In order to improve the accuracy of probabilistic power flow calculation with wind power system, a method combining Quasi Monte Carlo and semi invariant is proposed in this paper. The establishment of wind power model, the quasi Monte Carlo method (QMCS) to extract Sobol deterministic low discrepancy, combined with the semi invariant method, semi invariant computing node state variables and branch power flow, the introduction of Gram-Charlier series expansion, and the probabilistic load flow calculation tool of node voltage is simulated and compared with the traditional Monte Carlo method and Monte Carlo simulation with semi invariant contrast method. The calculation results show that the results of QMCS combined with the semi invariants method under the same sampling times are closer to the true solution, and the calculation speed is relatively faster.