To address the impact of the uncertainty from distributed energy resources and loads on the reliability of fault recovery strategies in active distribution networks, this paper explores the potential of collaborative participation of distributed generation and load in fault recovery. A robust fault recovery method based on scenario clustering is proposed. This method first generates a joint scenario set using historical operational data from the resources and loads, and employs a probabilistic distance reduction technique to obtain a typical scenario library. Subsequently, a two-layer robust fault restoration model for active distribution networks, considering island partitioning and source-load interaction., is established. The model is solved using a nested iterative solution method, which generates a robust fault restoration scheme. Finally, simulations are conducted using the IEEE 123-node test distribution network, and the results validate the effectiveness and superiority of the proposed robust fault recovery method.