Digital Twins of hydroelectric generating units: AI insight from sensor data combined with physics-based simulation

水力发电机组的数字孪生：传感器数据的人工智能洞察与基于物理的模拟相结合

• 批准号: 556353-2020
• 负责人: Gosselin, Frederick
• 金额: $ 9.5万
• 依托单位: École Polytechnique de Montréal
• 依托单位国家: 加拿大
• 项目类别: Alliance Grants
• 财政年份: 2020
• 资助国家: 加拿大
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=706310
• 关键词: Digital; Twins; hydroelectric; generating; units

The operation of hydroelectric generating units is evolving, creating the need for better condition monitoring and accurate modeling of the power-generating equipment. Whereas hydro units used to be overdesigned and continuously operated at baseload conditions, today's new units and refurbished ones are designed with stringent performance and operational regime requirements. Notably, they are relied upon for power regulation to integrate more intermittent wind and solar energy in the grid and must undergo frequent starts and stops. The consequence of all this is greater stresses on the units, which can lead to failures and loss of productivity. It is thus essential to better monitor the equipment and predict its behaviour and outcome. Polytechnique Montreal, Hydro Quebec and Maya HTT will work together to develop the tools to build the Digital Twin of a hydro unit. Such a digital twin will combine live sensor data with physics-based modeling through artificial intelligence to achieve real-time simulation of a hydro unit. It will allow predicting failures, optimising maintenance schedules, and simulate scenarios of usage and wear of the equipment. To this end, a group of more than 10 highly qualified personnel will use Physics-Informed Neural Networks and Proper Generalized Decomposition to develop reduced-order models of academic model systems exhibiting some of the same physics as Hydro Quebec's equipment and incorporate them in Maya HTT's software solution. This gradual approach will enable us to overcome the same challenges required to combine experimental measurements and physical modeling with artificial intelligence to develop digital twins for real industrial equipment.

水力发电机组的运行正在发生变化，这就需要更好的状态监测和发电设备的准确建模。水力发电机组过去设计过度，在基本负荷条件下连续运行，而今天的新机组和翻新机组在设计上具有严格的性能和运行制度要求。值得注意的是，它们被依赖于电力调节，以将更多间歇性的风能和太阳能整合到电网中，并且必须经历频繁的启动和停止。所有这一切的后果是对设备造成更大的压力，这可能会导致故障和生产力损失。因此，必须更好地监测设备并预测其行为和结果。 蒙特利尔理工学院、魁北克水电公司和玛雅HTT公司将共同开发工具，构建水电机组的数字双胞胎。这样的数字孪生兄弟将通过人工智能将现场传感器数据与基于物理的建模相结合，实现水电机组的实时仿真。它将允许预测故障、优化维护计划，并模拟设备的使用和磨损情况。 为此，一个由10多名高素质人员组成的小组将使用物理信息神经网络和适当的广义分解来开发学术模型系统的降阶模型，这些系统展示了与魁北克水电公司的设备相同的一些物理特性，并将它们整合到Maya HTT的软件解决方案中。这种循序渐进的方法将使我们能够克服将实验测量和物理建模与人工智能相结合以开发用于真实工业设备的数字双胞胎所需的相同挑战。