Abstract: Digital twin technology is the key path to realize the development of main equipment protection from static assessment to dynamic assessment. In order to improve the accuracy of the main equipment protection status discrimination results, mitigate the adverse effects of insufficient and unbalanced training samples on discrimination results, and avoid the drawbacks of lagging results caused by relying solely on historical status data for status discrimination, a method for constructing a digital twin-driven expert model for main equipment protection status discrimination is proposed. The expert model is used to discriminate the operating status of the main equipment protection system. Firstly, the architecture of the intelligent substation main equipment protection digital twin system is established. Secondly, a stack sparse autoencoder is used to construct the expert model for main equipment protection status dis-crimination, and the deep transfer learning algorithm and digital twin technology are utilized to improve the expert model. Then, a set of indicators for main equipment protection status discrimination is constructed based on the multi-source information of protection devices and the actual monitoring requirements. Finally, the main equipment protection system is discriminated based on the improved expert model and real-time data. Case studies have shown that this method can construct an expert model for main equipment protection status discrimination, and the accuracy of the model discrimination reaches 97.28%, which is 7.02% higher than that of the traditional model without online monitoring data correction. The accuracy and flexibility of state discrimination are improved, and the stability of relay protection of main electrical equipment is enhanced.