【摘要】 变压器在电力系统中发挥着至关重要的作用，为了保证变压器故障诊断的可靠性，提出了基于合成少数类过采样技术(SMOTE)的TDBO-SVM变压器故障诊断模型。首先，通过SMOTE均衡数据集，降低不平衡故障数据对模型诊断精度的影响；其次，引入了SPM混沌映射、可变螺旋搜索策略、Levy飞行策略、自适应t分布扰动变异对蜣螂算法进行改进；随后利用TDBO对SVM的惩罚参数和核参数进行组合寻优，建立TDBO-SVM模型；最后，对不同变压器故障诊断模型进行实例仿真对比，验证了该模型在油浸式变压器故障诊断中具有较高的诊断精度和较好的收敛性。更多还原

【Abstract】 Transformers play a vital role in power systems, and in order to ensure the reliability of transformer fault diagnosis, a TDBO-SVM transformer fault diagnosis model based on synthetic minority oversampling technique(SMOTE) is proposed. Firstly, the impact of unbalanced fault data on the diagnostic accuracy of the model is reduced by SMOTE balanced dataset. Secondly, SPM chaotic mapping, variable helix searching strategy, Levy flight strategy and adaptive t-distribution perturbation variantare introduced to improve the Dung Beetle algorithm. Then use TDBO to optimize the penalty parameters and kernel parameters of SVM, and establish the TDBO-SVM model. Finally, example simulation and comparison of different transformer fault diagnosis models are carried out, and verify that the model has a higher diagnosis accuracy and a better convergence in the fault diagnosis of oil-immersed transformer.更多还原