基于混合智能模型的城市污水处理优化控制研究

Urban sewage treatment is an effective method for reusing water resource, and it has great significance for alleviating the global water shortage. The main problems of the current wastewater treatment process (WWTP) lie in the high energy consumption and effluent equality out of standard due to the unsatisfactory control performance. The active sludge treatment process is a complex industrial system which includes many physical processes and intricate microbial biochemistry phenomenon, and thus it has obvious nonlinear characteristics of time-varying, time-delay, coupling and uncertain disturbances. Moreover, the precise mathematical model of WWTP is difficult to establish. Therefore, the implementation of optimization control in a WWTP is a challenging task. Revolving around the above problems, this project will study and establish a hybrid intelligent model based on the data-driven model, mechanism model and knowledge model, which can provide a basis of control and optimization of the WWTP; propose a multi-objective optimization control method to solve the multi-objective optimization problem of water quality, energy consumption and other important performance indexes; work on the dynamic neural network learning algorithm for a class of complex slow time-varying nonlinear system represented by the characteristics of WWTP, and the aim is to improve the network learning performance; design the intelligent controller based on neural network for WWTP, and realize the tracking control task of optimal set-point values. Based on simulation studies, apply the results into the real wastewater plants to improve the optimal performance of WWTP (such as energy consumption and effluent quality).

城市污水处理是水资源利用的有效手段，对于缓解全球水资源短缺具有重要意义。目前污水处理过程存在的主要问题是过程控制水平不高导致的出水水质超标以及能耗成本过高等。基于活性污泥法的污水处理过程包含复杂的物理及微生物生化反应现象，具有明显的时变、时滞、变量耦合以及干扰不确定因素众多等非线性特征，且模型建立较为困难，实施优化控制是一项具有挑战性的工作。项目将围绕以上问题，研究和建立基于数据驱动-机理模型-知识模型在内的混合智能模型，为污水处理过程控制及优化提供模型基础；研究多目标优化控制方法，解决污水处理过程水质、能耗等重要性能指标的多目标优化问题；研究污水处理过程所代表的一类复杂慢时变非线性系统的动态神经网络学习问题，提高网络学习性能；研究基于神经网络的污水处理过程智能控制方法，解决优化设定值的跟踪控制问题。结合仿真实验结果，将研究成果应用于实际污水处理过程，实现污水处理能耗等性能指标优化与提升。

城市污水处理是缓解全球水资源短缺的重要手段之一。基于活性污泥法的污水处理过程包含复杂的物理及微生物生化反应现象，实施优化控制是一项具有挑战性的工作。本项目开展污水处理过程多目标优化方法及智能控制方法研究，取得的主要成果包括：针对一类复杂慢时变系统的动态神经网络学习问题，提出一种其于SAPSO的神经网络自组织结构和参数学习算法，提高了网络学习性能；提出并获得一种基于机理-数据-知识的混合模型的建模方法，实现出水氨氮浓度预测，为污水处理过程控制及优化提供建模方面的理论支撑；开展污水处理优化控制研究，获得一种基于知识与数据信息决策的污水处理优化控制方法，研究成果可为污水处理过程混合智能建模及应用、污水处理过程智能优化控制提供相关支撑；研究多目标优化控制方法，提出一种多策略自适应多目标粒子群优化算法；获得基于神经网络的多变量污水处理过程控制方法，解决污水处理过程水质和能耗性能指标的多目标优化问题，丰富和发展了智能优化控制理论。该项目执行过程中申请相关国家发明专利3项，在本领域重要期刊和会议上发表相关学术论文13篇，其中SCI期刊和EI期刊论文8篇，培养毕业硕士生1名。