Abstract: Aiming at the intelligent prediction of reservoir parameters, a short-term memory neural network model with petrophysical domain knowledge as the constraint is established, and the intelligent reservoir parameter prediction driven by data and petrophysical knowledge is realized by reconstructing the loss function and improving the model weight updating process, and using the petrophysical knowledge constraints to guide the training of the short-term memory neural network. The results show that the long and short-term memory neural network reservoir parameter prediction model based on petrophysical domain knowledge constraints can improve the accuracy of reservoir parameter prediction, and the accuracy of the reservoir parameter prediction results can be improved by about 10% by comparing with the empirical formulas as well as the data-driven neural network. The study verifies the effectiveness of integrating petrophysical domain knowledge into long- and short-term memory neural networks, and provides a scientific basis for the application of deep learning in reservoir parameter prediction.