GPU Based Fracture Modeling And Pressure Distribution Simulation Of Shale Reservoirs

With continuous exploitation, a lot of existing oil and gas fields in our country have entered a late-time production stage, and the oil production is declining year by year. Further, within the new reservoirs whose reserves have been proven,70%of them are low permeability reservoirs. In order to enhance oil recovery, efforts must be made on the exploitation of oil fields with low and ultralow permeability. Among these fields, oil shale is the most commonly seen, which is typically characterized with complex geological structures. Oil shale typically involves two kinds of fracture systems, i.e. the natural fractures and the hydraulic fractures, and the modeling of such systems has direct impact on the subsequent development strategy. Thus, the studies on modeling and simulation of reservoirs with complex fracture systems have great implication on development of oil and gas fields with ultra-low permeability.Currently, studies of3-dimensional modeling are mostly based on the geostatistical theory. Due to limitations of the traditional geostatistical methods on description of the spatial distribution, the multi-point geostatistical methods are becoming a major tool for reservoir modeling. But during application of multi-point statistical methods to the large-scale geological modeling, CPU serial implementation is typically limited by some parameters, which may lead to the slow computational efficiency, excessive memory consumption and even program collapse. Above restrictions make it difficult to achieve the desired effect during practical application of point simulation.Study on fine scale structure and natural fracture distribution of shale reservoirs lead to a more comprehensive level of pressure simulation of fractured reservoirs. Current fluid flow simulation model of fractured reseroir include equivalent continuous model, discrete model and integrated model, and the key point during flow mechanism investigation of such fields is how to implement the fracture distribution obtained with geostatistics into the percolation simulation model and the study of fracture parameters on pressure distribution and production.Parallel implementation of the SNESIM multi-point geostatistics is carried out in this paper, and the fractured reservoir model is obtained with stochastic simulation. Investigations on the percolation mechanism and pressure distribution are then provided for vertical fracture wells and multi-stage fractured horizontal wells with both analytical and numerical methods. Main contents of the thesis are as follows:1. Multi-point geostatistics is introduced into the simulation of spatial distribution in shale reservoirs. Single grid and multi-grid implementation of SNESIM stochastic simulation strategy is given, and the module task of probability distribution function of the node to be simulated is allocated to GPU parallel execution. Taking the data template which contains300positions as example, the parallel methods is about25times faster than the serial version, and the memory cost reduces about800times, which leads to significant improvement of the three-dimensional modeling efficiency of naturally fractured reservoirs.2. Improvement is made to the GPU based SNESIM parallel approach. For multi-grid method, buffer is set in the GPU memory to reduce the interaction time of the data transfer between the CPU and GPU. Considering the large amount of repeated execution during calculation of data template offset, pre-buffer method is introduced to reduce the calculation load, the fast locating of the node is realized by changing the offset value, and optimization is made on the memory usage effciency of parallel algorithms.3. Research on seepage behavior of shale reservoir is conducted. Applying the transient source function, the analytical solutions to the vertical fracture well and the multi-stage fractured horizontal well with various boundary conditions are derived. Comparison of the calculated pressure to an in-situ field data in our country is made, which validates the analytical method proposed.4. Numerical simulation of reservoir pressure distribution is studied. The PEBI grid method is applied for mesh generation to the3-dimensional model obtained with stochastic simulation, and the seepage equation is then discretized. The obtained linear equations are solved with GMRES iterative method, and the pressure distribution of multi-fractured horizontal well is given out.5. With the numerical simulation method proposed, the pressure field of multi-stage fractured horizontal wells in shale reservoir is provided and then analyzed. The influence of permeability in the horizontal direction on homogeneous reservoirs with ultralow permeability, as well as influence of permeability on heterogeneous reservoir is investigated. The calculated results are consistent with in situ production data.6. Considering the natural fracture distribution model obtained with the stochastic simulation in fractured reservoirs, a dual-porosity single permeability quasi-steady seepage equation is proposed, and the bottom-hole pressure of the horizontal well is derived. The seven obvious flow regimes during production of multi-stage fractured horizontal well are analyzed.