In existing high?voltage scenarios, there are many harmonic data curves with similar CP95 values but significant differences in characteristics, which cannot be distinguished by the current single indicator of harmonic national standards. To solve this problem, a comprehensive evaluation method of high?voltage harmonics is proposed considering multi?dimensional probability and time?series characteristics of measured data. Firstly, based on the probability and statistics theory, fluctuation characteristic indices that characterize the data trend and local fluctuation conditions are proposed. Secondly, based on the wavelet transform theory, the time?series characteristics of the measured data are analyzed to construct an evaluation index system covering the data fluctuation characteristics and time?series energy characteristics. Then, in order to coordinate the objectivity of data and the subjectivity of human experience, a comprehensive evaluation model is established for high?voltage harmonics based on the weighted generalized Mahalanobis distance improved ideal solution combining the coefficient of variation method and the improved analytic hierarchy method. Finally, an calculation example is analyzed to verify the rationality of the proposed method. The analysis shows that the comprehensive evaluation model constructed can effectively characterize and evaluate the differences in the quality of high?voltage harmonics.