Two-layer PSDG based fault diagnosis for wind turbines

Fault diagnosis of the wind power system is a common problem in the routine maintenance of the wind farm. However, most researches and practical applications focus on separate components, which do not take correlations of different components into consideration. This paper presents a novel two-layer Probabilistic Signed Directed Graph (PSDG) method for the fault diagnosis of the overall wind power system. First, we establish a sensor data layer to detect the variations in the wind power system based on the complex network theory. Then we utilize the fault list to build a Bayesian network. We figure out the search sequence based on the comprehensive ranking index and introduce two evaluation indices to evaluate our methods. Finally, the proposed method is tested on actual fault data from a wind farm in the northwest of China. The result shows that the proposed method not only ensures the accuracy of the fault diagnosis, but also guarantees the quickness of the fault search.