In response to the common compound fault issues in outdoor photovoltaic arrays,a hybrid network model combining residual networks and vision transformers have been proposed and optimized using transfer learning meth-ods to enhance the reliability of fault diagnosis models in compound fault scenarios.Firstly,effective features are ex-tracted from the static I-V curves and environmental parameters of the photovoltaic arrays as inputs.Then,the model is pre-trained using simulation data.Finally,the reliability of the model in diagnosing real experimental data is veri-fied through transfer learning.The experimental results indicate that this hybrid model exhibits a higher convergence speed and accuracy when dealing with compound fault scenarios.
维普
万方数据