Abstract: Aiming at the problem that the existing model-driven methods are difficult to estimate parameters and the data-driven methods lack physical prior knowledge, resulting in low prediction accuracy and insufficient generalization ability of energy consumption in milling process, a milling energy consumption prediction method integrating particle swarm optimized physics-guided long short-term memory network (PSO-PG-LSTM) is proposed. Firstly, a physics-guided long short-term memory network (PG-LSTM) model is constructed, and the energy non-negativity constraint term is introduced in the loss function to make the training direction of the model conform to the physical law. Secondly, an adaptive search method for the optimal hyperparameters and physical constraint weights of the model based on the particle swarm optimization (PSO) algorithm is proposed to improve the prediction accuracy of the model. Finally, the above model and method are verified by the milling process of carbon fiber reinforced polymer and 45 steel. The results show that the constructed model is superior to the existing methods in terms of MAE, RMSE and R², and has good application prospects.