面向油田动态信息建模的PNN构建方法与应用技术研究
发布时间：2019-02-20 09:07
【摘要】：系统仿真、数据建模等信息处理方法随着科学技术的不断进步,其应用领域也在不断扩展。复杂非线性动态系统的辨识精度和建模性能的实际需求,也对系统仿真模型和信号变换机制的研究提出了更高的要求。本文面向油田勘探开发中的动态诊断、过程仿真、预测分析等典型问题,进行基于过程神经元网络的信息处理机制、数据建模方法与应用技术的研究。论文在对油田动态信息处理问题和智能数据建模方法总结分析的基础上,将其归纳为动态诊断、过程模拟、预测分析等典型应用,采用过程神经元网络方法进行建模处理。通过归纳分析油田动态系统信息汇聚和时间效应累积模式,针对不同的应用目的,提出了适合于不同典型应用的时空聚合算子、激励函数、网络模型的构建和选择策略,并对其理论性质进行了分析。对于网络结构优化和学习问题,提出了LMS算法与粒子群算法相结合、LM算法与量子遗传算法相结合的两种混合优化算法,可同时实现对过程神经元网络结构和学习性质的优化。在应用技术研究中,面向油田开发系统过程模拟和动态模式识别两个典型问题,建立了基于过程神经元网络的过程仿真模拟和动态诊断分析的实现方法和技术。课题在面向油田动态信息建模的过程神经元网络构建方法与应用技术研究结果基础上,以油田开发井组注采过程模拟和优势渗流场判别两个典型问题进行实际资料处理,取得了较好的应用结果。对于时空维信息处理方法的研究具有较大的理论和实际应用价值。
[Abstract]:With the development of science and technology, the application fields of system simulation, data modeling and other information processing methods are also expanding. The identification accuracy and modeling performance of complex nonlinear dynamic systems require higher requirements for the study of system simulation models and signal transformation mechanisms. Aiming at the typical problems of dynamic diagnosis, process simulation, prediction and analysis in oilfield exploration and development, the information processing mechanism, data modeling method and application technology based on process neural network are studied in this paper. On the basis of summarizing and analyzing the problem of oilfield dynamic information processing and the method of intelligent data modeling, this paper sums it up into typical applications such as dynamic diagnosis, process simulation, prediction and analysis, and adopts the method of process neural network to model and process. Based on the analysis of information gathering and time effect accumulation models of oilfield dynamic system, the strategies of constructing and selecting spatio-temporal aggregation operator, incentive function and network model suitable for different application purposes are put forward. The theoretical properties are analyzed. For the problem of network structure optimization and learning, two hybrid optimization algorithms, LMS algorithm and particle swarm optimization algorithm, and LM algorithm and quantum genetic algorithm are proposed, which can optimize the network structure and learning properties of process neurons at the same time. In the process simulation and dynamic pattern recognition of oilfield development system, the implementation method and technology of process simulation and dynamic diagnosis analysis based on process neural network are established. On the basis of the research results of process neural network and its application technology, two typical problems of injection and production process simulation and predominance seepage field discrimination of oilfield development well group are used to process the actual data. Good application results have been obtained. The research of spatiotemporal dimension information processing method has great theoretical and practical application value.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE331;TP18
本文编号：2427034
[Abstract]:With the development of science and technology, the application fields of system simulation, data modeling and other information processing methods are also expanding. The identification accuracy and modeling performance of complex nonlinear dynamic systems require higher requirements for the study of system simulation models and signal transformation mechanisms. Aiming at the typical problems of dynamic diagnosis, process simulation, prediction and analysis in oilfield exploration and development, the information processing mechanism, data modeling method and application technology based on process neural network are studied in this paper. On the basis of summarizing and analyzing the problem of oilfield dynamic information processing and the method of intelligent data modeling, this paper sums it up into typical applications such as dynamic diagnosis, process simulation, prediction and analysis, and adopts the method of process neural network to model and process. Based on the analysis of information gathering and time effect accumulation models of oilfield dynamic system, the strategies of constructing and selecting spatio-temporal aggregation operator, incentive function and network model suitable for different application purposes are put forward. The theoretical properties are analyzed. For the problem of network structure optimization and learning, two hybrid optimization algorithms, LMS algorithm and particle swarm optimization algorithm, and LM algorithm and quantum genetic algorithm are proposed, which can optimize the network structure and learning properties of process neurons at the same time. In the process simulation and dynamic pattern recognition of oilfield development system, the implementation method and technology of process simulation and dynamic diagnosis analysis based on process neural network are established. On the basis of the research results of process neural network and its application technology, two typical problems of injection and production process simulation and predominance seepage field discrimination of oilfield development well group are used to process the actual data. Good application results have been obtained. The research of spatiotemporal dimension information processing method has great theoretical and practical application value.
【学位授予单位】：东北石油大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TE331;TP18
【参考文献】
相关期刊论文 前3条
1 钟诗胜;朴树学;丁刚;;改进BP算法在过程神经网络中的应用[J];哈尔滨工业大学学报;2006年06期
2 许少华;何新贵;周继;王兵;;一种过程支持向量机及其在动态模式分类中的应用[J];控制与决策;2009年02期
3 许少华;李盼池;何新贵;;一种概率过程神经元网络模型及分类算法[J];智能系统学报;2009年04期
，本文编号：2427034
本文链接：https://www.wllwen.com/kejilunwen/shiyounenyuanlunwen/2427034.html
