Development of an integrated modeling framework for wind turbine health condition assessment

风电场的设计和运行遵循L3条件，即低成本、长寿命、低服务要求。风力涡轮机组件不仅因为其极端运行条件而故障率更高，而且由于需要大型起重机等特殊设备，与故障相关的成本也高于其他机械系统。此外，在极端天气条件下，风力涡轮机可能无法使用，导致长时间停机和生产损失。一个可靠、自动化的机器健康状态监测系统可以实现有效的预测性维护实践，并对L的三次故障都有直接贡献。这样的系统应该具有检测故障、将故障与根本原因联系起来并预测未来机器健康状态的能力。 在机器状态监测的一般领域，随着故障分类的可靠性和对操作和环境条件变化的稳健性的提高，故障诊断的解决方案变得越来越复杂。然而，几乎没有解决方案可以在系统级别确定故障的根本原因或预测故障。这些都是行业采用研发成果缓慢的主要原因，因为仅仅检测故障而没有可操作的信息，如故障将发生的时间和地点，并不一定能转化为成本节约。 包括基于底层物理原理的组件模型的系统级模型可以提供具有迫切需要的故障预测功能的解决方案的关键。在这个研究计划中，我们的目标是建立一个集成的风力涡轮机系统模型，能够代表实际系统的数字副本，特别是在其健康状况方面。在任何时间点，该模型都可以指示故障位置和严重程度。它可以用来模拟系统在未来时间作为故障进展的结果的行为，以实现故障预测。它还可以用来分析局部故障对系统的整体动态行为和其他组件的幸福感的影响，以便进行有效的干预。研究成果的成功交付将代表着朝着为风能和其他行业构建有效的预测性维护解决方案迈进了一大步。

Design and operation of wind power stations are guided by the “L3 conditions”, namely, low cost, long-lasting, and low service requirement. Not only do wind turbine components fail at higher rates because of their extreme operating conditions, the costs associated with the failure are also higher than other machinery systems due to the need for special equipment such as a large crane. Furthermore, wind turbines may not be accessible at times of extreme weather conditions, causing extended shutdown periods and loss of production. A reliable and automated machine health condition monitoring system can enable effective predictive maintenance practice and contribute directly to all three L's. Such a system should have the ability to detect faults, link them to root causes, and predict machine health state at future times. In the general area of machine condition monitoring, solutions in fault diagnosis are becoming increasingly sophisticated with improved reliability in fault classification and robustness to operating and environmental condition variations. However, very few solutions exist that can determine root causes of faults or predict failure at the system level. These are the main reasons for the slow industry adoption of research and development results because mere detection of faults without actionable information, such as when and where failure will occur, does not necessarily translate to cost savings. A system level model that includes component models based on underlying physical principles can provide the key to a solution with the much needed fault prediction capabilities. In this research program, we aim to establish an integrated wind turbine system model that can represent a digital copy of the actual system particularly in terms of its health condition. At any point of time, this model can indicate fault location and severity. It can be used to simulate the system's behavior as the result of fault progression in future times to enable failure prediction. It can also be used to analyze the effect of a localized fault on system's overall dynamics behavior and the well-being of other components to inform effective intervention. The successful delivery of the research outcome will represent a leap forward toward building an effective predictive maintenance solution for the wind energy and other industries.