Privacy-Preserving Energy Efficiency Optimization Using Federated Learning-Based IoT Networks

Baran E.

4. International Congress Of Modern Scientific Research July 01-03, 2026, Ankara/Türkiye, Ankara, Türkiye, 1 - 03 Temmuz 2026, ss.1-5, (Tam Metin Bildiri)

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Ankara
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.1-5
  • Gazi Üniversitesi Adresli: Evet

Özet

The increasing deployment of Internet of Things (IoT) devices in smart buildings has created new opportunities for improving energy efficiency through real-time monitoring and intelligent control. However, centralized energy management solutions often require continuous transmission of sensitive user data, leading to privacy concerns and increased communication overhead. This study proposes an Adaptive Federated Reinforcement Learning for Energy Efficiency (AFRLEE) framework that combines IoT sensing, Federated Learning (FL), and Reinforcement Learning (RL) to optimize building energy consumption while preserving user privacy. In the proposed architecture, IoT devices collect environmental and operational data, and local models are trained independently without sharing raw information. Model parameters are aggregated through a federated server to construct a global prediction model. The outputs of the federated model are then utilized by an RL-based optimization agent to dynamically manage HVAC systems, lighting units, and flexible electrical loads. A smart campus scenario involving multiple university buildings is considered for performance evaluation. The proposed framework is assessed in terms of energy savings, prediction accuracy, communication cost, and privacy preservation. Results indicate that the AFRLEE framework achieves significant reductions in energy consumption and network traffic while maintaining high forecasting performance. The findings demonstrate that integrating federated learning and adaptive control mechanisms can provide a secure, scalable, and efficient solution for next-generation smart energy management systems.