中文摘要：

针对标准差分进化算法自身存在的贪婪而易于早熟的问题,提出一种深度捕食二次梯度加速差分进化算法.混合算法首先把整个搜索空间作为整体进行广泛搜索,并预设一个梯度加速触发参数,对种群中较优的个体采用基于梯度加速的局部搜索,使算法能够快速收敛到全局最优值.同时为了保持种群的多样性,设计了一种差分变异算子.通过与已有的改进算法仿真对比可知:该算法能够有效地跳出局部极值,防止算法早熟,且收敛速度快.结合相关文献对深度捕食二次梯度加速差分进化算法的工程应用进行了仿真研究,仿真结果验证了该方法的可行性和有效性.

英文摘要：

T he standard differential evolution algorithm is too greedy to easy premature convergence . To avoid this drawback ,a kind of deep predation and secondary gradient acceleration differential evo-lution algorithm was proposed .The entire search space was first put as a whole wide search ,and a gradient acceleration trigger parameters were preset ,the local search based on gradient acceleration was taken for better individuals in population ,so the algorithm can quickly converge to the global op-timal value .At the same time ,in order to maintain the population diversity ,a new differential muta-tion operator was designed .By comparison with existing improvement algorithm ,the algorithm here can effectively escape from local minima ,and prevent premature convergence .And the convergence speed is fast .Finally ,in combined with the relevant literature ,the simulation research was conducted on the engineering application of the algorithm ,and the results verified the feasibility and the effec-tiveness of the method .