Abstract: If defects in drainage plates are not detected and addressed in a timely manner, they can cause a gradual increase in contact resistance, which may lead to fuse blowouts and affect the stability of the power system. Ultrasonic detection methods identify defects in drainage plates by analyzing differences in acoustic impedance; however, noise and interlayer echo interference make it difficult to extract defect features from ultrasonic B-scan images. To address this, this paper proposes a defect detection and intelligent recognition method for drainage plates based on the Swin Transformer model. The Swin Transformer effectively mitigates noise interference and feature extraction difficulties in B-scan images, enabling accurate and efficient identification of drainage plate defects. First, the optimal ultrasonic frequency range is determined through simulations, and an experimental platform is set up for ultrasonic testing of drainage plates. Then, synthetic aperture focusing technology (SAFT) and phase coherence techniques are employed to improve image quality and resolution while reducing noise interference. Finally, comparative experiments with five deep learning models demonstrate the superiority of the Swin Transformer in drainage plate defect detection.