To achieve optimized allocation of power resources on a large scale, the country is gradually developing a wide-area complex power grid that includes transmitting electricity from west to east and nationwide interconnections. In the development plan of power grid companies, the monitoring of the condition of wind power generation and substation equipment is an important research area for smart grids. It serves as a crucial technology for online perception and intelligent sensing in smart grids. Once faults occur in power system equipment, they not only cause damage to the equipment but also impact normal production in various industries and pose a threat to the safety of electricity consumption in society. Therefore, monitoring the condition of wind power generation and substation equipment, as well as predicting faults, has become a significant topic. This paper proposes a fault detection method for substation equipment based on a deep learning model. By utilizing multi-scale receptive fields and compression-with-activation modules, an improved one-dimensional convolutional neural network-based fault diagnosis model for substation equipment is introduced. Multiple-dimensional features are extracted using multi-scale receptive fields in the network, while the channel attention mechanism of the SE (Squeeze-and-Excitation) module learns the weights of feature channels. By recalibrating the weights to assign higher weights to useful features, the accuracy of the final classification results is improved.