Abstract: For large system harmonic impedance estimation errors caused by the random fluctuations of background harmonic in existing methods,this paper proposes a system harmonic impedance estimation method based on probability statistics.First, a probability model of system harmonic voltage is constructed using a 3-order Weighted Gaussian Mixture Distribution function to reflect the random fluctuations of background harmonic. Second, the logarithmic likelihood equation of the system harmonic impedance is constructed,and the Expectation Maximization method is improved using Deterministic Annealing algorithm so that the estimation accuracy and iteration speed of the hidden variable is improved,and the estimation of the utility harmonic impedance under background harmonic random fluctuations is achieved. The KL divergence and error are used to evaluate the accuracy of the background harmonic probability model and the estimation accuracy of the system harmonic impedance of the proposed method. Simulation experiments and measured data experiment are used to analyze the estimation effect of the proposed method. The experimental analyses show that the proposed method has strong robustness and high accuracy in estimating the system harmonic impedance.