As one of the important equipment of fossil-fuel power station, air cooler is widely used in the three north regions of China, where coal is rich and water is scarce. Its heat exchange efficiency has a considerable impact on unit efficiency. Therefore, conducting intelligent identification and analysis of ash accumulation in air coolers and effectively cleaning them is of great significance for improving the power generation efficiency of power plants. Considering the color and texture changes caused by the degree of dust accumulation in the visible light image of the surface of the air cooler. A method of identifying and analyzing ash deposition in air cooler based on convolutional neural network transfer learning is proposed. We split the training set, test set, and validation set using 6:2:2. Using a convolutional neural network algorithm model to identify air cooler images with varying degrees of fouling, and mining and analyzing the corresponding relationship between air cooler fouling images and fin cleaning. The results show that for the existing 5 levels of air coolers with ash accumulation, the model with higher recognition rate is Vgg19, with recognition accuracy of 0.98 and recall rate of 0.98, respectively, while the model with lower recognition accuracy is AlexNet, with an accuracy rate of 0.78. The proposed method for analyzing the ash accumulation status of air coolers can directly, real-time, and quantitatively analyze the on-site ash accumulation of air coolers, and provide reference and new considerations for achieving intelligent cleaning and operation and maintenance on site.