@Article{ee.2026.085888,
AUTHOR = {Xiao Liang, Yang Wang, Tianhang Long, Gang Li},
TITLE = {Data-Driven RUL Prediction of Oil-Immersed Power Transformer in Energy Systems via IDOA-BiGRU},
JOURNAL = {Energy Engineering},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/energy/online/detail/27901},
ISSN = {1546-0118},
ABSTRACT = {Accurate prediction of the Remaining Useful Life (RUL) of power transformers is essential for maintaining the reliability and stability of modern electrical energy systems. This paper proposes a data-driven RUL prediction method for oil-immersed power transformers by integrating an Improved Dhole Optimization Algorithm (IDOA) with a Bidirectional Gated Recurrent Unit (BiGRU) network. The proposed framework leverages multi-source condition-monitoring data, including electrical characteristics, dissolved gases in transformer oil, insulation aging indicators, oil-quality parameters, and mechanical condition indicators. Logistic chaotic mapping is introduced to improve the diversity of the initial IDOA population, while stage-dependent parameter scheduling is used to guide the transition from global exploration to local refinement during BiGRU hyperparameter optimization. The optimized BiGRU model is then used to capture temporal dependencies in transformer degradation sequences and predict RUL values. Experimental results obtained from 110 kV and 220 kV oil-immersed transformers show that the proposed IDOA-BiGRU model achieves improved prediction accuracy compared with the baseline models. For the decommissioned transformer samples with actual RUL labels, the proposed model achieves a mean absolute error of 0.8897 years. These results indicate that the proposed method can provide a data-driven auxiliary reference for condition-based maintenance, transformer asset management, and reliability-oriented operation of power grids.},
DOI = {10.32604/ee.2026.085888}
}