The traditional voltage sag frequency estimation considering protection action characteristics needs to obtain detailed protection configuration information. However,the protection configuration of distribution network is diverse. Under the influence of different factors such as transition resistance,generator scheduling and fault type,the protection areas of each stage of stage protection may change greatly,and the traditional method may produce large error in the evaluation of voltage sag duration. In this paper,a method of voltage sag frequency estimation based on improved K-means clustering algorithm is proposed. In the case of unknown line protection configuration,based on the historical monitoring data of voltage sag and protection action information,the improved K-means clustering algorithm is used to cluster the voltage sag amplitude duration,infer the line protection configuration,and calculate the protection action time and protection action sag threshold. According to the calculation results,the distribution network voltage sag frequency is estimated considering different fault types,different operation modes and transition impedance. The simulation of the bus 5 distribution network in IEEE RBTS-6 bus test system verifies the effectiveness and superiority of the proposed method.