城市污水处理建模理论及关键技术研究

Soft-sensors are considered a good alternative to predict hard-to-measure variables in the wastewater process. The bottlenecks limiting its widespread application, however, mainly lie on the fact that the wastewater process exhibits multi-rate、multiple data sources and multi-output features. The objective of this project is to predict amounts of hard-to-measure variables in the wastewater process. Firstly, in order to make sure the accuracy and robustness of soft-sensors prediction, Gaussian processes models are performed to deal with multi-rate、multiple data sources and multi-output issues; Secondly, for the purpose of exploring the multi-step prediction of multi-rate、multiple data sources and multi-output soft-sensors, Recursive、Direct and Recursive-Direct multi-step prediction strategies are implemented to improve Gaussian processes models. Finally, on this basis, uncertainty intervals obtained from Gaussian processes models are able to establish three-level for fault diagnosis and prognosis. These methods will be better guidance for monitoring system operation, it will be also make a greater contribution for system health management in a quantitative manner theoretically and technically.

软测量技术作为解决污水处理过程难以实时测量变量的有效途径正受到广泛的关注。然而，多采样率、多数据源以及大量难以测量变量的同时存在成为了制约该项技术推广应用的一个主要瓶颈。项目拟以城市污水处理为研究对象，首先依据高斯过程模型的基本特性，开展多采样率、多数据源、多输出软测量建模方法的研究，保证软测量在污水处理过程复杂情况下的准确而稳定的预测；其次，研究基于高斯过程模型的递归、直接和递归-直接方式的多步预测，同时探索污水处理过程多采样率、多数据源、多输出软测量建模方法的多步预测；最后，在上述研究的基础上，结合高斯过程模型的不确定信息描述能力，对三级故障预警机制进行深入分析从而提高故障诊断和预测的完整性。该项目旨在为有效的监控污水处理系统运行和定量评价系统健康状态奠定理论和技术基础。

软测量技术作为解决污水处理过程难以实时测量变量的有效途径正受到广泛的关注。然而，多采样率、多数据源以及大量难以测量变量的同时存在成为了制约该项技术推广应用的一个主要瓶颈。项目以城市污水处理为研究对象，首先依据高斯过程模型的基本特性，实现了多采样率、多数据源、多输出软测量建模方法，保证软测量在污水处理过程复杂情况下的准确而稳定的预测；其次，实现了基于高斯过程模型的递归、直接和递归-直接方式的多步预测，同时探索污水处理过程多采样率、多数据源、多输出软测量建模方法的多步预测，并充分应用在系统的故障诊断中；最后，在上述研究的基础上，结合高斯过程模型的不确定信息描述能力，对三级故障预警机制进行深入分析从而提高故障诊断和预测的完整性。该项目为有效的监控污水处理系统运行和定量评价系统健康状态奠定了理论和技术基础。