WITHDRAWN: A Novel Multi-Feature Information Fusion Method for Fault Diagnosis to Wind Power Converters

Fault diagnosis technique facilitates the development of wind turbines with low-maintenance costs and high-reliability. High accuracy is crucial to efficient fault-tolerant control, it enhances the system continuous operation capability in faulty states and improves economic revenue. At present, most high accuracy fault diagnosis approaches of wind turbines are founded on single feature representation. However, single feature-based method is not comprehensive enough, and the ability to describe faults is limited. In this article, a new multi-feature information fusion method Dempster-Shafer fusion multiscale approximate entropy (DSFMAE) is proposed for feature extraction. Firstly, multiscale approximate entropy is calculated to extract features of the signal at different scales, comprehensively characterizing the complexity of the signal, overcoming the incompleteness of single-feature extraction. Next, Dempster-Shafer theory is applied to fuse features in each scale to achieve feature-level multi-information fusion, and solve the conflict and uncertainty between features. Then the DSFMAE algorithm is used to diagnose converters faults in wind power system. Under the condition of changing wind speed, the average diagnostic accuracy reaches 99.81%. In addition, the noise robustness is quite strong with an accuracy of over 73% at 5dB. The standard deviations of accuracy are lower than 1.6% for different noise levels. Thus, the presented method can strongly tolerate changes in wind speed and interference from noise, and has excellent stability.