Applications of Artificial Intelligence Techniques in the Petroleum Industry

Abdolhossein Hemmati-Sarapardeh,Aydin Larestani,Nait Amar Menad,Sassan Hajirezaie

• Publisher: Gulf Professional Publishing
• Year: 2020
• ISBN:
  9780128223857
• Topic: Scienze fisiche
• Subtopic: Energia

Applications of Artificial Intelligence Techniques in the Petroleum Industry gives engineers a critical resource to help them understand the machine learning that will solve specific engineering challenges. The reference begins with fundamentals, covering preprocessing of data, types of intelligent models, and training and optimization algorithms. The book moves on to methodically address artificial intelligence technology and applications by the upstream sector, covering exploration, drilling, reservoir and production engineering. Final sections cover current gaps and future challenges.

• Teaches how to apply machine learning algorithms that work best in exploration, drilling, reservoir or production engineering
• Helps readers increase their existing knowledge on intelligent data modeling, machine learning and artificial intelligence, with foundational chapters covering the preprocessing of data and training on algorithms
• Provides tactics on how to cover complex projects such as shale gas, tight oils, and other types of unconventional reservoirs with more advanced model input

Chapter 1

Introduction

Abstract

In this chapter, the main statistical and graphical approaches used to analyze the performance of artificial predictive models in the oil and gas industry are described. In addition, data preprocessing steps as necessary steps to refine a data bank before performing statistical and graphical error analyses are presented. These processes eliminate unreliable data points to ensure the development of more accurate predictive models. These unreliable data points could be false data or outliers that should be removed from a dataset. Error analysis techniques are used to evaluate the performance and accuracy of predictive models. The basis of these techniques is measuring the deviation of predictions from the measured data points using different mathematical formulations. Graphical error analyses, on the other hand, are used to enable an easier way to compare the performance of multiple predictive models and select the most accurate one. These techniques are a representation of the outcome from statistical techniques and are used to facilitate the process of model selection. In this chapter, first, a brief description of different procedures for data analysis is provided, and then some examples of these procedures are given. The definitions and formulations of statistical error analyses are presented along with specific examples using data from real oil and gas operations. In addition, different graphical error analyses with graphical examples are presented. Dew point pressure was chosen as a candidate parameter to describe these statistical and graphical methods better.

Keywords

Data processing; data cleaning; data integration; applicability domain; sensitivity analysis

Contents

Outline

• 1.1 Overview 1
• 1.2 Preprocessing of data 2
  • 1.2.1 Data cleaning 2
  • 1.2.2 Data integration 2
  • 1.2.3 Data transformation 3
  • 1.2.4 Data reduction 3
  • 1.2.5 Data discretization 4
  • 1.2.6 Data statistics 4
• 1.3 Processing of data 5
  • 1.3.1 Data training 5
  • 1.3.2 Data validation and testing 5
• 1.4 Postprocessing of data 6
  • 1.4.1 Statistical analyses for models’ evaluation 6
  • 1.4.2 Graphical error analysis for models’ evaluation 9
• 1.5 Applicability domain of a model 19
  • 1.5.1 Identification of experimental data outliers 19
• 1.6 Sensitivity analysis on models’ inputs 21
  • 1.6.1 Relevancy factor analysis 21
• 1.7 The areas of intelligent models applications in the petroleum industry 21
• References 22

1.1 Overview

In this chapter, the main statistical and graphical approaches used to analyze the performance of artificial predictive models in the oil and gas industry are described. In addition, data preprocessing steps, as necessary steps to refine a data bank before performing statistical and graphical error analyses, are presented. These processes eliminate unreliable data points to ensure the development of more accurate predictive models. These unreliable data points could be false data or outliers that should be removed from a dataset. Error analysis techniques are used to evaluate the performance and accuracy of predictive models. The basis of these techniques is measuring the deviation of predictions from the measured data points using different mathematical formulations. Graphical error analyses, on the other hand, are used to enable an easier way to compare the performance of multiple predictive models and select the most accurate one. These techniques are a representation of the outcome from statistical techniques and are used to facilitate the process of model selection. In this chapter, first, a brief description of diffe...