油氣行業(yè)對(duì)數(shù)據(jù)并不陌生。數(shù)字、智能化領(lǐng)域在非常規(guī)油氣生產(chǎn)上游行業(yè)運(yùn)用中的迅速發(fā)展促進(jìn)了千兆字節(jié)、兆字節(jié)和1015字節(jié)中產(chǎn)生的數(shù)據(jù)量（大數(shù)據(jù)）在石油行業(yè)中的應(yīng)用的急劇增加。與傳統(tǒng)油氣領(lǐng)域的發(fā)展和生產(chǎn)不同的是,常規(guī)油氣中的多相流基本規(guī)律并不適用于非常規(guī)油氣。此外,為了提高單井自然產(chǎn)能,非常規(guī)油氣開采需要采用多段壓裂和整體壓裂來提高經(jīng)濟(jì)效益。根據(jù)非常規(guī)油氣生產(chǎn)過程中的大量數(shù)據(jù)建立一個(gè)生產(chǎn)模型,利用大數(shù)據(jù)進(jìn)行分析預(yù)測(cè)和管理。這就強(qiáng)調(diào)了數(shù)據(jù)分析對(duì)石油和天然氣工業(yè)生產(chǎn)的重要性。因此,本文提出了機(jī)器學(xué)習(xí)算法（預(yù)測(cè)分析）來分析致密氣的生產(chǎn)數(shù)據(jù),建立一個(gè)分析模型,可以通過改變儲(chǔ)層/流體性質(zhì)以及氣井的水力壓裂參數(shù)來預(yù)測(cè)氣井產(chǎn)能。利用K-均值聚類理論,根據(jù)氣井原始生產(chǎn)速率,將壓裂井分為兩類。第一類分為5個(gè)等級(jí)（極好、很好、好、平均和差）,第二類分為3個(gè)等級(jí)（好、平均和差）。每個(gè)類別中的各組代表井的性能。利用人工神經(jīng)網(wǎng)絡(luò)理論來建立預(yù)測(cè)模型,并與線性模型進(jìn)行比較分析,利用均方差確定擬合程度衡量模型的預(yù)測(cè)準(zhǔn)確性,并優(yōu)選出預(yù)測(cè)模型,然后利用蒙特卡洛模擬進(jìn)行動(dòng)態(tài)分析氣井的生產(chǎn)效果并重新對(duì)氣井的生產(chǎn)等級(jí)進(jìn)行劃分。最后對(duì)... 

【文章來源】：西安石油大學(xué)陜西省

【文章頁(yè)數(shù)】：129 頁(yè)

【學(xué)位級(jí)別】：碩士

【文章目錄】：
Abstract
摘要
NOMENCLATURE
CHAPTER 1 INTRODUCTION
    1.1 Unconventional Resources and Predictive Analytics
    1.2 Unconventional Resources
    1.3 Hydraulic Fracturing
    1.4 Definition of Analytics
    1.5 Objective of Study
CHAPTER 2 BACKGROUND AND LITERATURE REVIEW
    2.1 Big Data
    2.2 Big Data Analytics
        2.2.1 Types of Analytics
            2.2.1.1 Descriptive Analytics
            2.2.1.2 Predictive Analytics
            2.2.1.3 Prescriptive Analytics
        2.2.2 Big Data Analytics Platform
    2.3 Data Warehouse/Cloud Computing
        2.3.1 Benefits of Cloud Computing
    2.4 Data Mining
        2.4.1 Types of Data
        2.4.2 Overfiting and Underfitting
        2.4.3 Noise and Attribute Importance
    2.5 Predictive Analytics in the Oil and Gas Industry
        2.5.1 Drilling and Optimization
        2.5.2 Production Optimization
        2.5.3 Reservoir and Asset Management
        2.5.4 Asset Maintenance Business Management
    2.6 K-means Clustering Algorithm
    2.7 Artificial Neural Network (ANN)
        2.7.1 ANN Transfer Function
        2.7.2 ANN Activation Function
        2.7.3 Types of Artificial Neural Network
        2.7.4 Neural Network Algorithms
            2.7.4.1 Forward Propagation
            2.7.4.2 Backpropagation
            2.7.4.3 Adaptative Learning Algorithms (Resilient Backpropagation (RPROP))
    2.8 Generalized Linear Model
    2.9 Measuring the Quality of Fit
CHAPTER 3 RESEARCH METHODOLOGY
    3.1 Data Mining
    3.2 Data Exploration
    3.3 Clustering
    3.4 Predictive Modeling
        3.4.1 Measure of Quality of Fit
        3.4.2 Key Performance Index (KPI)
        3.4.3 The Sensitivity Analysis of the Model
    3.5 The Look-back Modeling
    3.6 Software Used for the Research
CHAPTER 4 PREDICTIVE ANALYTICAL MODELS
    4.1 Exploratory Data Analysis
        4.1.1 Data Set
        4.1.2 Correlation Analysis
    4.2 K-Means Clustering Analysis
    4.3 Predictive Model Analysis
        4.3.1 Artificial Neural Network Model
            4.3.1.1 Sensitivity Analysis
            4.3.1.2 Explanation of the Sensitivity Analysis
        4.3.2 GLM Model
        4.3.3 Key Performance Index (KPI)
    4.4 Look-back Analysis
        4.4.1 Monte Carlo Simulation
CHAPTER 5 CONCLUSION AND RECOMMENDATION
    5.1 Conclusion
    5.2 Recommendation
ACKNOWLEDGEMENT
REFERENCES



本文編號(hào)：3207471