The invention discloses a power distribution network topology calibration method based on a k nearest neighbor classification algorithm, comprising the following steps: (1) acquiring other users in a court where a calibration user is located and all users in adjacent courts to form a training sample set of the calibration user; (2) extracting the voltage curve data of the calibration user and of all the users in the training sample set in a most recent period of time; (3) calculating the voltage curve similarity between the calibration user and each user in the training sample set; and (4) selecting k nearest neighbors of the calibration user based on the similarity, calculating the correct court category of the calibration user, and judging whether the power distribution network topology of the calibration user is correct or not. Through the method, the correctness of power distribution network topology data in a power grid GIS platform can be verified effectively. Moreover, the method has the advantages of low cost of required materials and manpower, high accuracy, and high operability.