Development of optimization software to improve the efficiencies of desalination and wastewater treatment

开发优化软件以提高海水淡化和废水处理效率

• 批准号: 544088-2019
• 负责人: Dong, Zuomin
• 金额: $ 1.82万
• 依托单位: University of Victoria
• 依托单位国家: 加拿大
• 项目类别: Engage Grants Program
• 财政年份: 2019
• 资助国家: 加拿大
• 起止时间: 2019-01-01 至 2020-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=680510
• 关键词: Development; optimization; software; improve; efficiencies

Pani Energy is developing a novel membrane process design technology and optimization software ("Pani Digital") that enhances cost efficiencies of water processes such as desalination or wastewater treatment plants. A key component of the process design technology is a real-time responsive control system that ensures optimal operations based on computational models of plant components. The model components specify the plant operating rules which control component off/on schedules (on hourly update rates). Preliminary modeling work has highlighted a number of key technical development challenges: 1) the number of free control variables; 2) the number and complexity of the governing equations; 3) a mixture of non-linear and implicit equations, and; 4) time dependency due to component degradation. Traditional physical model-based optimization techniques are challenged to solve this type of problem. Predictive analytics, using historical plant data along with analysis, statistics, and machine learning techniques, may provide a more effective means to control the optimizations required. In a predictive analytics workflow, time enters the model on a fundamental level. Supervised machine learning techniques can take both controllable and observable/unobservable variables to simulate plant operation, and thereby enable effective control of the plant.

Pani Energy正在开发一种新型膜工艺设计技术和优化软件（“Pani Digital”），以提高海水淡化或废水处理厂等水处理工艺的成本效益。过程设计技术的一个关键组成部分是一个实时响应控制系统，该系统确保基于工厂组件的计算模型的最佳操作。模型组件指定控制组件关闭/打开计划（每小时更新率）的工厂操作规则。初步建模工作凸显了许多关键技术开发挑战：1）自由控制变量的数量; 2）控制方程的数量和复杂性; 3）非线性和隐式方程的混合; 4）由于时间依赖性组件退化。 传统的基于物理模型的优化技术在解决这类问题时面临挑战。使用历史工厂数据沿着分析、统计和机器学习技术的预测分析可以提供更有效的手段来控制所需的优化。在预测分析工作流程中，时间在基本层面上进入模型。监督机器学习技术可以采用可控和可观察/不可观察变量来模拟工厂操作，从而实现工厂的有效控制。