| dc.description.abstract | Diagnosing hydraulic fracture performance is essential to evaluate and optimize fracturing treatment designs in horizontal wells. Distributed temperature sensing (DTS) is a valuable tool to monitor downhole conditions and diagnose hydraulic fractures. Although various temperature prediction models have been proposed to interpret the measured temperature data, quantitative interpretation is still challenging. To predict temperature in near-wellbore regions accurately, a forward model is needed to consider both reservoir and wellbore domains in transient conditions. In addition, the model has to be computationally efficient to implement history matching for field-scale reservoirs. Yoshida et al. (2016) developed a comprehensive thermal and flow model and successfully interpreted the DTS temperature data. This numerical model consists of a reservoir model and a wellbore model, which are coupled iteratively through boundary conditions. In each domain, mass, momentum and energy conservation are solved in transient conditions to obtain profiles of wellbore and sand=face temperature during fracturing treatment, shut-in, and production in a fractured well. This model enables us to interpret the DTS temperature quantitatively; however it is not practical for field applications from the point of view of computational efficiency. This study presents a parallel version of the numerical thermal and flow model. Parallel computing is generally used as an effective way to improve computational speed. A parallel computing interface, MPI (Message Passing Interface) is implemented in this study because of its flexibility. The parallel model allows us to simulate the temperature in field-scale reservoirs efficiently. Results of improvement are shown as comparisons of computational speed between the original model and the parallel model during the processors of water injection and production. | en |
