Well Placement Optimization and History Matching Using Hybrid Methods
MetadataShow full item record
For field development, hydrocarbon recovery is considered as one of the main objectives through the whole life cycle of the field. The reservoir management during the field development ranges from planning for the field development, drilling and well optimization and rejuvenation of the mature field till abandonment. The ultimate goal of field development comes down to maximizing profits. Better understanding of the reservoir is essential to achieve this goal. Among the many problems of reservoir management, well location optimization for maximum recovery and prediction of the reservoir performance are important ones to solve. We propose a hybrid sampling method for the well placement problem. To decide on the well location, evolutionary algorithms were used and updated starting from an initial response surface consisting of the candidate well locations selected using random sampling method as well as a dynamic measure probability map serving as an indicator of remaining hydrocarbon in the reservoir. We applied this approach to a mature field case because the dynamic measure can capture the complex response of the reservoir and provide information of sweep and drainage areas. We presented a well placement optimization method for primary depletion in green fields. Unlike the complex responses in the mature field, pressure depletion is the main recovery mechanism in the green field. For pressure depletion in green fields, we adopted diffusive time of flight, instead of convective time of flight for the dynamic measure. By using a fast marching method, we can get the propagation of the pressure front very fast with a single non-iterative calculation. This diffusive time of flight was consolidated into the dynamic measure probability map which is the starting point of our search space, for the evolutionary algorithm. This method was extended to a dual porosity model by considering the flow between matrix and fracture. Finally, a structured history matching approach which consists of global calibration and local calibration is presented. Key global parameters which heavily affect the model response are selected through a sensitivity analysis. Design of experiments and response surface methodology with evolutionary algorithms such as genetic algorithm are used to calibrate these key global parameters. Then, local calibration using streamline based sensitivity and generalized travel time inversion technique is performed.
SubjectWell Placement Optimization
Fast Marching Method
Kim, Jeong Min (2015). Well Placement Optimization and History Matching Using Hybrid Methods. Doctoral dissertation, Texas A & M University. Available electronically from