In this study, we propose a multi-agent glowworm swarm optimization (MAGSO) algorithm for the economic load dispatch (ELD) problem of a large-scale hydropower station. The MAGSO integrates the idea of the evolution of glowworm swarm optimization (GSO) and the interindividual cooperation of multi-agent system (MAS). In this structure, each glowworm of the GSO is treated as an agent cooperating in the MAS before entering the evolutionary process. The proposed mechanisms enable the algorithm to escape from local extremum, capture the optimal solution and expand the diversity of solution space. The test results of four benchmark functions verify the advantages of the algorithm in global search, convergence speed and avoiding prematurity. In comparison of two case studies with the real genetic algorithm (RGA), the evolution programming (EP), the particle swarm optimization (PSO), the ant colony optimization (ACO), and the standard glowworm swarm optimization (GSO), the MAGSO provides significantly robust solutions of ELD problem of a large-scale hydropower station.