Journal of Engineering, Project, and Production Management, 2026, 16(5), 2026-0056
A Machine Learning-Based Temperature Monitoring Method for Power Equipment by Fusing Morphological Features and Chromaticity Analysis
1
Engineer, Electric Power Research Institute of Guangdong Power Grid Co.,
Ltd., China, E-mail: chenyuze@dky.gd.csg.cn (corresponding author).
2
Senior Engineer, Electric Power Research Institute, Guangdong Power Grid
Co., Ltd., China, E-mail: xieshanyi@dky.gd.csg.cn
3
Engineer, Electric Power Research Institute, Guangdong Power Grid Co.,
Ltd., China, E-mail: zhangkai@dky.gd.csg.cn
4
Senior Engineer, Electric Power Research Institute, Guangdong Power Grid
Co., Ltd., China, E-mail: fanying@dky.gd.csg.cn
5
Senior Engineer, Electric Power Research Institute, Guangdong Power Grid
Co., Ltd., China, E-mail: zhougang@dky.gd.csg.cn
Project Management
Received May 12, 2026; revised June 7, 2026; July 10, 2026; accepted July 13, 2026
Available online July 25, 2026
Abstract: A bi-level deep reinforcement learning-based, geographically digitized, thermal-state-aware energy coordination framework is proposed to address the limitation that temperature monitoring is typically treated as an external diagnostic rather than a control variable. The model embeds aerial electric-drone temperature monitoring, morphological features, and chromaticity analysis directly into the dispatch logic within a thermally coupled green energy hub-aerial e-drone charging system. A hierarchical Feudal Reinforcement Learning structure supported by the twin delayed deep deterministic policy gradient co-optimizes generation priority, cost reduction, and inspection strategy under uncertainty. Results show a mean daily operating cost reduction of 8 % compared with the twin delayed deep deterministic policy gradient and 15 % relative to the distributed distributional deep deterministic policy gradient, remaining within 5 % of the optimal solution obtained using a perfect mixed-integer linear programming solution. Over 100 days, coordinated drone-enabled energy exchange decreases total expenditure by 9 %. Full feature-based monitoring reduces the integrated cost by more than 11% and the Thermal Risk Index by 61%. Forced combined heat and power derating events decrease from 47 to 12, while inspection coverage reaches 96 %.
Keywords: Deep reinforcement learning; temperature monitoring; geographically digitalized energy coordination; thermal-state-aware.
Copyright © Journal of Engineering, Project, and Production Management (EPPM-Journal).
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
Requests for reprints and permissions at eppm.journal@gmail.com.
Citation: Chen, Y., Xie, S., Zhang, K., Fan, Y., and Zhou, G. (2026). A Machine Learning-Based Temperature Monitoring Method for Power Equipment by Fusing Morphological Features and Chromaticity Analysis. Journal of Engineering, Project, and Production Management, 16(5), 2026-0056.
