False data injection attacks (FDIA) are attacks targeting power system state estimation, which are difficult to detect by static residuals and mostly rely on dynamic time series data for detection. Therefore, a FDIA detection and state restoration based on power system interval dynamic state estimation (IDSE) is proposed. The FDIA detection is based on time series data, using its time-varying and asymmetrical consistency intervals as the consistency features of time series, and replaces the traditional fixed threshold for FDIA detection. The state restoration is based on the IDSE model. Compared with the conventional point estimation model, IDSE can obtain more secure and reliable state restoration results. The simulations based on the IEEE 14 case and data from the New York independent system operator have demonstrated the effectiveness of the proposed detection and state restoration method, and the advantages of the proposed method have been demonstrated through comparison with deterministic state restoration methods.