To address the mechanical fault diagnosis of high-voltage disconnect switches in substations, which are affected by complex environments and prolonged operation, a fault diagnosis method based on Gramian angular field (GAF) combined with convolutional neural network (CNN) for feature extraction and multi-strategy improved sparrow search algorithm (MISSA) to optimize extreme gradient boosting (XGBoost) parameters was proposed. Vibration signals were converted into two-dimensional images using GAF, features were extracted through CNN, and MISSA was employed to optimize the XGBoost classifier, thereby constructing a fault diagnosis model. Experimental results indicate that the proposed method achieves an identification accuracy of over 96%, effectively diagnosing issues like mechanism jamming, looseness, and phase mismatch, providing a theoretical basis for condition-based maintenance of high-voltage disconnect switches.