Randomness and correlation of multiple distributed generators and fluctuating loads in close proximity brings multiple uncertainties to the model of distribution network. In allusion to the problem, a probabilistic reactive power optimization model considering multiple uncertainties is built and solved as a multi-scene deterministic optimization problem. In the model, third-order polynomial normal transformation (TPNT) simplified Nataf transformation is used to decouple correlation of distributed generators and fluctuating loads, and Gaussian-Hermite integration method is used for probabilistic power flow calculation in objective functions and constraints. The model is optimized by MSI-PSO, and a certain optimal scheme of control variables and statistical moments of state variables are achieved. The simulation is carried out on IEEE 33-bus system with multiple photovoltaic generators and fluctuating loads, and the result verifies the adaptability and availability of the proposed model and strategy for probabilistic reactive power optimization.