Prediction Of Coalbed Methane Wells Desorption Pressure Based On BP Neural Networks And Its Applications In The Central-southern Qinshui Basin

Desorption pressure is the key parameter of coalbed methane(CBM)wells productivity prediction,which plays an important role in gas wells drainaging.The production of CBM wells shows that there are differences between actual desorption pressure obtained from the drainage and critical desorption pressure calculated by the adsorption theory,which is detrimental to the development of drainage system.It is of great significance to predict the actual desorption pressure of gas wells accurately and to discuss the relationship between the actual desorption pressure and the critical desorption pressure.In this paper,the No.3 coal seam of Changzhi-Anze block in central-southern Qinshui basin was selected as research object.Based on the geological settings of coalbed methane and the production data of CBM wells,the distribution characteristics and differences of the desorption pressure of CBM wells were analyzed,and the main factors affecting the actual desorption pressure and their influence mechanism were investigated.Further,the prediction model of actual desorption pressure were established based on BP neural network,and the mathematical model for the reasonable drainage intensity in uniflow stage of CBM wells was constructed.Results show that the main coal types of No.3 coal seam in the study area is lean coal,which has a greater buried depth,low permeability,strong adsorption capacity,short adsorption time and high deformation degree of coal body;The gas content gradually increased from the east and the west part to the central of the basin;The No.3 coal seam is the under pressure reservoir,and the pressure gradient of the Changzhi block in the eastern part of the study area is obviously lower than that of the western Anze block.The desorption pressure of CBM wells in the east and the west is lower than that in the central,showing a banding distribution.The actual desorption pressure of CBM wells in the Anze block is generally greater than the critical desorption pressure;however,for Changzhi block,the desorption pressure is approximately equal to the critical desorption pressure.Gas content,reservoir pressure and adsorption capacity are the fundamental factors affecting the desorption pressure.Desorption hysteresis and the gas content measurement error are the main reasons for the differences between autual and critical desorption pressure,the differences of the reservoir pressure gradient are the main reasons for the discrepancy of the actual desorption pressure between the east and the west part.According to the influence weight of each factor on desorption pressure,this paper optimized the reservoir pressure,the minimum horizontal principal stress,reservoir temperature,buried depth,gas content and water production before gas production as the input layer parameters,constructed the BP neural network prediction model with the topological structure of 6-12-1,this model has high prediction precision,and the average error is 9.12%.Based on the percolation theory,considering starting pressure gradients,a mathematical model was established to calculate the reasonable drainage intensity of CBM wells in uniflow stage,the model has fairly good applicability.