A Novel Differential Evolution Algorithm with Multi-Strategy Differential Mutation Mechanism

By combining the application mode of several effective differential mutation strategies, a novel DE algorithm (ANMSDE) is proposed. First, with population evolution, population individuals are divided into three groups by the normal distribution of population fitness, and the three groups adopt different mutation operators. Second, a random method and a simple roulette wheel method based on affinity matrix are alternatively used in selecting the individuals involved in mutation operation. Extensive experiments have been conducted to compare ANMSDE with several state-of-the-art DE variants proposed in pertinent literatures on nine well-known benchmark functions. The simulation results show that ANMSDE can efficiently balance the exploration and exploitation capabilities of algorithm and promises competitive performance not only in the convergence speed but also in the quality of solution.