State Recognition of Operating Mechanism Based on Transfer Learning of Simulated Electromagnetic Coil Current Features

The state recognition of distribution switchgear operating mechanisms is crucial for power grid safety. The current of the electromagnetic coil contained a wealth of status information about the operating mechanism, which can reflect its operational condition. Aiming at the typical faults in electromagnetic coils and latching mechanisms of 10 kV distribution switchgear operating mechanisms, a state recognition method is proposed, utilizing electromagnetic coil simulation and current characteristic transfer learning to analyze the operating mechanism status. A three-dimensional transient multi-physics coupling simulation model is established using ANSYS Maxwell. Two kinds of faults are considered including the increase of coil contact resistance and latch jamming. Current waveform characteristics are analyzed under these state scenarios, and a joint feature set is constructed based on the extracted parameters. To solve the distribution discrepancy between simulation data and real data, a hybrid kernel transfer component analysis(TCA) method is proposed. The method integrates dynamic regularization processing and a pseudo-label iterative optimization strategy. These techniques enable effective cross-domain transfer of simulation features to real-world scenarios. The experimental results show that the simulation model has an average error below 10%, which verifies the model's accuracy. The TCA-based K-nearest neighbors(KNN) algorithm achieved a state identification accuracy of 94.67% in real data. This provides a high-accuracy and low-data-dependent solution for operating mechanism state recognition and has great potential for engineering application.