There are problems such as uncertain power output and load fluctuation when the distribution network with distributed generator (DG) operates on an island. In order to ensure stable and reliable power supply for islands, a dynamic island partition strategy is proposed which takes into account the standby demand of islands and gives full play to the flexible load regulation ability. Considering the uncertainty of wind power and photovoltaics, a prediction error model is established at first. The depth-first search algorithm is used to obtain the maximum power supply range for each time period, and chance constrained programming is used to deal with the random variables in the model, the island standby demands for each time period is obtained by the maximum power supply range. Then, modified particle swarm optimization (PSO) algorithm is used to give full play to the control ability of flexible load, and the optimal islanding results that meets the standby demand under each confidence levels is obtained with the objective of maximizing the comprehensive recovery value during the failure recovery period. Finally, a modified IEEE 33 node power distribution system is taken as an example to verify that the proposed strategy can effectively increase the value of islands while ensuring the reliability of islands.