Modern power systems are featured with complicated structure, strong connection, and changeable operation patterns. Therefore, the result of traditional N-1 security analysis based on single study case is not sufficient to ensure the secure operation of the power system. Based on the forecasted nodal curve, this paper proposes a method of improved N-1 analysis, which is time process oriented and considers the load tendency, and hence provides a new way for early warnings and corrective control of subsequent fault. There are two steps in the method, firstly, a hierarchical clustering method is used to partition the nodal curve, and then the technology of data mining is used to extract the effective characteristic data that can represent the load variation of each time segment. Secondly, security analysis of initial contingency is carried out to perform early warnings for subsequent fault, which is triggered by the overloads caused by initial contingency and changeable load tendency. Adjustment measures for the unsecure time segment are developed with optimal power flow. To verify the feasibility and superiority of this method, a testing scenario is provided and the result shows the correctness.