This paper presents a chaotic firefly algorithm (CFA) to solve economic load dispatch (ELD) problems. In the proposed algorithm, chaotic variables generated using tent map are used in the control parameters of FA algorithm, instead of rand number generator to improve the searching capacity of FA and to find the better optimum solution. The effectiveness of the proposed algorithm has been tested with 13, 15, 40, 10 units with multiple fuel options and IEEE-30 bus system with different constraints, such as power balance, prohibited operating zones, ramp-rate limits, bus voltage constraints and line flow limits. The results obtained by the proposed algorithm are compared with the results of FA and other soft computing techniques reported in literature. The results obtained reveal that the optimal results obtained by the proposed algorithm are better than FA and other soft computing techniques reported in the literature. Furthermore, the proposed algorithm has good convergence characteristic in all the test cases considered.