A Predictive Model for CBM Wells Productivity Based on Artificial Neural Network
Production from coalbed methane (CBM) wells is a complex nonlinear system controlled by geological,engineering and human-made factors,which leads inevitable difficult to describe the relationship between production behavior and its influencing factors.Due to some unavailable parameters (such as permeability),the conventional methods including type curves,material balance and numerical simulation with idealistic assumptions do not work well in matching and predicting the production performance of CBM wells.Based on BP neural network with the strong ability of approximation in matching the nonlinear,unstable and complicated system,a predictive model was established to match the history production data of CBM wells and predict the futural performance in short term.The results show the BP neural model matches production performance of CBM wells successfully.Case studies show this model has high accuracy and good reliability in matching gas production and the fitting accuracy increases as the number of outliers in gas production data decreases.Therefore,BP network can quantitatively predict CBM well performance without clear knowledge of coalbed reservoir and even without sufficient production data during the early stage of development.
predictive model CBM well productivity BP neural network
Yumin LV Dazhen TANG Hao XU Shenhua JIAO
School of Energy Resources,China University of Geosciences (Beijing),Beijing,China Civil & Environment Engineering School University of Science & Technology Beijing,Beijing,China
国内会议
北京
英文
275-279
2011-10-20（万方平台首次上网日期，不代表论文的发表时间）