Aiming at the problems of low efficiency and insufficient accuracy of existing relay protection cabinet drawing verification methods, this paper proposes a multi-scale convolutional neural network (MSCNN-SAM) method based on wavelet convolution kernel and semantic attention mechanism. The method uses the multi-scale feature of wavelet transform to construct a convolutional kernel, effectively extracts multi-scale features in the drawing image, and combines with the semantic attention mechanism to focus on the key semantic information regions in the drawings, improving the ability to capture detailed features and important parts. Compared with traditional convolutional neural networks, MSCNN-SAM achieves higher accuracy and recall in the drawing verification task. The experimental results show that the method can significantly improve the efficiency and accuracy of relay protection cabinet drawing verification, which provides strong support for ensuring the safe operation of power systems.