Study Of Multi-point Geostatistical Method Base On Distributed Computing

Multi-point geostatistics is a major method in reservoir modeling.It is also being extended to digital core remodeling,electrical resistivity tomography and other areas.The popularity of multi-point geostatistical algorithm is based on the continuous enhancement of the scale and resolution requirement of the object model.But with the sharp increase in the cost of computing,which has become the main limitation in the practical application of multi-point geostatistics.To solve this problem,this thesis proposes a distributed multi-point geostatistics algorithm based on the distributed computing system and Spark computing platform.According to the characteristics of Apache Spark computing architecture,this thesis proposes a distributed decomposition strategy for the SNESIM method,and implements find-grained problem partition and highly parallel computing tasks.The distributed strategy significantly improves the computational efficiency,and it doesn't require high-performance hardware and software,so it enhanced the adaptability of the algorithm.In addition,to reduce generation time and query time,we propose a solution based on the features of the data structure in Apache Spark,and adjust the data structure of search tree and store it in memory.The experimental results show the effectiveness and efficiency of the data structures.Comparing with existing serial geostatistics simulation method,this thesis introduces new ideas to the combination of the distributed multiple-point geostatistics simulation method.It greatly improves the efficiency and the adaptability of multi-point geostatistics stochastic simulation in the applications of the large-scale geological modeling.The experimental results show that the proposed method not only ensures the accuracy of the simulation,but also improves the efficiency of the algorithm,especially for the simulation of large high precision data.