Crude oil viscosity is an important parameter which is considered in both porous media and in pipeline channels. Estimating oil viscosity at various operating conditions of pressure and temperature is of great value to petroleum engineers.
In petroleum and reservoir engineering, we are often faced with the analysis of processes which involve petroleum fluid properties, but in many cases no experimental data is available. At such cases, empirical correlations are used to estimate the oil viscosity but these correlations are either too simple or too complex and very few of these correlations are generic. Oil viscosity correlations have been barely generated for the Niger Delta region because of inappropriate coefficients obtained for the correlations. Hence, Artificial Neural Network (ANN) has been applied as the technique for re-calibrating these coefficients.
Feed-forward back propagation networks were used with a training algorithm of Levenberg-Maquardt to develop the ANN model for re-calibrating oil viscosity correlation coefficients.
A total of 350 data points were used in the development of the ANN model for oil viscosity above bubble point pressure with inputs of T, Pb, P, µob. The newly developed ANN model shows good results when compared to the existing oil viscosity correlations. The average absolute relative error for the ANN model was obtained as 5.46094 and the regression coefficient, R, was obtained as 0.9999. The input weights and biases from the ANN model serve as the re-calibrated coefficients and constants which can be used in the existing correlations to enhance applicability
TABLE OF CONTENTS
CERTIFICATION ii
DEDICATION iii
ACKNOWLEDGEMENT iv
ABSTRACT v
LIST OF FIGURES vi
LIST OF TABLES vii
LIST OF SYMBOLS viii
TABLE OF CONTENTS x
CHAPTER ONE 1
INTRODUCTION 1
1.1 Background of Study 1
1.2 Petroleum Reservoir Classifications 2
1.3 Reservoir Fluid Properties (PVT properties) 5
1.3.1 Solution Gas-Oil-Ratio (Solution GOR) 5
1.3.2 Oil Formation Volume Factor (FVF) 5
1.3.3 Bubble Point Pressure 6
1.3.4 Coefficient of Isothermal Compressibility of oil 8
1.3.5 Oil Viscosity 9
1.3.6 Gas Formation Volume Factor (FVF) 11
1.3.7 Critical Point 12
1.4 Laboratory Experiments 13
1.5 Equation of State (EoS) 14
1.6 Empirical Correlations 14
1.7 Aim and Objectives 15
1.8 Problem Statement 15
1.9 Significance of Study 16
1.10 Artificial Neural Networks (ANN) 16
1.10.1 Neurons 18
1.10.2 Weights and Biases 19
1.10.3 Transfer Functions 19
1.10.4 Layers 20
CHAPTER 2 21
LITERATURE REVIEW 21
2.1 Oil Viscosity 21
2.2 Under-saturated oil viscosity correlations applicable worldwide 22
2.2.1 Khan et al Correlations 22
2.2.2 Chew and Connally Correlation 23
2.2.3 Vasquez and Beggs Correlation 23
2.2.4 Beal’s Correlation 24
2.2.5 Ishenuwa et al Correlation 24
2.2.6 Kartoatmodjo and Schmidt Correlation 25
2.3 Review of past works relative to ANN models 25
CHAPTER THREE 27
METHODOLOGY 27
3.1 PROCEDURES 27
3.2 Performance Evaluation of Developed Network 33
CHAPTER FOUR 34
RESULTS AND DISCUSSION 34
4.1 ANN Model Development and Testing 34
4.2 Analysis of Results from ANN tool in MATLAB 35
4.3 Statistical Analysis of the ANN model 43
CHAPTER FIVE 45
CONCLUSION AND RECOMMENDATIONS 45
5.1 Conclusion 45
5.2 Recommendations 46
REFERENCES 47
