Abstract: Multi point faults lead to complex changes in distribution network topology, which greatly increases the difficulty of fault recovery, resulting in poor power supply recovery effect. To solve this problem, a fast fault recovery technology for distribution network based on breadth first search is proposed. Firstly, the topology of distribution network after fault is analyzed, and the matrix method is used to accurately locate the fault, and then the fault area to be recovered is divided to ensure the topology consistency of recovery path search; Secondly, taking the fault point as the starting point, the breadth first search algorithm is used to traverse the distribution network topology layer by layer to quickly explore all feasible power supply recovery paths; Finally, through the comprehensive evaluation of multi-objective indicators such as switch action times, network loss and power loss load recovery rate, the optimal recovery path is selected and the switch operation is performed to restore the power supply in the non fault area. The experimental results show that: compared with the three traditional methods based on deep learning, quantum firefly algorithm and Harris hawk algorithm, after the application of this technology, the number of switch actions is the least (only 4), the network loss is the lowest (below 50 kW), and the recovery rate of power loss load is the highest (up to 75%). The conclusion shows that this technology can efficiently search the optimal recovery path with the least switching operation, effectively reduce the network loss and improve the power supply reliability; The topology aware path search mechanism adopted by it effectively avoids the problem of load omission caused by the lack of topology information or misjudgment, which is the key to significantly improve the recovery rate of power loss load.