Multivariate approach to the gas production forecast using early production data for Barnett shale reservoir

Abstract

This study examined the relationship between the early production data (peak production rate, three-, six-, nine-, and 12-month cumulative gas production) and the long term cumulative gas production (CGP) at three, five, seven, and ten years through a numerical simulation and production data analysis of 546 shale gas production datasets. Forecast models for shale gas production were developed using two multivariate approaches: a polynomial approach and a response surface methodology approach (RSM). The developed forecast models showed that the early production data strongly correlated with the long-term shale gas production (R2 > 0.99). The field applicability of the developed models was examined using these models to forecast the production of 237 shale gas wells in the Barnett reservoir. The field application results revealed the forecasted value from the multivariate polynomial forecast models to have good agreement with the measured shale gas production (R2 > 0.80). The RSM model showed a better prediction result for three-year CGP than the polynomial model. In contrast, the RSM forecast models yield less predictive results for five, seven, and ten-year CGP. Therefore, the RSM model was used to forecast the three-year CGP and the multivariate polynomial models were used to forecast the five-, seven-, and ten-year CGP in the Barnett reservoir. The proposed models could predict the long-term production of new shale gas wells from a minimum one-year production history without requiring geological and well/hydraulic fracturing data. The proposed method can be applied to the development of forecast models for other shale gas reservoirs.