The robotic grinding process is affected by both dynamic and static factors and has dynamic characteristics such as complex coupling and high time-varying nonlinearity. To solve the problems of difficult feature selection of dynamic factors and low prediction accuracy of surface roughness caused by only considering static factors, a prediction method of surface roughness of robotic grinding considering the influence of dynamic factors is proposed by combining deep learning technology. Firstly, the convolutional neural network is used to automatically extract the spatial features of dynamic factors in the grinding process, and capture complex dynamic behaviors of robotic grinding. The temporal features are extracted from the obtained spatial features through the bidirectional long short-term memory network to characterize the dynamic changes of robotic grinding. The attention mechanism is introduced to realize the automatic weight distribution of spatial features, temporal features and static factors. The improved whale optimization algorithm is used to adaptively optimize the hyperparameters of the bidirectional long short-term memory network to improve the convergence speed and adapt to the dynamic changes of robotic grinding. Secondly, according to the proposed prediction method, an IWOA-CNN-BiLSTM-Attention surface roughness prediction model considering the influence of dynamic factors is formulated. Thirdly, the robotic grinding experiment is carried out. The spatial and temporal characteristics of the extracted dynamic factors, the collected static factors, and the measured values of surface roughness are normalized to construct the experiment dataset. Finally, the experimental dataset is input into the prediction model for model training, and the surface roughness prediction of robotic grinding considering both dynamic and static factors is realized. The effectiveness of the proposed method is evaluated by comparative experiments. The mean absolute percentage error, root mean square error, and coefficient of determination of the corresponding prediction model are 0.027 6, 0.029 5, and 0.998 8, respectively. Compared with the comparison prediction model, the prediction accuracy is improved by 17.14%, 13.65%, and 21.35%, respectively. compared with the comparison prediction model.