王鼎, 赵慧玲, 李鑫. 基于多目标粒子群优化的污水处理系统自适应评判控制[J]. 工程科学学报, 2024, 46(5): 908-917. DOI: 10.13374/j.issn2095-9389.2023.04.15.001
引用本文: 王鼎, 赵慧玲, 李鑫. 基于多目标粒子群优化的污水处理系统自适应评判控制[J]. 工程科学学报, 2024, 46(5): 908-917. DOI: 10.13374/j.issn2095-9389.2023.04.15.001
WANG Ding, ZHAO Huiling, LI Xin. Adaptive critic control for wastewater treatment systems based on multiobjective particle swarm optimization[J]. Chinese Journal of Engineering, 2024, 46(5): 908-917. DOI: 10.13374/j.issn2095-9389.2023.04.15.001
Citation: WANG Ding, ZHAO Huiling, LI Xin. Adaptive critic control for wastewater treatment systems based on multiobjective particle swarm optimization[J]. Chinese Journal of Engineering, 2024, 46(5): 908-917. DOI: 10.13374/j.issn2095-9389.2023.04.15.001

基于多目标粒子群优化的污水处理系统自适应评判控制

Adaptive critic control for wastewater treatment systems based on multiobjective particle swarm optimization

  • 摘要: 考虑到城市污水处理系统存在保证出水水质达标和降低能耗的需要, 将其运行过程视为一个多目标优化控制问题. 针对此问题, 提出一种基于多目标粒子群优化(Multi-objective particle swarm optimization, MOPSO)算法的污水处理系统自适应评判控制方案, 该方案分为上层优化和底层跟踪控制两部分. 首先, 污水处理过程存在非线性、多变量、大时变等特点, 结合数据驱动思想对入水及出水组分数据进行分析, 构建关于出水水质和运行能耗的多目标优化模型. 采用径向基函数(Radial basis function, RBF)神经网络进行建模, 并与反向传播(Back propagation, BP)神经网络进行了对比. 然后, 结合MOPSO算法强大的优化能力, 采用MOPSO算法对优化目标进行求解, 并设计一个决策方式从最优解集中选出偏好解, 作为溶解氧与硝态氮浓度的最优设定值. 接下来, 底层跟踪控制部分采用基于自适应动态规划的辅助控制器对比例–积分–微分算法的控制策略进行补充, 弥补了传统控制算法自适应能力差的不足. 此外, 比例–积分–微分算法也为自适应动态规划算法提供了初始的稳定控制策略,克服了学习算法前期控制效果差的缺陷,保证了污水处理过程的安全性和可靠性. 最终, 该控制器成功实现了对最优设定值的跟踪控制. 将所提算法在污水处理基准仿真平台上进行验证, 结果表明所提算法能有效地提高污水处理过程的运行性能, 不仅能保证出水水质达标, 同时能有效地降低污水处理过程产生的能耗.

     

    Abstract: Given the need to ensure that effluent quality meets the standards and reduces energy consumption in urban wastewater treatment systems, the operation process is considered a multiobjective optimization control problem. An adaptive critic control scheme is developed based on multiobjective particle swarm optimization. This scheme is divided into two parts: upper optimization and bottom tracking control. First, considering the characteristics of nonlinear, multivariable, and large time variation in a wastewater treatment system, the mechanism model is difficult to establish accurately. To preserve quality and reduce consumption, an accurate operation index model of the wastewater treatment process must be designed. The data of the inlet and outlet components are analyzed using a data-driven framework. A multiobjective optimization model reflecting effluent quality and energy consumption is constructed. A radial basis function neural network is used for modeling and compared with a back-propagation neural network. Then, combined with powerful optimization capabilities, the multiobjective particle swarm optimization algorithm is used to solve the multiobjective optimization problem. Combining the practical importance of the two indicators of energy consumption and water quality, a decision method is designed to select the preferred solutions from the optimal solution set. The preferred solutions can be defined as the optimal set concentrations of dissolved oxygen and nitrate nitrogen. Next, the bottom tracking control part adopts an auxiliary controller based on adaptive dynamic planning to supplement the control strategy of a proportional–integral–derivative algorithm, compensating for the shortcomings of the poor adaptive ability of the traditional control algorithm. In addition, this proportional–integral–differential algorithm provides an initial stable control strategy for the adaptive dynamic programming algorithm, overcoming the poor control effect of the learning algorithm in the early stage and ensuring the safety and reliability of the wastewater treatment process. Ultimately, the controller successfully achieves the tracking control of the optimal setting value. To verify the optimization effect and control performance of the proposed scheme, we use benchmark simulation model no. 1 to complete the simulation. Using the indicators of water quality and energy consumption, we also compare the proposed scheme with other multiobjective optimization schemes. The results show that the proposed algorithm effectively improves the operational performance of the wastewater treatment process. It not only ensures that the effluent water quality meets the standards but also effectively reduces the energy consumption of wastewater treatment.

     

/

返回文章
返回