Production Prediction of Natural Gas Well Based on Parallel Neural Network

Accurate production prediction of natural gas wells is crucial for optimizing resource utilization and improving gas field management. However, the gas production data, influenced by multiple factors such as oil pressure and casing pressure, poses significant challenges to conventional predictive models. In this study, we propose a novel neural network model, Dual-LSTM, which employs two identical Long Short-Term Memory (LSTM) networks in parallel to process oil pressure and casing pressure data separately. This parallel architecture decouples the complex interdependencies between these factors and improves the accuracy of production prediction. We conduct extensive experiments on a dataset comprising several years of production data from four natural gas wells, and the results demonstrate that the Dual-LSTM model outperforms traditional Convolutional Neural Network (CNN) and LSTM models. The proposed Dual-LSTM model offers a robust and reliable framework for enhancing production prediction in the natural gas industry.