With the increasing number of distributed power sources, power electronic transformers, flexible loads and their corresponding monitoring and measuring devices in distribution network, the difficulty of decision-making control in distribution network is increasing day by day. There are many differences in sampling time scale, data structure and measurement accuracy in new distribution network multi-source measurement, so the efficient fusion of multi-source measurement data becomes an important prerequisite for the integration and application of multi-source information data in distribution network. In this paper, the measurement data characteristics of the new distribution network measurement system are analyzed, and the multi-source data fusion strategy based on the phasor measurement unit (PMU) is adopted. On this basis, a multi-source measurement fusion method based on untrace Kalman filter is proposed in order to solve the problems of nonlinearization error and high dimension difficulty in calculating Jacobian matrix in extended Kalman filtering. Finally, the mixed quantity measurement of IEEE33-node power distribution test system is used to verify that the data fusion method can improve the efficiency of multi-source heterogeneous data fusion and reduce the relative error of node voltage and power estimation.