基于深度强化学习的造纸废水处理过程多目标优化

Multi-objective Optimization of Papermaking Wastewater Treatment Process Based on Deep Reinforcement Learning

本课题提出了一种基于多智能体深度强化学习的动态优化方法,以期实现造纸废水处理过程的运行成本和能耗的协同优化。实验采用了基准仿真1号模型(BSM1)模拟造纸废水处理过程的生化反应和沉淀过程,并利用模型数据对强化学习智能体进行训练,最后用实际的造纸废水数据对搭建的模型系统进行验证。结果表明,基于多智能体深度强化学习的废水处理系统能够保障排水质量,实现成本与能耗的多目标优化控制,其性能表现优于传统方法。

In this study,a dynamic optimization method based on multi-intelligent deep reinforcement learning was proposed to realize the col‑laborative optimization,of the operation cost and energy consumption of papermaking wastewater treatment process.The BSM1 benchmark sim‑ulation model was used to simulate the biochemical reaction and precipitation process of papermaking wastewater treatment process,the rein‑forcement learning intelligences were trained,and the actual paper making wastewater data was used to verify the model system.The results showed that the wastewater treatment system based on multi-intelligent deep reinforcement learning system could guarantee the effluent quality,realized the multi-objective optimization control of cost and energy consumption,and its performance was better than the traditional methods.