Parameter identification of chaotic systems using improved differential evolution algorithm

In this paper, an improved differential evolution algorithm, named the Taguchi-sliding-based differential evolution algorithm (TSBDEA), is proposed to solve the problem of parameter identification for Chen, Lü and Rossler chaotic systems. The TSBDEA, a powerful global numerical optimization method, combines the differential evolution algorithm (DEA) with the Taguchi-sliding-level method (TSLM). The TSLM is used as the crossover operation of the DEA. Then, the systematic reasoning ability of the TSLM is provided to select the better offspring to achieve the crossover, and consequently enhance the DEA. Therefore, the TSBDEA can be more robust, statistically sound, and quickly convergent. Three illustrative examples of parameter identification for Chen, Lü and Rossler chaotic systems are given to demonstrate the applicability of the proposed TSBDEA, and the computational experimental results show that the proposed TSBDEA not only can find optimal or close-to-optimal solutions but also can obtain both better and more robust results than the DEA.