本项目针对动态环境多目标优化问题的几个关键问题进行研究，包括动态EMO中忽视动态环境对进化的作用的问题，研究基于动态环境进化模型的动态多目标进化算法，针对动态EMO中不能有指导性地增加种群多样性的问题，运用神经网络等机器学习的方法，在已建立的动态环境模型的基础上，有指导性地增加种群多样性，帮助算法能更好地适应新的动态环境；针对动态EMO中既存在周期性变化，又在第一个变化周期的环境下种群难于收敛的问题，研究新的混合策略，将记忆与预测两种策略有机的结合；针对动态EMO中缺乏合理的子种群划分和进化模型的问题，研究基于动态环境模型的多种群策略的动态多目标进化算法，提出更适合于优化问题的子种群划分和进化策略。

This project studies for severalkey issues of dynamic multi-objective optimization problems, including that dynamic EMO is not instructive to increase the diversity of population, we use the method of machine learning such as neural networks etc., to increase the diversity of population by guided fashion on the basis of the prior knowledge, and then help the algorithm can better adapt to the new dynamic environment. For the existence of cyclical changes in dynamic EMO, and the dynamic EMO is difficult to convergence in the first period change, we research new hybrid strategy, which combining the prediction and memory. For the lack of reasonable classification and evolutionary model about sub-populations, we research the dynamic MOEA, and propose more suitable strategies of dividing sub-populations and evolution for optimization problems. For the issue of that current dynamic MOEA did not consider the role of dynamic environment for the evolutionary population, we research the dynamic MOEA based on dynamic environment evolutionary model.