Model-Based Temperature Fault Detection for Individual Measuring Points

For wind turbines, condition monitoring is an essential part of research. To minimize downtime of the plants and to prevent extensive damage, it becomes inevitable to detect faults at an early stage, especially, since plants are increasingly placed in poorly accessible areas, like deserts, arctic or offshore regions. The following paper concentrates on temperature monitoring and discusses an observer-based approach reconsidering the model's accuracy and reliability for condition monitoring purposes. As novel approach the fault detection is focused on individual measuring points rather on the overall system. The presented model is developed as an equivalent thermal circuit. To minimize the difference between modeled and measured temperature curve (residual), a observer is implemented and configured. The advantages and disadvantages as well as the improvement of the presented approach compared to presently used methods are outlined.