With the continuous development of the power Internet of Things, various types of power terminals are widely constructed and applied. Massive data is aggregated and processed on cloud platforms, which puts great pressure on network bandwidth and cloud platforms. Edge computing has effectively alleviated the computing pressure of cloud platforms and reduced the consumption of network transmission bandwidth, but the introduction of edge computing has also brought new security problems, and resource constrained power terminals face more security challenges. Access authentication is an important part of network security mechanisms. Currently, passive device fingerprint recognition methods based on data traffic do not consider the temporal arrival order of data packets and cannot extract their deep features. This paper proposes an identity authentication method for power digital intelligent terminal based on 1D-CNN transformer. This method extracts device features from data packets generated during the device network configuration phase to construct passive device fingerprints, and uses a power device fingerprint authentication method based on a 1D-CNN Transformer hybrid model to extract deep features from device fingerprints. To improve the device recognition ability, the article uses a fixed window sliding mechanism and SMOTE algorithm to enhance data from two stages of feature extraction and vectorization processing, solving the problem of data imbalance and removing interference vectors. The simulation results show that this method can effectively identify device identity, and compared with traditional machine learning and deep learning, the device recognition accuracy of the method proposed in the article has been improved by 6%.