Abstract: With the large-scale integration of distributed generation sources into distribution networks, traditional protection methods face challenges such as bidirectional power flow and complex fault characteristics. This paper proposes a self-healing method for differential protection in active distribution networks based on multiagent technology. By constructing a hierarchical multi-agent architecture, rapid fault identification and isolation are achieved, and power supply restoration for non-faulted areas is realized through coordinated optimization. Considering the characteristics of active distribution networks with distributed generation sources, the method establishes a differential protection model that accounts for the dynamic output of distributed sources. It formulates protection criteria using the amplitude ratio and phase difference of current fault components at both ends of the protected zone and develops a multi-agent collaborative self-healing optimization model aimed at maximizing load restoration priority and minimizing network losses. Simulation results show that the proposed method can effectively handle different types of faults, exhibits strong tolerance to transition resistance and anti-interference performance, and demonstrates excellent performance in terms of operation counts, line losses, and power factor in self-healing.