Abstract: With increasing fuel prices and restructuring in the power industry, nonconvex economic dispatch (NED) is expected to become crucial because of nonsmooth cost functions. This paper presents an intelligent particle swarm optimization for economic dispatch with valve-point effect. A new index, another particle best (Pbestap), is incorporated into the particle swarm optimization to further improve social behavior. Moreover, a novel diversity-based judgment mechanism for evaluating Pbestap behavior is proposed for maintaining population diversity, which facilitates identification of the near-global region. The direct search algorithm is used to fine-tune and determine the eventual global optimal solution at low computational expense. Numerical experiments demonstrate that the proposed approach offers higher quality solutions than do several existing techniques.