Research On Reverse Time Migration Data Processing Method Based On Cloud Computing

Along with some oil and gas reservoir with traditional simple structure is gradually reduced, the difficulty of oil and gas exploration is increasing. The main performance is that formation has large dip angle, reservoir is deeply buried, the velocity change of medium in vertical and horizontal direction is strongly, etc. Therefore, the requirement of calculation speed and precision of migration imaging method is also increasing. At present, prestack reverse time migration is considered as the most effective method to precise obtain the internal image of complex structures. But in the whole migration imaging process, seismic data has huge amount of computation and memory, which make it difficult to be large-scale applied in industrial production.In the migration imaging of oil-gas exploration, how to reduce the storage amount and improve the computational efficiency of the seismic data is an emphatic research problem. At present, high performance processing of seismic data is mainly realized by cluster computing method. But the cluster computing is limited by resources and construction cost is high, it is difficult to obtain a wide range of applications. In recent years, the development and maturity of cloud computing technology provides a new kind method of efficient data storage and data management for reverse time migration. This thesis analyzes the characteristics and application requirements of data processing in reverse time migration imaging. Based on the analysis of techniques and methods of data processing and cloud computing at home and abroad, a new data processing method is proposed for the inverse time migration in cloud computing environment. This method provides more extensive high-performance computing applications.First, designed data processing overall architecture model of reverse time migration in cloud computing environment, and analyzed the problem which is need to be solved in the model; Secondly, to solve these problems, carried out in-depth study for distributed storage, distributed parallel processing and task scheduling aleorithm of seismic data in cloud environment, and seek a solution to provide the technical support for the overall model; Finally, according to business requirements of seismic data migration imaging, designed prestack reverse time migration imaging application system, and applied it to the cloud environment to provide services. The main research content of this thesis is as follows:1. Proposed the data processing overall architecture model of reverse time migration in cloud computing environmentThis thesis carried out the relevant research and analysis on data processing business characteristics, application requirements and existing problems of reverse time migration; Focused researched the key technology of cloud computing architecture and data processing, and designed data processing process of reverse time migration in the cloud computing environment; Proposes an overall architecture model of reverse time migration data processing in cloud environment, and detailed described the function and the problems need to be solved of each layer in model; Proposed related technologies introduction such as distributed storage, parallel computing and task scheduling in the cloud computing environment, which realize the high-efficient storage and processing of reverse time migration data, and provide a new method for reverse time migration data processing.2. Focused researched data storage, data processing and task scheduling methods in C-RTMDPMFirstly, this paper researched the distributed storage method of seismic data in cloud environment. The method analyzed the seismic data access characteristics, introduced the distributed storage idea, considered the data storage structure, data distribution strategy and the node selection mechanism, and designed data distributed storage structure model in cloud environment under (C-RTMDCM); On the basic of model, it carried out in-depth research and application for seismic data, designed a data storage algorithm based on dynamic nodes selection, which can realize the high-efficient storage of seismic data in the cloud environment.Secondly, it researched the parallel processing method of inverse time migration data in cloud computing environment. The method analyzed the parallel processing characteristics of the reverse time migration data, and proposed a process procedure of distributed parallel computing and a parallel data processing method of on granularity partition. Using GPU computational advantages, the parallel working mode based on CPU/GPU collaborative and multi GPU card union reverse time migration computing is designed. Meanwhile, a GPU-MapReduce parallel computing model in the cloud environment is proposed, which is combined with the MapReduce programming model of cloud computing. It realized high-efficient data processing mechanism of CPU/GPU resources collaborative parallel work between the nodes in the heterogeneous cloud environment.Finally, the task scheduling optimization strategy in cloud environment is studied. The advantages and disadvantages of the commonly used scheduling optimization algorithms are analyzed, and the advantages and disadvantages of genetic algorithm and ant colony optimization are studied. The task scheduling strategy (GA-ACOHSP) is proposed. The dual advantages of genetic algorithm and ant colony optimization algorithm in solving the task scheduling optimization problem are proposed, and the implementation of large-scale task scheduling in cloud computing environment is realized.3. An empirical analysis of the processing method of inverse time migration data in cloud computing environmentBased on the reverse time migration data processing method, combined with the research results of data distributed access model, data parallel computing model and task scheduling optimization strategy, the inverse time migration system platform of seismic data is developed. Provides a visual interface for the reverse time migration data processing, and applies the platform to the cloud environment. The feasibility and effectiveness of the method are verified by real data experiments.The research results show that the proposed method can solve the problem of data storage and computation in the current seismic data processing method, and provides efficient data storage, data management and high performance services for other seismic data processing.