Abstract: This paper proposed a method to calculate the utility harmonic impedance based on Bayes theorem. Firstly, according to the maximum entropy principle, prior distribution of utility harmonic impedance was a uniform distribution. Secondly, the real probability distribution of utility harmonic voltage at the downtime was estimated by kernel density estimation, thus conditional probability of utility harmonic voltage corresponding to the prior impedance was obtained. Thirdly, the posterior distribution was corrected from prior distribution by use of conditional probability on the basis of Bayes theorem. Finally, the utility harmonic impedance was calculated by minimizing the loss function. The proposed method is applicable to arbitrary background harmonic distribution and still accurate under large interference of background harmonic. In addition, the reliability of results was quantified by comentropy decrement and width of confidence interval from a probabilistic perspective. Based on analysis of simulation results and field test cases, it proves that the proposed method is effective and accurate.