Prediction of Deep Formation Pressure Using CNN-LSTM Method

Accurate prediction of reservoir pressure, particularly in deep formations, is crucial for ensuring drilling safety and optimizing hydrocarbon exploration efficiency. Traditional empirical or semi-empirical methods often suffer from limited adaptability and weak generalization capability due to their reliance on simplified assumptions and predefined parameters. To address these limitations, this study proposes a deep learning framework based on a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) architecture for reservoir pressure prediction. The model leverages the feature extraction capability of CNNand the temporal sequence learning strength of LSTM to effectively capture nonlinear patterns in well log data. A dataset comprising logging and measured pressure data from ten wells in Block A of a Chinese basin was used for training and validation. Principal Component Analysis (PCA) was employed for feature selection and dimensionality reduction. The performance of the proposed CNN-LSTM model was benchmarked against BP, CNN-BP, and CNN-RNN models using standard evaluation metrics including Mean Squared Error (MSE), Mean Absolute Error (MAE), and the coefficient of determination (R²). Results indicate that the CNN-LSTM model achieved superior predictive accuracy, demonstrating its effectiveness in handling complex geological conditions and offering a robust approach for overpressure identification in deep reservoirs.