Reasonable production prediction of coalbed methane (CBM) is of great significance for improving the economic benefit of CBM reservoirs. Current prediction methods for CBM production focus on the later stages of development, with few studies on early production forecasting. The objective of this work is to provide a reliable new idea for the early production prediction of CBM through various analyses and demonstrations. First, the CBM development modes are classified according to the production characteristics of the Panhe demonstration block of Shaanxi Province, China. Second, an efficient and feasible early production prediction model is established based on the geological potential and development potential. Finally, using the proposed model, different modes’ production characteristics and optimization strategies are analyzed. The research shows that the gas production profiles can be divided into two modes: single-peak mode (SPM) and multipeak mode (MPM). The peak production and average EUR of the SPM are 49.6% and 32.4% higher than those of the MPM, but the stable production period is only 0.2~1 year. In terms of the geological potential of CBM wells, the gas content, critical desorption pressure, and formation coefficient of the SPM are 6.7%, 13.3%, and 37.8% higher than those of the MPM, and the gas wells are mainly located in the high part of the coal seam (the average height difference is about 20 m). Besides, the concept of quasidesorption degree Pdq is innovatively introduced to characterize the development potential of gas well. The Pdq has an exponential relationship with CBM production, and the coefficient of the exponential term in SPM is approximately 22% larger than that in MPM. Moreover, the production of gas wells is greatly affected by the continuity of production. In the process of gas production, the influence of factors such as equipment shutdown should be minimized. To examine the applicability of the proposed method, the model is applied to an actual CBM well in Panhe, and the prediction accuracy is higher than 85%.
Read full abstract