DDDAS - SMRP: A Framework For the Dynamic Data-Driven Fault Diagnosis of Wind Turbine Systems

摘要本合作研究（0540132,PI: Ding Yu, Texas a&m University；和0540278，PI: Tang Jiong, University of Connecticut）将为风电机组诊断提供一个动态数据驱动的框架。这种新方法与目前的实践有着根本的不同，目前的实践由于建模方法和数据收集和处理策略的非动态和非鲁棒性而受到限制。该框架包括两个鲁棒数据预处理模块，用于突出故障特征和去除传感器异常，三个相互关联的多级模型，描述系统行为的不同细节，以及一个用于鲁棒局部询问的动态策略，该策略允许根据特定的物理条件和相关的风险级别自适应地采取测量。它将历史数据和在线信号结合到系统建模中，并能够自适应地改变数据收集过程，以最佳地捕获关键系统特征。总的来说，这些组件构成了一个强大而敏感的风力涡轮机诊断系统。通过与工业界的密切合作，这项研究将得到加强，这将提供丰富的历史传感器数据和详细的系统特性，并提供现场实施的机会。拟议的研究将对风能的利用具有战略重要性，风能是目前最可行的清洁能源替代品。如今，在广大的低风速地区，风能的成本更高，无法与传统能源竞争，主要是由于风能的维护成本高，对诊断技术的信心不高。这种动态和数据驱动的故障诊断将在实现具有成本效益的风力发电方面发挥关键作用。叶片和齿轮箱故障诊断的进展也将有利于发电、汽车、航空航天和发动机工业。同时，这项研究的合作性质将为学生提供多学科的培训，并将为大学带来工业视角。该项目将通过课程开发对教育产生长期影响，并将通过外展到高中促进公众对清洁能源概念的认识。

CMS-0540132, PI: Yu Ding, Texas A&M UniversityCMS-0540278, PI: Jiong Tang, University of ConnecticutAbstractThis collaborative research (0540132, PI: Yu Ding, Texas A&M University; and 0540278, PI: Jiong Tang, University of Connecticut) will provide a dynamic data-driven framework for wind turbine diagnosis. This new methodology is fundamentally different from the current practice whose performance is limited due to the non-dynamic and non-robust nature in the modeling approaches and in the data collection and processing strategies. This framework consists of two robust data pre-processing modules for highlighting fault features and removing sensor anomaly, three interrelated, multi-level models that describe different details of the system behaviors, and one dynamic strategy for the robust local interrogation that allows for measurements to be adaptively taken according to specific physical conditions and the associated risk level. It incorporates both historical data and on-line signals into the system modeling, and enables the ability to adaptively alter data collection procedures to best capture the critical system features. Collectively, these components lead to a robust and sensitive diagnosis system for wind turbines. This research is strengthened by a close collaboration with industry that will provide abundant historical sensor data and detailed system characterization, and also offer in-field implementation opportunities. The proposed research will have strategic importance on the utilization of wind energy that is currently the most viable clean energy alternative. Today, in the vast areas that have low wind speed, wind energy cannot compete with traditional energy sources as it has a higher cost, mainly owing to its high maintenance costs and low confidence in the diagnosis technology. This dynamic and data-driven fault diagnosis will play a key role in enabling a cost-effective generation of wind electricity. Progress in the fault diagnosis of blades and gearboxes will also benefit the power generation, automobile, aerospace, and engine industries. Meanwhile, the collaborative nature of this research will provide students with a multidisciplinary training and will bring industrial perspective to the universities. This project will have a long-term impact on education through the curriculum development and will promote the public awareness of clean energy concept through outreaches to high schools.