Improved Oil Production and Waterflood Performance by Water Allocation Management

Improved Oil Production and
Waterflood Performance by Water

Allocation Management
Yousef Rafiei
A thesis submitted for the degree of Doctor of Philosophy
Institute of Petroleum Engineering
Heriot-Watt University
Edinburgh, Scotland, UK
March 2014
The copyright in this thesis is owned by the author. Any quotation from the
thesis or use of any of the information contained in it must acknowledge
this thesis as the source of the quotation or information.
ABSTRACT
This thesis has evaluated a wide range of techniques to militate one of the big
challenges of petroleum industry, water production. The techniques discussed
waterflooding management, all aim to reduce excessive water production. The injection
and production history at a well and field level are most common available data in any
oil field, especially when nowadays we can have these data in real time with the
implementation of the digital oil field and the intelligent well completion. This research
aims to understand the strength and weaknesses of the existing techniques and
“repackage” them to provide an optimum combination for more effective waterflood
management by analysing injection and production data history.
The first part of this research reviewed, tested and compared the analytical techniques
that have been previously used for analysing the injection and production. The methods
studied fell in to two distinct classes: those that monitor the waterflood performance
secondly, methods for determining the inter-well connectivity.
The second part of this thesis showed that an improved workflow used the captured
information from the phase one methods could be combined to give more effective
waterflood management via combination of reservoir voidage management (RVM),
water allocation management (WAM) and production allocation management (PAM).
Finally, a semi-analytical method was introduced in this thesis for performing RVM.
Two approaches were defined for WAM and new techniques developed for PAM, all of
which employed only the production and injection history. The results from these
techniques were compared with the more advanced reservoir simulation methodologies
such as gradient free optimisation. This comparison showed the reliability of the
proposed techniques.
DEDICATION
I dedicate this thesis to my wife and my parents for their continuous

encouragement and love.
Acknowledgement
Foremost, I would like to express my gratitude to my supervisor Prof. David. R.

Davies for the continuous support of my study and research, for his patience,
motivation, enthusiasm, and immense knowledge. His guidance helped me throughout
all the time of the research and writing of this thesis. The good advice, support and
friendship of my second supervisor, Dr. Khafiz Muradov, has been invaluable, on both
an academic and a personal level, for which I am extremely grateful.
I would like to thank my internal and external examiners Dr. James M. Somerville and
Prof. Ian Main for their encouragement, insightful comments, and hard questions.
I am very grateful to Steve Toddman for his valuable information about IPM softwares
and providing me with good example of reservoir simulation models. I would like to
thank Dr. A. Kuznetsov for his great comments and support during my study.
I would also like to offer my best regards and blessings to all of those who supported
me in my research: Morteza Haghighat Sefat, Dr. Ivan Grebenkin, Dr. Hamidreza
Shahverdi, Dr. Alireza Emadi, Dr. Alireza Kazemi, Dr. Hamidreza Hamdi, Amir
Jahanbakhsh, Mohammad Hoseinipour and all the members of the IWsT research
group.
In addition, a thank you to Dr. Ali Vatani, his advice on both research as well as on my
career have been priceless.
I would like to acknowledge the financial, academic and computer support of the
Institute of Petroleum Engineering of Heriot-Watt University and its staff that
provided the necessary support for this research.
I also want to thank to the Ministry of Research, Science and Technology of Islamic
Republic of Iran for their financial support granted through my study.
Last but not least I would like to express my love and gratitude to my beloved wife
Zainab Naeim Abadi and my parents for their understanding and endless love
throughout my research.
TABLE OF CONTENTS
TABLE OF CONTENTS............................................................................................. i
LISTS OF TABLES AND FIGURES...................................................................... vii
LIST OF ACRONYMS ............................................................................................ xx
NOMENCLATURE ................................................................................................. xxi
Chapter 1 – Introduction ................................................................................................... 1
1.1 Water production ................................................................................................ 1
1.2 Waterflooding management ............................................................................... 3
1.3 Intelligent Well Completion (IWC) and waterflooding management ................ 4
1.4 Research objectives ............................................................................................ 6
References ..................................................................................................................... 8
Chapter 2 - Monitoring The Performance Of Waterflooding ........................................... 9
2.1 Monitoring and surveillance of water injection projects .................................... 9
2.1.1 Semi-log plot of oil cut or water cut versus cumulative oil production (cut-
cum plot) ................................................................................................................. 10
2.1.2 Semi-log plot of Water oil ratio (WOR) versus cumulative oil production
10
2.1.3 The X-Plot ................................................................................................. 12
2.1.4 The Hall Plot ............................................................................................. 15
2.2 Case study......................................................................................................... 17
2.2.1 Model description ..................................................................................... 18
2.2.1.1 Base case scenario ............................................................................. 20

