Interaction of neural network and dung beetle optimizer for characterization of large power transformers considering through-fault protection capability

• Apply a hybrid DBO-ANN based methodology in capturing the steady state and faulty stamps of transformers with various categories. • Validate the proposed technique with other metaheuristic approaches in defining the optimal results. • Crop the transformers full load technical parameters identifying its operation with various loading conditions. • Allocate the through-fault protection characteristics of large power transformers considering both electrical and mechanical damage. Several computational methods aim to accurately identify unknown parameters of transformers (XFMRs) through appropriate objective functions, but there is a need for more accurate models for large power XFMRs in both normal and faulty conditions. This paper presents a novel approach for estimating unknown parameters of power XFMRs by combining metaheuristic optimization with artificial neural networks (ANN). Six unidentified parameters are determined using the dung beetle optimizer (DBO), whose effectiveness is validated against various challenging optimizers. The algorithms capture the steady-state fingerprints of test cases regarding voltage regulation and efficiency characteristics under varying loading conditions. Moreover, ANN is employed to assess XFMR operations in malfunctioning scenarios by predicting protection through-fault capabilities. This includes analyzing frequent fault characteristics due to mechanical damage and infrequent electrical degradation faults via actual measurements of XFMR parameters. The proposed method is applied to a real-case power XFMR of 40 MVA and a benchmark 4 kVA XFMR, confirming that frequent fault characteristics for the 40 MVA XFMR begin at 4.06 times the rated current. The findings illustrate the synergistic capabilities of ANN and DBO in characterizing key parameters and performance of power XFMRs.