Proceedings of the
The Nineteenth International Conference on Computational Intelligence and Security (CIS 2023)
December 1 – 4, 2023, Haikou, China
Coalbed Methane Production Forecasting Based on CNN-GRU Model
College of Electrical and Power Engineering, Taiyuan University of Technology, China.
ABSTRACT
To further improve the accuracy of coalbed methane production prediction, a coalbed methane production forecasting model based on Convolutional Neural Network and Gated Recurrent Unit (CNN-GRU) is proposed. Firstly, the Random Forest algorithm is used to analyze the proportion of feature importance in mining and drainage, determining that bottomhole flowing pressure, dynamic liquid level, wellhead casing pressure, and mining and drainage equipment speed are the main controlling factors affecting coalbed methane production during the mining and drainage process. Then, the CNN-GRU model for coalbed methane production prediction is established, utilizing 1D CNN to process sequential data and learn local features. These features are used as input to the GRU model to provide better long-term dependency modeling capabilities. Finally, the effectiveness of the model is verified by taking the southern Ma Bi area of Qinxian County, Shanxi Province, as the research object and predicting the gas production for the next 60 days. The prediction results show that compared to the other three models, the CNN-GRU model exhibits higher prediction accuracy. Accurate predictions can reduce the probability of accidents such as CBM explosions and fires, and protect the lives of workers and the integrity of CBM well facilities.
Keywords: Convolutional neural network, Gated recurrent unit, Coalbed methane, Production forecasts, Random forest, Drainage safety.
Download PDF
College of Electrical and Power Engineering, Taiyuan University of Technology, China.
ABSTRACT
To further improve the accuracy of coalbed methane production prediction, a coalbed methane production forecasting model based on Convolutional Neural Network and Gated Recurrent Unit (CNN-GRU) is proposed. Firstly, the Random Forest algorithm is used to analyze the proportion of feature importance in mining and drainage, determining that bottomhole flowing pressure, dynamic liquid level, wellhead casing pressure, and mining and drainage equipment speed are the main controlling factors affecting coalbed methane production during the mining and drainage process. Then, the CNN-GRU model for coalbed methane production prediction is established, utilizing 1D CNN to process sequential data and learn local features. These features are used as input to the GRU model to provide better long-term dependency modeling capabilities. Finally, the effectiveness of the model is verified by taking the southern Ma Bi area of Qinxian County, Shanxi Province, as the research object and predicting the gas production for the next 60 days. The prediction results show that compared to the other three models, the CNN-GRU model exhibits higher prediction accuracy. Accurate predictions can reduce the probability of accidents such as CBM explosions and fires, and protect the lives of workers and the integrity of CBM well facilities.
Keywords: Convolutional neural network, Gated recurrent unit, Coalbed methane, Production forecasts, Random forest, Drainage safety.
Download PDF