Abstract: The growing penetration of new energy sources in the power system has significantly increases the uncertainties in the grid, asking for higher requirements for the planning, operation and control of the distribution network. The distribution network planning is an important cornerstone for the safe and stable operation of the power system. The traditional distribution network planning, in which all parameters are determined in advance, lacks adaptability to uncertainties. In view of this, we proposed a method for distribution network planning based on probabilistic power flow analysis. The source-load output model was firstly established according to the quantitative modeling of uncertainties in the distribution network by using our method. Secondly, we utilized the rank correlation coefficient matrix to characterize the correlation between wind speed, light intensity and load, and developed a semi-invariant probabilistic power flow calculation method with correlation taken into account. Finally, with the objective function of reducing the comprehensive cost, we constructed a distribution network planning model with the constraints of feeder capacity, node voltage, tidal balance, and radial structure of the grid. And the particle swarm algorithm was improved by optimization of inertia parameters and incorporation of variational operations. The improved algorithm was employed to solve the planning model. Simulations were conducted taking a 33-node system as an instance, and the results confirms the effectiveness of our method in reducing network loss and network planning costs.