Exploring Production Casing Pressure Time-Series Trends Based on an LSTM-CNN Model

Production casing pressure is a key parameter to ensure well integrity and safe operation in oil and gas fields. In this paper, we propose a hybrid LSTM-CNN model that combines a long short-term memory network with a convolutional neural network for predicting casing pressure trends. Data preprocessing includes missing value estimation, outlier detection, and data normalization to ensure high-quality inputs. The experimental results compare the performance of multiple models of RNN, CNN, LSTM, LSTM-CNN, and on RMSE, MAE, and R². The LSTM-CNN has a significant advantage in capturing the temporal dependence and local features of production casing pressures, and it can provide accurate and reliable prediction support for production safety and well control monitoring.