Adaptive Federated Learning for Industrial Wireless Energy Optimization

Abstract

The advancement in Industrial Wireless Sensor Networks (IWSNs) demands intelligent energy management strategies to enhance network longevity and reliability. This research proposes an adaptive federated learning (FL) framework to optimize energy consumption in IWSNs, incorporating predictive maintenance and dynamic clustering. Utilizing lightweight machine learning models integrated with FL, the framework predicts network conditions and proactively adjusts sensor operations. Dynamic clustering facilitates efficient data aggregation and reduces transmission energy. The proposed solution significantly minimizes unnecessary energy usage by predicting sensor node failures and scheduling optimal sleep states, thus extending network lifespan and reliability. Experimental evaluation demonstrates that this approach reduces energy consumption by up to 25% compared to traditional energy management schemes, validating its effectiveness in industrial environments.