Improved Siamese Neural Betwork Based on Feature Fusion for Wind Turbine Fault Warning and Identification of Key Components

In wind turbine critical component fault warning and identification, the small data sample size leads to the problem of model under-fitting and low accuracy. In this article, we propose an improved siamese neural network-based fault warning and identification method for wind turbine critical components, which is based on the traditional siamese neural network with the addition of fault identification classification discriminative labels and ensemble learning algorithm metrics, significantly improving the accuracy of equipment fault warning and identification with few samples and reducing the misclassification of equipment operation status. The test results on different datasets show the superior performance of our model.