In order to accurately estimate the unknown parameters for chaotic systems, the artificial bee colony optimization algorithm was improved, and an adaptive artificial bee colony optimization algorithm was proposed. The proposed method formatted the problem of parameter estimation for chaotic systems to a multidimensional variable optimization problem, and used the artificial bee colony optimization algorithm to search the unknown parameters in a guided random manner. During the search process, the method adaptively adjusted the step size and the solution trial limits based on the optimum degree of the population and the quality of the solutions. The numerical simulation on the classic Lorenz chaotic system demonstrates that the proposed method is robust and can obtain accurate estimation for chaotic systems without noise or with intensive noise.