Particle swarm optimization (PSO) has attracted interest in recent years to solve practical economic dispatch problems. However, as with traditional optimization methods, conventional PSO may converge to local optima and incurs significant computational overheads to produce a practical solution, and has a performance which strongly depends on the choice of internal parameters. To address these drawbacks, this paper applies a non-inertial weight based PSO method to solve nonconvex economic dispatch problems where valve-point effects, prohibited operating zones, ramp rates and transmission losses are taken into account. The method used adopts a simplified velocity update rule for particles by only employing the social and the cognitive parts of conventional PSO. Simulation results from four different systems, with 3, 6, 13 and 15 generating units, demonstrate its superiority over other optimization methods presented in the literature. Moreover, the results show that our method is able to speed up the convergence with savings of up to 70.11% computation time and also to improve the quality of the solutions with savings in fuel cost for the 13-unit test example of between 19.79 $/h up to 75.05$/h.