Comprehensive Evaluation Method of Smart Substation Relay Protection System Based on Improved Deep Transfer Learning

Aiming at the existing comprehensive evaluation method of relay protection system which does not effectively consider the condition of insufficient and unbalanced training samples and fails to effectively extract the deep features of data during feature extraction, a comprehensive evaluation method of smart substation relay protection system based on improved deep transfer learning is proposed. First, according to the composition of smart substation relay protection system and the actual monitoring needs, the state evaluation index set is constructed, and the expanded historical state data is used as twin data to construct the protection system evaluation model using sparse stack self-encoder. Secondly, the model is dynamically updated using online monitoring data to realize the two-way interaction between the twin and the physical entity. Finally, the improved deep transfer algorithm combined with the digital twin utilizes realtime data to drive the dynamic update of the model to obtain the real-time operation status of the protection system. Experiments show that the accuracy rate of the proposed method is 0.937, which are higher than the compared algorithms.