Abstract:

Accurate and efficient condition monitoring is the key to enhance the reliability and security of wind turbines. In recent years, the intelligent anomaly detection method based on deep neural networks (DNN) has been receiving increasing attention. Since accurately labeled data are usually difficult to obtain in real industries, this paper proposed a novel deep small-world neural network (DSWNN) on the basis of unsupervised learning to detect the early failure of wind turbines. During the deep belief network (DBN) construction, a regular auto-encoder network with multiple restricted Boltzmann machines (RBM) was first stacked and pre-trained by using unlabeled supervisory control and data acquisition (SCADA) data of wind turbines. After that, the trained network was transformed into a DSWNN model by the randomly add-edges method, where the network parameters are fine-tuned by using minimal amounts of labeled data. In order to deal with the disturbances of wind speed and reduce false alarms, an adaptive threshold based on extreme value theory was presented as the criterion of anomaly judgment. The DSWNN model is excellent in depth mining data characteristics and accurate measurements. Finally, two failure cases of wind turbine anomaly detection were given to demonstrate the validity and accuracy of the proposed DSWNN contrasted with the DBN and DNN algorithms.

Key words: wind turbine, supervisory control and data acquisition (SCADA) data, fault diagnosis, deep small-world neural network (DSWNN), adaptive threshold