Abstract: With the wide application of power electronic devices in new energy resources and loads, the harmonic distortion degree in power systems is increasing. Meanwhile, the uncertainty of source and load further increases the difficulty of harmonic analysis. To effectively evaluate the harmonics and fully mine the measured data characteristics, a data-driven probabilistic harmonic power flow (PHPF) calculation considering source and load uncertainties is proposed. First, based on the measured data, the dynamic harmonic coupling matrix model (DHCMM) is established for source and load to reveal the coupling between voltage and current harmonics. Then, the time-varying uncertainty characteristics are determined with the measured data, and the statistical features are extracted by using the improved point estimation method to overcome the estimation deviation caused by the interaction between variables. Finally, a PHPF calculation method considering source and load uncertainties is proposed to evaluate time-varying uncertain harmonics in the power system. According to the analysis results, the effectiveness of data-based DHCMM is verified, and the accuracy of the proposed PHPF can accurately evaluate the harmonics based on time-varying uncertain power states.