Exploring the Power of Neural Basis Expansion Analysis for Time-Series Forecasting Neural Network for Deep Coalbed Methane Production Prediction

Summary China’s coalbed methane (CBM) is widely distributed and very rich in resources, and the current proven reserves of deep CBM with a burial depth of more than 1500 m are abundant. As a clean unconventional resource, the accurate production prediction of deep CBM is of great significance to economic exploitation, environmental protection, and energy security. Compared with traditional decline curve analysis methods and numerical simulation algorithms, deep learning algorithms have significant advantages in feature extraction and production prediction of long time-series data. Thus, in this paper, we selected 16 deep CBM wells with a total of 2,136 production data in the main gas-producing areas of the Ordos Basin to explore and validate the prediction capability of neural basis expansion analysis for time series (N-BEATS) neural networks for deep CBM. The prediction accuracy of the N-BEATS neural network is high for both high- and low-producing CBM wells. Meanwhile, the blind test results on the test data set are stable after training using a mixture of high- and low-producing training data sets. Additionally, the newly introduced N-BEATS algorithm shows clear advantages in prediction accuracy, achieving the lowest mean absolute error (MAE) (0.0034) and root mean square error (RMSE) (0.0052) on CBM 12, significantly outperforming traditional methods such as autoregressive integrated moving average (ARIMA) (MAE: 0.0839, RMSE: 0.0846) and machine learning models like support vector machine (SVM) (MAE: 0.0733, RMSE: 0.0739), thereby demonstrating its superior capability in deep CBM production forecasting. More importantly, the accurate and stable prediction results achieved by the N-BEATS neural network can provide strong technical support for optimizing production strategies, reducing exploration risks, and improving the overall efficiency of deep CBM development, thereby promoting the intelligent and sustainable growth of the deep CBM industry.