A Fault Diagnosis Model Based on Attention Convolution Network

In recent years, with the continuous growth of the power market, the structure of power equipment has become increasingly complex and intelligent. Traditional equipment fault diagnosis methods are mainly achieved through manual judgment or detection of equipment parameters, which are prone to noise interference, low accuracy, and low efficiency. To improve the accuracy of fault diagnosis, a network model based on attention convolution is proposed to achieve adaptive power equipment fault diagnosis. This model mainly consists of three parts: data preprocessing, linear attention layer, and spatiotemporal convolutional network. Firstly, the data is standardized and expanded. Then, a dual-channel convolutional network with attention is used for training and testing. Finally, FC detection heads are used for fault classification. The experimental results show that under different training sample ratios, this method can accurately identify faults in power equipment and has strong generalization.