Affecting Factors on History Matching Field-Level Coal Seam Gas Production from the Surat Basin, Australia

Reservoir history matching (HM) is an effective means for unconventional prospect studies, which enables the calibration of reservoir models and lays a solid foundation for production forecasting. This work presents a comprehensive reservoir simulation model comprising key static properties and dynamic data from the Surat Basin. The simulation model is deployed to discuss the HM experience of coal seam gas (CSG) wells, i.e., the challenges encountered and how these challenges have been overcome during the HM process. The challenges are mainly caused by uncertainties related to reservoir properties, tenement boundaries, and production data set integrity. Sensitivity analysis was conducted on the reservoir properties as well as the model boundary extent to ascertain the range of uncertainty and the relative weight of each parameter. An artificial neural network (ANN), a commonly applied machine learning process, was employed to predict the missing downhole pressure (DHP) data points for the following HM control comparison. The results demonstrate that the source and direction of water influx have played a central role in CSG production in the Surat Basin, especially for the well batches close to the simulation model boundary. Meanwhile, the novelty of ANN application for DHP prediction allows the HM the flexibility to switch from water rate constraint to bottom-hole pressure constraint; this eventually allows the HM simulation to reach the required degree of accuracy despite having wells with incomplete DHP data.