A wind turbine voiceprint anomaly early warning method and system based on small sample learning

Abstract

The application discloses a kind of wind turbine voiceprint abnormal early warning method and system based on small sample learning.The method comprises the following steps: collecting the voiceprint signal under normal and abnormal state of wind turbine and pre-processing;Construct a deep feature extraction network, use triplet loss function to measure learning training, map the voiceprint signal to high-dimensional feature space;Calculate the prototype vector of each category;Real-time acquisition of voiceprint signal and extract its eigenvector, calculate its Euclidean distance with each prototype vector, according to the nearest neighbor principle and preset threshold value, abnormality is judged and early warning is carried out.The present application overcomes the defects of the prior art, such as strong dependence on a large number of fault samples, poor generalization ability under small sample conditions, and high false alarm rate caused by environmental noise and working condition interference, realizes high-precision and high-robustness early warning for known and unknown abnormalities, and reduces operation and maintenance cost.

Images