i
2.2.1.2 WAM scenario ................................................................................... 20
2.2.2 Plot Analysis ............................................................................................. 20
2.2.2.1 Cut-Cum plot ..................................................................................... 22
2.2.2.2 WOR plot ........................................................................................... 26
2.2.2.3 X-plot ................................................................................................. 30
2.2.2.4 Hall plot ............................................................................................. 32
2.3 Reservoir voidage management ....................................................................... 35
2.3.1 Model description ..................................................................................... 35
2.3.2 Methodology for reservoir voidage management ..................................... 36
2.4 Results and discussion ...................................................................................... 44
2.5 References ........................................................................................................ 46
Chapter 3 – Producer-Injector Inter-well Connectivity Measurement; Statistical

Approach ......................................................................................................................... 48
3.1 Producer-injector inter-well connectivity ......................................................... 48
3.2 Inter-connectivity measurement ....................................................................... 49
3.2.1 Spearman rank correlation coefficient ...................................................... 49
3.2.2 Multi-linear regression (MLR).................................................................. 50
3.2.3 Capacitive resistive model (CRM) ............................................................ 52
3.3 Case study......................................................................................................... 58
3.3.1 Spear Rank correlation coefficient ............................................................ 60
3.3.2 Multi-linear regression .............................................................................. 62
ii
3.3.3 Capacitive resistive model ........................................................................ 68
3.4 Allocation management .................................................................................... 73
3.5 Results and discussion ...................................................................................... 76
3.6 References ........................................................................................................ 79
Chapter 4 – Producer-Injector Inter-well Connectivity Measurement; Artificial

Intelligence Techniques .................................................................................................. 81
4.1 Artificial neural networks (ANNs) .................................................................... 81
4.1.1 Structure of ANNs .................................................................................... 82
4.1.2 Classification of ANNs ............................................................................. 83
4.1.3 Application of ANN .................................................................................. 84
4.1.3.1 Pattern classification .......................................................................... 84
4.1.3.2 Clustering ........................................................................................... 84
4.1.3.3 Function approximation (modelling) ................................................. 85
4.1.3.4 Forecasting ......................................................................................... 86
4.1.4 Application of ANN in petroleum engineering......................................... 86
4.2 ANN and inter-well connectivity measurement ............................................... 87
4.2.1 Feed-forward back propagation (FFBP) ................................................... 88
4.2.2 Fuzzy logic network (Co-adoptive neuro-fuzzy inference system

