Answers are grounded in this paper and cite their sources.
As a key component of the power grid, accurate fault diagnosis of power transformers is essential for reliable operation. Data-driven methods have become important tools for transformer condition monitoring, but their performance can degrade when training and testing data come from different operating conditions. This limitation becomes more evident when target-domain samples are unavailable. To improve diagnostic accuracy and mechanism-informed representation under unseen load-voltage conditions, this paper proposes a Mechanism-guided Semantic Heterogeneous Meta-generalization Model (MS-HMSM) framework. The proposed method integrates mechanistic knowledge with deep learning through three modules. First, the Heterogeneous Physical-Mechanism Semantic Distillation (HPMSD) module extracts domain-invariant semantic representations from operating-condition information and transformer physical priors. Second, the Cross-Semantic Adaptive Enhancement Fusion (CSAEF) module uses cross-attention and adaptive gating to fuse complementary data-driven and semantic features. Third, the Domain-Consistent Meta-Regularized Identification (DC-MRI) module combines prototype learning and semantic projection within a meta-learning framework to improve generalization under unseen load-voltage conditions. Experiments on an operating 110 kilovolt (kV) transformer platform show that the proposed method achieves average diagnostic accuracies of 95.16% and 96.67% under unloaded and loaded conditions, respectively, and outperforms the compared benchmark methods in the tested settings.
No takes yet. Share an insight, caveat, or question.
Wang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: