Multi Kernel Fusion Convolutional Neural Network for Wind Turbine Fault Diagnosis

To accurately diagnose the type of failure, make full use of computing resources and automatically identify different health conditions of wind turbine (WT), a new multi-kernel fusion convolutional neural network (MKFCNN) is proposed in this paper. The proposed method is based on a one-dimensional convolutional neural network (1-D CNN). Convolution kernels of different sizes are used in each layer of the network to extract features of different scales of data, which is inspired by the inception v1 model. Compared with ordinary CNN, its unique network design reduces a lot of network parameters, reduces the risk of network overfitting, and saves a lot of computing resources. The superiority of the proposed method is verified on a generic WT benchmark simulation model and compares with support vector machine (SVM), decision tree, random forest and CNN.