The access of distributed renewable energy enhances the uncertainty of the distribution network reliability, whereas the distribution network reliability evaluation using existing methods cannot fully reflect the stochastic volatility of islanding power supply and loads. For this reason the approach of active distribution network probabilistic reliability evaluation based on point estimation method is proposed. In the framework of Monte Carlo simulation, the active distribution network failed subzone is determined and the probabilistic models of islanding random variables are established firstly; then the islanding probabilistic reliability is assessed with a nonlinear transformation method of independent random variables based on point estimate method; finally, system probabilistic reliability indexes can be obtained from the probabilistic superimposing of the results of repeating simulated failures, achieving a comprehensive assessment and analysis to the active distribution network reliability. The simulation on the transformed multi-branch feeder model of the IEEE RBTS Bus6 verifies the validity and accuracy of this assessment approach.