Application of artificial intelligence and optimisation techniques for fault detection and classification in power transformers

Altayef, Ehsan 2023. Application of artificial intelligence and optimisation techniques for fault detection and classification in power transformers. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (9MB) | Preview
	PDF (Cardiff University Electronic Publication Form) - Supplemental Material Restricted to Repository staff only Download (145kB)

Abstract

The thesis reviews the core lamination faults in power transformers and the testing and condition monitoring methods to confirm that artificial intelligence has not yet been applied to diagnose the edge burr faults and the degradation of the lamination insulation fault in the power transformer core. Experimental tests are conducted on the transformer core to observe the effects of faults. A clamping system is designed to apply artificial burrs in a repeatable manner in the range of 0.5T to 1.8T, while insulation faults are created by removing insulation material, short-circuits of 2, 6, 8, and 12 laminations were considered for flux densities of 0.5, 1.0, 1.5, 1.7, and 1.8 T. The faults were generated at different locations on the transformer core, and a total of 5 sites were selected. Artificial intelligence techniques are applied to investigate the impact of edge burr faults and insulation degradation on the total power loss in a 15 kVA three-phase transformer core. Features are extracted from current signals and thermal images and used as input vectors for training and testing using KNN, SVM, and DT classifiers. The accuracy rate obtained from the classifiers was over 90%. A new feature extraction technique called the RGB technique was presented for thermal images. A hybrid model of SVM with BOA and PSO optimization algorithms was developed to enhance the SVM classifier. The results from BOA-SVM were more appropriate for current signals data, while the results from BOA-SVM were the same for thermal image data. The results showed that for a large number of laminations affected by both faults, the overall core losses were doubled. And the results showed a satisfactory accuracy rate for fault detection and classification.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Engineering
Uncontrolled Keywords:	Power transformer Lamination insulation fault Lamination edge burrs Power losses Leakage current Artificial intelligence
Date of First Compliant Deposit:	14 April 2023
Last Modified:	14 Apr 2024 01:30
URI:	https://orca.cardiff.ac.uk/id/eprint/158881

Actions (repository staff only)

Edit Item