CAJ | 학술논문

标准差分进化( DE)算法在高维多峰等复杂函数优化时易出现早熟现象，并且算法后期收敛速度较慢。为此，研究2种标准差分进化算法的变异策略( DE/rand/1和DE/best/1)，并将其进行串行组合，提出一种多变异策略的差分进化算法( MDE)。在4个Benchmark函数上的测试结果表明，在多变异策略下，通过对MDE算法控制参数的调整能有效拓展和平衡改进后算法的全局与局部搜索能力，其所得最优解的精度、算法的收敛速度都较标准差分进化算法有明显优势，能较好地解决电力负载分配问题。
표준차분진화( DE)산법재고유다봉등복잡함수우화시역출현조숙현상，병차산법후기수렴속도교만。위차，연구2충표준차분진화산법적변이책략( DE/rand/1화DE/best/1)，병장기진행천행조합，제출일충다변이책략적차분진화산법( MDE)。재4개Benchmark함수상적측시결과표명，재다변이책략하，통과대MDE산법공제삼수적조정능유효탁전화평형개진후산법적전국여국부수색능력，기소득최우해적정도、산법적수렴속도도교표준차분진화산법유명현우세，능교호지해결전력부재분배문제。
In order to overcome the shortcomings of the standard Differential Evolution ( DE ) algorithm in the optimization of complex functions like dimension multi-modal functions,such as the problem of premature and slow later convergence,this paper proposes a DE algorithm based on the Mutation strategy( MDE) through serial combination of DE/rand/1 and DE/best/1. It makes an in-depth study of this algorithms,and finally the algorithm is tested on the four Benchmark functions. Result shows that through the modulation of the control parameters of MDE can effectively expands and balances the global and local search capabilities of the improved algorithm,and its resultant optimal accuracy,and convergence speed are better than standard DE algorithm. It can be well applied in electric power load distribution.