The traditional rolling bearing residual life prediction methods lack a clear learning mechanism and low model prediction accuracy, which cannot effectively extract the important degradation information features contained in the differences between different timing features. In order to further improve the accuracy of the prediction model, this paper proposes a prediction model of the temporal convolutional network (BiLSTM). Firstly, a multi-dimensional typical degradation feature set is built; Secondly, the time convolution network (TCN) integrating a self-attention mechanism to capture the dependence on various time scales, learn feature weights and improve the model’s attention to key features; finally, the BiLSTM time series prediction model is used to predict the bearing degradation trend and realize high-precision bearing life prediction. Using the data of seven motor bearings to evaluate the effectiveness, the results show that the TCN-SA-BiLSTM model showed excellent performance in the bearing life prediction task, which can significantly enhance the robustness of prediction.