An improved competitive swarm optimizer (ICSO) is proposed for large scale power system economic dispatch. The CSO is fundamentally inspired by the particle swarm optimization (PSO). It is with advantages of simple principle, easy implementation, and strong search ability. Based on the analysis of the disadvantages of the CSO, an improved loser updating strategy and an adaptive adjusting strategy of control parameter are proposed to balance the local and the global researching abilities. Four cases are employed to comparatively demonstrate the ICSO’s viability and effectiveness from different perspectives of economy, convergence, and robustness. Simulation results indicate that the ICSO is with better performance and can be used as a reliable method for large-scale ED.