融合组合采样和IBKA-KELM的油浸式变压器故障诊断方法OA

To address the issue of low diagnostic accuracy caused by the insufficient number of transformer fault sam-ples and the imbalance characteristics,a transformer fault diagnosis method is proposed,which integrates a combined sampling method based on synthetic minority over-sampling technique(SMOTE)and the weighted edited nearest neigh-bor algorithm and uses an improved black-winged kite algorithm(IBKA)to optimize the kernel-based extreme learning machine(KELM).First,a weighted heterogeneous value difference metric space is constructed based on the imbal-anced fault dataset to reconstruct the sample neighborhood relationships.Then,the edited nearest neighbor algorithm is incorporated to remove the over-sampling noise,enhance the local density of insufficient samples and inter-class feature distinction,and provide sufficient and balanced samples for training the diagnostic model.In addition,a gas feature ra-tio matrix is established,and the maximum information coefficient combined with random forest is utilized to select the optimal feature subset,so as to enhance the capability of feature representation.Finally,the IBKA improved by a hy-brid strategy is used to optimize the structural parameters of KELM,and an IBKA-KELM fault diagnosis model is con-structed to achieve an accurate identification of multi-class fault samples.Experimental results demonstrate that com-pared with the conventional over-sampling techniques such as SMOTE and adaptive synthetic sampling(ADASYN),the proposed combined sampling method effectively improves the feature representation in imbalanced samples and achieves diagnostic accuracy of 96.05%,thereby validating its effectiveness.