In the event of a network failure that compromises energy supply, the characteristic of smart grid self-healing consists of finding a proposal for reconfiguration of the grid, aiming at restoring the power, partially or completely to supply all network nodes. The search for a satisfactory reconfiguration is a combinatorial problem whose complexity is proportional to the network size. An exhaustive search-based method is time-consuming and often computationally non-viable. To overcome this difficulty, techniques for generating minimal spanning trees (MSTs) of the graph that represents the smart grid, are exploited. However, existing studies provide centralized implementations. In this work, we propose a distributed implementation, where each of the network switch collaborates in developing of the recovery solution. The proposed decentralized approach seeks a reduction of reconfiguration time requirements, thus increasing network intelligence. For this purpose, a distributed algorithm for building the MST is embedded in the processing elements available at the commutation nodes. The evaluated case studies show that, whenever possible, the proposed solution allows for a successful reconfiguration, regardless of the number of simultaneous failures. Moreover, the network reconfiguration time is not significantly impacted by the number of buses and included lines. The implementation presents results of communication cost and reconfiguration time significantly lower than the expected upper bounds. Notably, for the case studies, the proposed implementation achieves 70% and 69% speedup regarding the reconfiguration time from simple and multiple failures, respectively, when compared to the expected theoretical performance. Furthermore, when compared to existing multiagentbased self-healing systems, the proposed implementation recovers from failures twice as fast, making it more desirable in a smart-grid real implementation.