A main feature of future smart grids is their self-healing capability, which allows fault location, isolation and service restoration to be done automatically. Restoration is a combinatorial non-linear optimization problem for which conventional mathematical programming techniques and heuristic methods become computationally very costly. This paper proposes a decentralized restoration method that avoids the computational burden of fully centralized optimization techniques while still providing good solutions and avoiding single points of failure. By pre-calculating the amount of power that can be transferred from other feeders without violating their limits, restoration within the outage area is converted into a binary integer linear optimization problem which can be easily solved. To prove its effectiveness, different case scenarios were tested on 14-bus and 70-bus systems. The multi-agent system was implemented in JAVA Agent Development Framework (JADE) and MATLAB was employed for the main algorithms within each agent. The outcomes were compared to those of previously proposed centralized and decentralized approaches, proving that the method proposed shows better computational efficiency than the centralized approach and better results than the decentralized approach. Therefore, the proposed algorithm provides an overall improved solution.

中文翻译：

---

未来智能电网中配电网恢复的多智能体系统

未来智能电网的一个主要特点是具有自愈能力，可以自动完成故障定位、隔离和服务恢复。恢复是一个组合非线性优化问题，传统的数学规划技术和启发式方法的计算成本非常高。本文提出了一种分散式恢复方法，该方法避免了完全集中式优化技术的计算负担，同时仍然提供良好的解决方案并避免单点故障。通过预先计算在不违反其限制的情况下可以从其他馈线传输的电量，停电区域内的恢复被转换为可以轻松解决的二进制整数线性优化问题。为了证明其有效性，在14总线和70总线系统上测试了不同的案例场景。多智能体系统在JAVA智能体开发框架（JADE）中实现，每个智能体内的主要算法采用MATLAB。将结果与之前提出的集中式和分散式方法的结果进行比较，证明所提出的方法比集中式方法具有更好的计算效率，并且比分散式方法具有更好的结果。因此，所提出的算法提供了整体改进的解决方案。