Because of the problem that celestial observation information is affected by the external environment and photoelectric signals easily, which leads to performance degradation of starlight refraction navigation system, fault detection algorithm for starlight refraction angle using star sensor is proposed. The error model of starlight refraction angle based on inertial navigation information is established. Combined with the measurement noise characteristics of the star sensor, the algorithms for range rationality detection, trend rationality detection, and jump detection are designed, and the faults such as field value and jump change in the starlight refractive angle measurement information are detected respectively. The experimental verification of the actual observation data shows that the proposed method effectively solves the problem of fault detection of the starlight refraction angle in starlight refraction navigation system, and lays a foundation for the engineering implementation of the dual-field starlight refraction navigation system.