CANFIS) ................................................................................................................. 88
4.3 Case Study ........................................................................................................ 88
4.3.1 Feed-forward back propagation (FFBP) ................................................... 89
iii
4.3.1.1 Designing the optimum network ....................................................... 89
4.3.1.2 Inter-well connectivity analysis ......................................................... 97
4.3.2 Co-Active Neuro-Fuzzy Inference System (CANFIS) ........................... 100
4.3.3 Determination of water allocation factor ................................................ 101
4.4 Results and discussion .................................................................................... 102
4.5 References ...................................................................................................... 104
Chapter 5 – New Allocation Management Methodologies ........................................... 105
5.1 Water allocation management (WAM) ........................................................... 105
5.1.1 Water Cut ................................................................................................ 105
5.1.2 Cumulative Water Cut (CWC) ................................................................ 110
5.1.3 Relative Oil Production Ratio (OPR)...................................................... 114
5.1.3.1 All AFt have positive values ............................................................ 115
5.1.3.2 All AFt have negative values ........................................................... 115
5.1.3.3 Both positive (𝑨𝑭𝒕+) and negative (𝑨𝑭𝒕−) values of AFt are present
........................................................................................................................... 115
5.1.4 Oil Production Index (OPI) ..................................................................... 119
5.2 Production Allocation Management (PAM) ................................................... 124
5.2.1 Oil ratio (OR) .......................................................................................... 124
5.2.2 Relative oil production ratio (OPR): ....................................................... 125
5.2.2.1 All OPR values are positive ............................................................. 125
5.2.2.2 All OPR values are negatives .......................................................... 125
iv
5.2.2.3 Both positive (𝑶𝑷𝑹𝒋+) and negative (𝑶𝑷𝑹𝒋𝒕−) values of OPR are
present 126
5.2.3 Oil production index (OPI) ..................................................................... 127
5.3 Production and injection allocation management (W&P AM) ....................... 130
5.4 References ...................................................................................................... 135
Chapter 6 – Water Allocation Optimisation ................................................................. 136
6.1 Introduction .................................................................................................... 136
6.2 MLR review .................................................................................................... 136
6.3 CRM Review .................................................................................................. 140
6.4 Linear Programming (LP) .............................................................................. 143
6.5 Case study....................................................................................................... 143
6.6 New optimisation using Genetic Algorithm (GA) and reservoir simulator.... 150
6.7 References ...................................................................................................... 154
Chapter 7 – Waterflooding Management: New Case Study ......................................... 156
7.1 Model description ........................................................................................... 156
7.2 Reservoir voidage management (RVM).......................................................... 158
7.3 Production allocation management (PAM) .................................................... 160
7.4 Water allocation management (WAM) ........................................................... 161
7.5 Water flooding management (WFM).............................................................. 163
7.6 Results and Discussions ................................................................................. 163
Chapter 8 – Conclusions and Recommendations .......................................................... 167
v
8.1 Research story ................................................................................................ 167
8.2 Monitoring the performance of waterflooding ............................................... 168
8.2.1 Summary ................................................................................................. 168
8.2.2 Conclusions ............................................................................................. 168
8.3 Inter-well connectivity measurement: Statistical techniques ......................... 169
8.3.1 Summary ................................................................................................. 169
8.3.2 Conclusions ............................................................................................. 170
8.4 Inter-well connectivity measurement: Artificial intelligence techniques....... 171
8.4.1 Summary ................................................................................................. 171
8.4.2 Conclusion .............................................................................................. 171
8.5 New development in water allocation management (WAM) and production

allocation management (PAM).................................................................................. 171
8.5.1 Summary ................................................................................................. 171
8.5.2 Conclusions ............................................................................................. 172
8.6 Water allocation optimisation......................................................................... 173
8.6.1 Summary ................................................................................................. 173
8.6.2 Conclusions ............................................................................................. 174
8.7 Water flooding management: new case study: ............................................... 174
8.7.1 Summary ................................................................................................. 174
8.7.2 Conclusions ............................................................................................. 175
8.8 Recommendations for future works: .............................................................. 175
vi
LISTS OF TABLES AND FIGURES
Figure 1. 1: Waterflood performance at different periods [7]. ......................................... 2
Figure 1. 2: Different components of an intelligent well completion. .............................. 5
Figure 1. 3: Closed loop reservoir management by IWC.................................................. 6
Figure 2. 1: An example of the plot of water cut versus cumulative oil production [8]. 11
Figure 2. 2: Typical example of WOR plot for poor, average and good waterflood
performances[13]. ........................................................................................................... 12
Figure 2. 3 An example of plot of X values versus Recovery obtained from Equation

2.2 [17]. ........................................................................................................................... 15
Figure 2. 4: Typical Hall plot for various injection well conditions[2]. ......................... 17
Figure 2. 5: Location of the wells in the reservoir (colours represent the reservoir
porosity). ......................................................................................................................... 19
Figure 2. 6: Plot of oil recovery versus time for the base injection scenario and the
WAM injection scenario. ................................................................................................ 21
Figure 2. 7: Plot of water cut versus time for the base injection scenario and the WAM
injection scenario. ........................................................................................................... 21
Figure 2. 8: Semi log plot of water cut versus cumulative oil production plot for base
injection scenario and the WAM injection scenario. ...................................................... 22
Figure 2. 9: Extrapolation of cut-cum plot for estimating future performance by

