Discrimination Method of Aging Degree of Oil Paper Insulation Based on Raman Spectroscopy and Machine Learning Interpretable Model
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
In view of the destructiveness and limited information sources of traditional transformer oil paper insulation aging detection, this study proposes a nondestructive and interpretable oil paper insulation aging degree method based on Raman spectroscopy combined with advanced machine learning algorithm Firstly, the oil samples at different aging stages were obtained by accelerated thermal aging test and Raman spectrum detection; Then, combined with the improved competitive adaptive reweighted sampling (CARS) and Support vector machine (SVM) algorithm, the discrimination model of oil paper insulation aging degree is established; Finally, the key wavelength variables are resolved by Shapley additive interpretation (SHAP) The experimental results show that the improved CARS-SVM model is superior to the full spectrum-SVM in accuracy, accuracy, recall and F1 value, which can accurately identify the aging degree of oil samples, and the selected characteristic wavelength can effectively characterize the change of chemical bond vibration characteristics during the aging process of insulating oil.
Fields of Study & Subjects
Power Transformer Diagnostics and Insulation
Spectroscopy and Chemometric Analyses
High voltage insulation and dielectric phenomena
Publication details
- Journal / source
- International Conference on Electrical Automation and Artificial Intelligence (ICEAAI)
- Pages
- 1233–1236
- Publication date
- 16 January 2026
- Language
- EN