Abstract: With the expansion of the scale of the interconnected power grid, the requirements for the calculation speed in the power system analysis are increasing. The Advanced Vector Extensions (AVX) is an efficient instruction set proposed for large-scale floating-point operations, which can increase the floating-point computing capability by up to 16 times. Taking the three-phase power flow calculation as the research object, the fast YBUS power flow algorithm is adopted for calculation, and the vector instruction set is introduced to improve the calculation process. By transforming the process of solving iterative equations, and taking the 8×8 matrix and 8×1 vector as the operation units, the AVX instruction set is used to realize the basic calculation and the inverse vectorized algorithm of the operation units, and the equations are divided into blocks to be solved by lower-upper (LU) decomposition based on the vectorized operation unit. The 6×6 matrix and 6×1 vector corresponding to the three-phase model of the node admittance matrix of the three-phase power flow equation are extended into the operation unit of the 8×8 matrix and 8×1 vector, and other parts are filled into the operation unit. By adopting the Tinney-2 ordering scheme, a sparse partitioned matrix based on the operation unit is established, and the parallel solution of the three-phase power flow equation is realized based on the vectorized overloading of the operation unit operator. The three-phase system cases IEEE24 and IEEE57 are used for testing. The results show that the three-phase power flow vectorization parallel method has accurate calculation results. Compared with the traditional power flow method, the vectorization parallel method has high efficiency in time performance.