There are many single-phase loads in the low-voltage distribution network, such that an unbalanced three-phase load will increase the line loss in the substation distribution network and endanger the safety of power grid operation. In this paper, a method of user phase sequence optimization based on historical data is proposed. The historical voltage and current data of users in the distribution network are used to construct the unbalanced power flow model of the distribution network. The objective function of the minimum average voltage imbalance degree of the user node, the minimum line loss and the minimum commutation times in the station area are established according to the operating conditions of the station area within one day. An improved non-dominated sorting genetic algorithm-3(NSGA-3) with a normal distribution crossover (NDX) operator is proposed, and the phase sequence of the user nodes is optimized to obtain a better population distribution and reduce the optimization time. Then the most suitable set of solutions is selected from the solution set as the phase sequence adjustment strategy. Finally, the results validate that the proposed method can effectively reduce the degree of three-phase voltage imbalance and reduce line loss by taking the actual load data of a distribution station area in Anhui Province as an example. This work is supported by the National Natural Science Foundation of China (No. 51607001, No. 52177027, and No. 52107034).