Abstract
To address the issues of fault localisation and recovery in medium-voltage direct current power systems for ships, this paper proposes a hybrid reconstruction algorithm that combines the entropy weighting method with grey relational analysis, and embeds them within a framework based on chaotic particle swarm optimisation. During initialisation, the algorithm employs chaotic mapping to enhance the population diversity. The entropy weighting method is used for adaptive weighting of the objective functions, while grey relational analysis is integrated to evaluate the particle fitness and determine optimal reconstruction paths, thereby accomplishing fault diagnosis and system restoration. Simulations and case studies demonstrate that the proposed method has advantages such as rapid convergence and strong ability to escape local optima. Compared with traditional methods, it achieves a 41% increase in the reconstruction success rate and a 4.7% improvement in the system restoration capability, effectively enhancing post-fault power supply capacity and overall reliability.