Abstract: The reliability of the secondary system directly affects the safe and reliable operation of the entire substation and even the power grid. With the high proportion of renewable energy sources integrated into power systems, it is increasingly important to effectively detect the secondary system faults. However, existing research faces two challenges: on the one hand, the existing fault detection methods for logic circuits have high requirements for data completeness and are difficult to apply in practice; on the other hand, the existing fault detection methods for secondary equipment are difficult to identify the small differences between normal and fault data, and the computational accuracy is difficult to guarantee. Therefore this paper proposes a secondary system fault detection method based on association rules and reconstruction errors. Firstly, the Apriori algorithm is used to derive the association rules between fault alarm information and fault devices in the logic circuit, achieving rapid diagnosis of logic circuit faults. Then, the individual discriminator is trained using the operational data of normal secondary equipment, and the operating status of the secondary equipment is determined by measuring the reconstruction error of the data to be discriminated. The ensemble learning model is used to quantify the current fault detection probability of the equipment. Finally, the ensemble learning model is optimized to improve the accuracy of secondary equipment anomaly warning. The effectiveness and accuracy of the proposed method was verified by simulating the dataset from a substation in Henan Province.