Multi-objective evolutionary algorithm for wastewater treatment process optimization control

DOI编号  10.7641/CTA.2019.80408
2020,37(1):169-175

 作者 单位 E-mail 杨壮 北京工业大学信息学部 892836748@qq.com 杨翠丽 北京工业大学信息学部 顾锞 北京工业大学信息学部 乔俊飞 北京工业大学信息学部

污水处理过程中，能耗与出水水质是两个相互矛盾的评价指标。为了找出这两个目标的最优解，本文在基于分解的多目标进化算法（multi-objective evolutionary algorithm based on decomposition, MOEA/D）的基础上进行改进，期望用更少的进化次数得到分布均匀的近似帕累托前沿。针对MOEA/D 算法每一次产生的新解，本文中改进的算法从所有子问题中找到最合适新解的子问题，并在其邻域范围内进行种群的更替，在原本子问题的基础上进行二次寻优，提高子代利用率，进而用更少的迭代次数找到优化问题中的近似帕累前沿。实验证明，该算法明显减少了找到帕累托前沿的步数，使得MOEA/D 算法的性能明显提升，在污水处理过程优化问题中达到了优化目标的作用。

In the process of sewage treatment, energy consumption and effluent quality are a pair of contradictory indicators. In order to find the optimal solution of these two objectives, this paper improves multi-objective evolutionary algorithm based on decomposition(MOEA/D) that expects even distribution with fewer evolution times for an approximate Pareto front. This algorithm aims at the new solution by using the MOEA/D algorithm each time, finds the most suitable sub-problem of the new solution from all the sub-problems, and carries out replacement of the population within its neighborhood, based on the original sub-problem. Secondary search improves the utilization of the child generation and finds the approximate Pareto front in the optimization problem with fewer iterations. Experiments show that the algorithm significantly reduces the number of steps to find the Pareto front, which results in a significant increase in the performance of the MOEA/D algorithm and achieves the goal of optimization in the wastewater treatment process.