Optimal experimental design for decision-making with dynamic water quality models

利用动态水质模型进行决策的优化实验设计

• 批准号: 327591-2011
• 负责人: Vanrolleghem, Peter
• 金额: $ 2.48万
• 依托单位: Université Laval
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2015
• 资助国家: 加拿大
• 起止时间: 2015-01-01 至 2016-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=572064
• 关键词: Optimal; experimental; design; decision; making

Dynamic models increasingly become key ingredients of decision-making in water infrastructure planning and receiving water management. It is generally accepted that within the three preparatory steps of a modelling exercise, (i) definition of the objective, (ii) structuring of prior knowledge and (iii) analysis of available data, most modelling issues are raised due to limitations in quantity and quality of the available data. It is therefore surprising to have to conclude from the current state-of-practice that relatively little research effort is conducted to provide methodological support to collect good quality data within the constraints imposed by available resources (time, funding, people, equipment, etc.). This discovery grant proposal therefore aims at developing such methodological support in the form of what is termed "model-based optimal experimental design" (mbOED). This will be developed for two types of dynamic systems where water quality aspects are central: biological wastewater treatment and river water quality. Decision-making always occurs under uncertainty and the sources of uncertainty can be attributed to a number of sources, e.g. the model structure, the parameter values used and the input data. mbOED can assist in a methodological reduction of these uncertainties. The discovery grant research proposal thus has three research lines: (i) OED for Model Discrimination (mbOED/MD), (ii) OED for Parameter Estimation (mbOED/PE) and (iii) OED for overall reduction of uncertainty in Decision Variables (mbOED/DV). Thus, the proposed research programme should deliver a methodological framework by which experiments can be designed that allow (i) deciding on the best model structure, (ii) obtaining low uncertainty parameter estimates and (iii) predicting decision variables with least possible uncertainty. The work will be supported experimentally by CFI grants for river measuring stations and a wastewater treatment pilot plant.

动态模型日益成为水基础设施规划和受水管理决策的关键因素。人们普遍认为，在建模工作的三个准备步骤中，（一）确定目标，（二）构建现有知识，（三）分析现有数据，大多数建模问题是由于现有数据的数量和质量有限而提出的。因此，令人惊讶的是，从目前的实践状况得出的结论是，在现有资源（时间、资金、人员、设备等）的限制范围内，为收集高质量数据提供方法支持的研究工作相对较少。 因此，这项发现补助金提案旨在以所谓的“基于模型的最佳实验设计”（mbOED）的形式开发这种方法学支持。这将被开发为两种类型的动态系统，其中水质方面是中心：生物废水处理和河流水质。决策总是在不确定性的情况下进行，不确定性的来源可以归因于许多来源，例如模型结构，使用的参数值和输入数据。mbOED可以帮助在方法上减少这些不确定性。因此，发现补助金研究计划有三个研究方向：（i）模型识别OED（mbOED/MD），（ii）参数估计OED（mbOED/PE）和（iii）决策变量不确定性总体减少OED（mbOED/DV）。 因此，拟议的研究方案应提供一个方法框架，通过该框架可以设计实验，以便（i）决定最佳模型结构，（ii）获得低不确定性参数估计值，（iii）以最小可能的不确定性预测决策变量。这项工作将得到CFI赠款的实验性支持，用于河流测量站和废水处理试验厂。