For comprehensive evaluation of a distribution network, this paper proposes an improved AHP-CRITIC combination weighting and extension evaluation model based on least squares. Since the single weighting method has a certain bias, the subjective weight and the objective weight are determined by an Analytic Hierarchical Process (AHP) and CRITIC method respectively. A combined weighting model based on least squares is constructed and solved by a genetic algorithm to determine the comprehensive weight. The subjectivity of the classic domain determination in the extension evaluation model is constructed by introducing triangular fuzzy numbers. To ensure the accuracy of the existing correlation degree and the calculation accuracy of the eigenvalues of the rank variables, we propose a correlation degree correction method and the rank expansion method. The calculation example shows that the method can classify the distribution network scientifically and reasonably, and the evaluation result has high credibility. This work is supported by the National Natural Science Foundation of China (No. 51777142) and the Electric Power Planning Special Research Project of Guizhou Power Grid Co., Ltd. (No. 060000QQ00190011).