A hybrid modeling, monitoring and control approach for wastewater treatment plants

目前污水处理厂的设计使用相当详细的第一性原理模型，这些模型能够捕获过程的主要动态，但不一定非常准确地预测变量。为新的厂址校准这些模型仍然是一个巨大的挑战。缺乏精确的模型意味着不容易实现对这些电厂的可靠监测和最佳运行。工厂运行数据正变得越来越容易获得，并已努力使用完全数据驱动的方法来模拟工厂行为。然而，数据驱动方法中缺乏第一性原理信息，使得数据驱动模型容易过度拟合，而不适合外推。出于这些考虑，本提案将开发混合模式，使数据驱动的方法与第一原则模式无缝协同，以利用第一原则模式中存在的信息。这个项目是对一项成功的参与资助的后续行动，在那里，拟议的想法的概念证明得到了证明。本提案将以实际案例研究为基础，并推广到大大扩大监测和反馈控制的应用。对行业合作伙伴海德曼蒂斯的直接好处是极大地扩大了客户群。鉴于正在开发的工具的广泛性质，通过将该方法适应从石化到制药到钢铁制造等广泛的生产设施，将为加拿大经济带来额外的好处。

Wastewater treatment plants are currently designed using fairly detailed first principles models that are able to capture the dominant dynamics of the process, but not necessarily predict the variables very accurately. The calibration of these models for a new plant site remains a significant challenge. The lack of precise models means that reliable monitoring and optimum operation of these plants is not readily achieved. Plant operation data is becoming increasingly more available, and efforts have been made to use completely data driven approaches to model the plant behaviour. The lack of first principles information in the data driven approaches, however, makes the data driven models susceptible to overfitting and not amenable to extrapolation. Motivated by these considerations, the present proposal will develop hybrid models that seamlessly synergies data driven approaches with first principles models in a way to leverage the information present in the first principles model. This project is a followup to a successful engage grant where a proof of concept for the proposed idea was demonstrated. The present proposal will build on that with real case studies, and generalize to significantly expanded applications of monitoring and feedback control. The direct benefit to the industrial partner Hydromantis is an immensely broadened clientele. Given the broad nature of the tool being developed, additional benefit to Canadian economy will be realized through adaptation of the approach to a broad ranges of production facilities ranging from petrochemicals to pharmaceuticals to steel manufacturing.