes, are advanced, and data privacy is ensured. Remarkable results are observed, especially in mid-term and long-term solar forecasting, highlighting the framework's transformative potential in reshaping energy management. This innovation holds the promise of revolutionizing how renewable energy is harnessed within homes and communities, paving the way for a more sustainable and efficient future. With this comprehensive approach, the intersection of renewable energy and edge computing takes a significant step forward in enhancing energy management.