Abstract: Multi-strategy differential evolutionary algorithm oriented by excellent individual is proposed. Firstly, According to the fitness value, the population is divided to three equal sub-population, different mutation strategies and control parameters are used for different populations. Secondly, a new mutation strategy is proposed for the population with poor fitness value, by introducing the learning factor and balance factor to balance the improvement of convergence speed, accuracy and falling into the local optimal state. An adaptive mechanism is adopted to the control parameters of the individual to reduce the probability of the population stagnating. In addition, a new population will be formed among the three populations after each iteration, thus the information interaction between different populations is realized. Finally, the performance of the proposed algorithm is tested by 19 standard test functions and compared with some mainstream differential evolutionary algorithms. The experimental results show that the convergence speed and accuracy of most functions are improved obviously.

Key words: differential evolution（DE）, multi-strategy, excellent individual, learning factor, balance factor

摘要： 提出一种以优秀个体为导向的多策略差分进化算法。根据适应度值将种群等分为三个子种群，针对不同的种群使用不同的变异策略和控制参数。针对适应度值较差的种群提出了一种新的变异策略，通过引入学习因子和平衡因子，对提高收敛速度、精度和易陷入局部最优状态进行平衡，并对其中个体的控制参数采取自适应的机制，降低种群陷入停滞状态的概率。除此，在每次迭代完成之后，三个种群会重新组成一个新的种群，从而实现了不同种群之间信息的交互。用19个标准测试函数对所提出算法的性能进行了测试，并将其与一些主流差分算法进行比较。实验结果表明，所提出的算法在大部分函数的收敛速度以及精度上有明显的提升。

关键词: 差分进化（DE）, 多策略, 优秀个体, 学习因子, 平衡因子