A rolling bearing remaining life prediction method based on the combination of multi-scale degradation features and CNN-GRU algorithm is proposed in order to accurately describe the degradation trend of rolling bearings in rotating machinery and improve the prediction accuracy of the remaining life of rolling bearings. The method firstly extracts typical time-domain, frequency-domain and time-frequency-domain features from the original vibration signal, normalizes them and constructs a training feature set; secondly, divides the feature set into training set and test set, and inputs them into the CNN-GRU model for training respectively, and the multi-scale feature set is input into the CNN-GRU to fully explore the relationship between the bearing degradation features and the degradation trend to realize the rolling bearing remaining life prediction. The experimental results on XJTU-SY bearing data and IEEE PHM2012 Challenge dataset show that the method can effectively improve the prediction accuracy of the remaining life of bearings and provide a research basis for rolling bearing health management and performance evaluation.