中文摘要：

基于梯度信息的线性搜索法具有快速的收敛性,但易陷入局部最优。当优化目标不可解析时,基于梯度信息的算法便不易应用。多目标进化算法以其优秀的全局特性广泛地应用于多目标优化问题,但其算法比较耗时,收敛速度慢。对此,本文提出一种基于进化梯度搜索的多目标混合算法。首先,结合单目标优化中的爬山算法与进化梯度搜索法,得到一种多目标局部搜索算法。其次,在算法前期采用适应度概率策略选择个体进行局部搜索。最后,在非支配集个体数达到种群个体数后,应用多目标进化算法保证其分布性。通过ZDT系列测试函数验证并与NSGA-II及EGS-NSGA-II混合算法比较,结果显示本算法具有更好的全局性及收敛快速性。

英文摘要：

Linear search algorithm based on gradient information has fast convergence, but it is easy to fall into local optimum. When the optimization goal is not differentiable, the algorithm based on gradient information will be no longer easy to apply. Multi-objective evolutionary algorithms are widely applied to multi-objective optimization problems for excellent global features, but this algorithm is time-consuming with slow convergence speed. For solving this problem, this paper proposes a hybrid multi-objective evolution algo- rithm based on gradient search. First of all, we promote a multi-objective local search algorithm compared with HCS and EGS algorithm which perform well in single objective problem. Secondly, at the early stage of algorithm, we selects individual species using fitness probability choosing strategy. Finally, when the non-dominated set number reaching the population individual number, multi-objective evolutionary algorithms return. According testing the algorithm through the ZDT series test function, and comparing with the NSGA-II and EGS-NSGA-II algorithms, results show that this algorithm has better global ability and fast convergence speed.