Aiming at the comprehensive evaluation of the impact of power failure on power systems, a new Improved Extension Analytic Hierarchy Process (IEAHP) is presented in this paper. Firstly, the indicator system is established to fully characterize the impact of each power failure. Secondly, the power failure levels are divided based on historical data of power failures and experience of evaluation experts, and then the extension set theory is used to calculate the membership of the indicators to the levels. Thirdly, the judgment matrix is constructed by the extension interval in order to reduce the experts' subjective empowerment vagueness, then the judgment matrices are combined through the objective allocation of expert weights so as to provide a comprehensive assessment of the impact of power falure. Finally, the proposed method is applied to the assessment of real power failures, and the results prove its effectiveness. This work is supported by National Natural Science Foundation of China (No. 51307064), Science and Technology Project of China Southern Power Grid Company Limited (No. 000000KK52150001), and Fundamental Research Fund for the Central Universities (No. 2017ZM0022).