Diagnostic study of insulation defects in zero-carbon DC system based on GRU-ATT

The current insulation defect detection method of DC system has problems such as complicated calculation and large error.For this reason, this paper proposes a GRU-ATT based insulation defect diagnosis method for DC systems based on the Gated Recurrent Unit (GRU) model, combined with the attention mechanism (ATT).First, the simulation model of DC system is built to simulate the insulation defects of each branch and collect the current signals; then the collected signals are subjected to noise reduction and feature extraction, and the feature data are downscaled by Principal Component Analysis (PCA) to generate new feature vectors; Finally, based on the GRU-ATT neural network model, three different neural network models are established to conduct comparative experiments for insulation defect diagnosis using the new feature vectors as input data.The experiments show that the identification accuracy of insulation defects in all branches using the GRU-ATT model reaches more than 96.77%.