Energy-Aware Hybrid Machine Learning on Edge Telemetry for Dynamic Resource Allocation and Carbon-Neutral Smart Factories

• Ada John

  ladoke Akintola university of technology

  Author

The industrial sector's transition toward carbon neutrality requires intelligent systems capable of balancing energy efficiency with production performance in real time. This study addresses the gap in edge-deployable hybrid architectures that simultaneously optimize resource allocation and carbon footprint in smart manufacturing environments. We propose an Energy-Aware Hybrid Machine Learning (EA-HML) framework that integrates Temporal Convolutional Networks for energy demand forecasting, Random Forest classifiers for workload prioritization, and a Dueling Deep Q-Network controller for dynamic resource allocation on edge devices. The framework was validated on real-time telemetry data from discrete manufacturing operations, achieving 89.4% resource allocation accuracy while reducing energy consumption by 23.7% and operational costs by 19.2% compared to baseline methods. Statistical significance was confirmed (p < 0.01), with the hybrid architecture demonstrating superior convergence stability and edge deployment feasibility. These findings establish a replicable pathway for deploying energy-aware intelligence at the industrial edge, enabling factories to pursue carbon-neutral operations without sacrificing production throughput.

• pdf

Copyright (c) 2026 Ada John (Author)

This work is licensed under a Creative Commons Attribution 4.0 International License.