Towards digital twin based control of water resource recovery facilities - Methods supporting the use of adaptive hybrid digital twins

水资源回收设施（WRRF）受到越来越多的废水负荷和更严格的目标，对环境保护和质量的回收资源。过程控制是一种经过验证的方法，可以在不增加设备成本的情况下提高过程性能。然而，系统的复杂性和更严格的要求要求过程控制逐步升级，并超越无所不在的开/关和PID控制器。在接下来的5年里，这项发现资助的研究计划将利用以前对数据质量评估工具和WRRF模型的研究，以及2014年在拉瓦尔大学安装的独特试点WRRF，培训14名HQP，开发可以集成到基于模型的过程控制中的数字孪生模型，参与水务部门的数字化。该研究计划的长期目标是开发和应用新的数据驱动方法，利用机器学习和人工智能，以WRRF操作和控制。结合了联合收割机机械见解和数据驱动关系的混合数字孪生将成为这种新过程控制方法的核心要素。短期目标旨在处理数据管道中的一些缺陷，高质量的数据必须通过这些缺陷流动，以不断更新数字孪生模型，使其保持模仿WRRF，并为控制提供例如流入废水的信息。提高工艺性能和质量的处理废水和回收资源将需要发展：1）方法的实时检测设备和工艺故障（故障检测）; 2）表征主要WRRF的干扰3）在线过程模型修正算法（自适应混合数字孪生）; 4）利用数字孪生的预测并配备故障情况下的回退策略的高效控制系统（基于数字孪生的故障适应控制）。拟议的研究计划并不是专注于构建数字孪生模型本身，而是确保数字孪生模型是最新的，提供适当的数据，并可用于基于模型的控制。这些发展将在试验性WRRF（与平行列车直接比较方法）和两个全尺寸系统上进行测试，HQP可以在这些系统上暴露于实际实施问题。该项目将培养8名研究生（4名博士，4名硕士），5名本科生和一名博士后。由于参与研究项目的不同方面，这些HQP将获得广泛的技能，接触国家和国际研究，并通过行业标准试点设施和两个全面WRRF的工作进行实践。该研究计划包括公平，多样性和包容性（EDI）在HQP招聘/培训，团队组成和研究规划/管理/影响的关键要素。

Water Resource Recovery Facilities (WRRF) are subject to increasing wastewater loads and stricter objectives with respect to environmental protection and quality of recovered resources. Process control is a proven methodology to improve process performance without costly expansion of the facilities. However, the complexity of the systems and the stricter requirements ask for process control to step-up and go beyond the omnipresent on/off and PID controllers. Over the next 5 years the research program of this Discovery grant will capitalize on previous research on data quality assessment tools and WRRF models, and the unique pilot WRRF installed at Université Laval in 2014, to train 14 HQP, to develop digital twins that can be integrated in model-based process control, to take part in the digitalization of the water sector. The long-term objectives of this research program are to develop and apply new data-driven methods that leverage machine learning and AI, to WRRF operation and control. Hybrid digital twins that combine mechanistic insights and data-driven relations will be the central ingredient of this new process control approach. The short-term objectives aim at dealing with some of the deficiencies in the data pipeline through which high quality data must flow to continuously update the digital twin to keep it mimicing the WRRF and to feed the control with, for instance, information on the influent wastewater. Improving process performance and quality of treated wastewater and recovered resources will require developing: 1)Methods for real-time detection of equipment and process faults (Fault detection); 2)Characterization of the main WRRF's disturbances (Real-time influent generator); 3)Algorithms for on-line process model updating (Adaptive hybrid digital twin); 4)Efficient control systems taking advantage of predictions by the digital twin and equipped with fallback strategies in case of faults (Digital twin-based, fault-accommodating control). The proposed research program is not focusing on building the digital twin itself, but rather on methods ensuring digital twins are up-to-date, fed with proper data and can be used for model-based control. The developments will be tested on the pilot WRRF (with parallel trains for direct comparison of approaches) and two full-scale systems on which HQP can be exposed to practical implementation issues. The project will train 8 graduate students (4 PhD, 4 MSc), 5 undergraduates and a postdoctoral fellow. These HQP will acquire a wide range of skills thanks to the involvement in the different aspects of the research project, the exposure to national and international research and to practice through the work on an industry standard pilot facility and two full-scale WRRFs. The research programme includes key elements of equity, diversity, and inclusion (EDI) in HQP recruitment/training, team composition and research programming/management/impact.