Fitting and predicting models for coalbed methane wells dynamic productivity

Abstract：Based on modern artificial intelligence theory and mathematical statistics theory,BP neural network model and monthly/cumulative production model for fitting and predicting coalbed methane（CBM） wells productivity were established to verify the validity of these models by examples.The application results show that two models can match production data of CBM wells and quantitatively predict it.BP neural network model has high accuracy in matching the data points of gas production and predicting well productivity in a short term but not in a long term.Thus,this model is appropriate for the short-term predictions for productivity of CBM wells,even though the wells with unsteady gas production.Monthly/cumulative production ratio model has high accuracy in matching the change trend of monthly/cumulative production ratio and predicting well productivity not only in a short term but also in a medium-long term.However,the validation of this model is determined by the exponential relationship between monthly/cumulative production ratio and production time.Therefore,this model is very suitable for predicting the futural productivity of CBM wells with stable gas production during the past.