Optimization of QCNN-ResNet Based on Sparrow Search Algorithm for Fault Diagnosis of Current Monitoring Circuits

Current monitoring circuits are integral to the inspection of wind power transmission lines, with their fault diagnosis being essential for the early identification of potential issues, thus mitigating safety risks. This paper addresses the fault diagnosis challenges in current monitoring circuits by introducing a novel diagnostic approach for analog current monitoring circuits, leveraging a Sparrow Search Algorithm (SSA) optimized Quadratic Convolutional Neural-Residual Neural Network (QCNN-ResNet). The proposed methodology employs the attention mechanism inherent in the Quadratic Convolutional Neural Network (QCNN) to bolster the model’s interpretability. Concurrently, it incorporates the skip connections of the Residual Neural Network (ResNet) to counteract gradient vanishing, thereby expediting model convergence. Furthermore, the SSA is applied to refine the iteration count and dropout rate, enhancing the network’s overall performance. The efficacy of the proposed method was substantiated through diagnostic tests on typical fault components within current detection circuits. Empirical results affirm that the proposed method significantly augments diagnostic precision.