Aiming at the problem that it is difficult to identify the damage degree of rolling bearings under strong noise and variable load conditions， a new method based on improved anti-interference convolutional neural network （ACNN） for bearing damage identification under variable load conditions is proposed. First， the one-dimensional vibration signal of the rolling bearing is pre-processed to obtain labeled data samples， which are divided into training set and test set. Then the attention mechanism is introduced into each feature extraction layer of the convolutional neural network to establish the relationship between the feature extraction channels， and an improved model of bearing damage recognition under variable load conditions based on improved ACNN is obtained. After that， the training set data is input into the improved ACNN for learning. The obtained recognition model is applied to the test set， and the results of damage degree recognition are output. During the training process， in order to improve the anti-interference ability of the model， Dropout operation is performed on the convolutional layer. In addition， the original training samples are denoised to suppress overfitting. Finally， through the rolling bearing damage degree simulation test， the test is performed under variable working conditions. The results show that the proposed method can more accurately identify the degree of bearing damage under variable load conditions in noisy environments.