employing first, second and third order polynomials...................................................... 23
Figure 2. 10: Second order polynomial fitted for water cut values higher that 70%. ..... 24
Figure 2. 11: Curve fitting of cut-cum plot for the WAM scenario when all data points
are used. .......................................................................................................................... 25
vii
Figure 2. 12: Curve fitting of cut-cum plot for WAM scenario for data points after start
of injection management. ................................................................................................ 26
Figure 2. 13: WOR plots for base case and WAM injection schemes. ........................... 27
Figure 2. 14: WOR plot curve fitting for base case scenario when all data points are
used and when WOR values higher than of 1 are employed. ......................................... 28
Figure 2. 15: WOR plot curve fitting for WAM injection scenario. ............................... 29
Figure 2. 16: Plot of oil recovery versus X values (X–plots) for base case and WAM
injection schemes. ........................................................................................................... 30
Figure 2. 17: X-plot curve fitting for base case scenario for all data points and for fw
higher than 65%. ............................................................................................................. 31
Figure 2. 18: X-plot curve fitting for WAM scenario for all data points and for fw higher
than 65%. ........................................................................................................................ 31
Figure 2. 19: Determined Hall plots of each injection well for base case injection
scenario. .......................................................................................................................... 32
Figure 2. 20: Determined Hall plots of each injection well for WAM injection scenario.
......................................................................................................................................... 33
Figure 2. 21: Comparison of the Hall plot of each injector in base case and WAM
injection scenarios. .......................................................................................................... 34
Figure 2. 22: Schematic of the reservoir model (colours representing oil saturation). ... 35
Figure 2. 23: Oil production rate versus the time for injection with different VRs. ...... 36
Figure 2. 24: Oil recovery versus time, obtained from injection with different VRs. .... 37
Figure 2. 25: Cumulative water production from injection with different VRs.............. 37
Figure 2. 26: Change in the reservoir pressure versus the time during injection with
different VRs. .................................................................................................................. 38
viii
Figure 2. 27: Sensitivity analysis on the well outflow performance to determine the
minimum required reservoir pressure. ............................................................................ 39
Figure 2. 28: Calculated breakthrough time for different starting injection VRs. .......... 41
Figure 2. 29: Oil production rate versus time for base case and voidage management
Figure 2. 30: Plot of oil recovery versus time for base case and voidage management
Figure 2. 31: Plot of cumulative water production versus time for base case and voidage
management injection scenarios. .................................................................................... 43
Figure 2. 32: Cumulative injected water versus time for base case and voidage
management injection scenarios. .................................................................................... 43
Figure 3. 1: An example of calculated inter-well connectivity between injectors (blue

circles with arrows) and producers (black circles) [3]. ................................................... 48
Figure 3. 2: Schematic representation of control volume for producer j, CRMP [21]. .. 55
Figure 3. 3: Schematic representation of a field with one injector and one producer,
CRMT [21]. ..................................................................................................................... 56
Figure 3. 4: Schematic representation of control volume between each injector/producer

pair [21]. .......................................................................................................................... 58
Figure 3. 5: Reservoir model (colours represent the porosity)........................................ 58
Figure 3. 6: 20-year down-hole injection rate history of all the injectors from the
reservoir model................................................................................................................ 59
Figure 3. 7: 20-year down-hole production rate history of all the producers from the
reservoir model................................................................................................................ 60
ix
Figure 3. 8: Calculated Spearman correlation coefficients for each injector and its
associated producers. ...................................................................................................... 61
Figure 3. 9: Calculated Spearman correlation coefficients for each producer and its
associated injectors.......................................................................................................... 61
Figure 3. 10: Calculated inter-well connectivity for Producer 1, assuming two different
injector-producer combinations. ..................................................................................... 63
Figure 3. 11: Calculated error of estimation of production rate of Producer 1 for

different combinations of injectors. ................................................................................ 63
Figure 3. 12: Calculated estimation error of the production rate of Producer 2 for
Figure 3. 13: Calculated estimation error of the production rate of Producer 3 for
Figure 3. 14: Determined MLR coefficients for the injectors connected to Producer 1.
......................................................................................................................................... 65
......................................................................................................................................... 66
......................................................................................................................................... 66
Figure 3. 17: Calculated production rate from MLR versus production history from the
reservoir simulation model for Producer 1..................................................................... 67
Figure 3. 18: Calculated production rate from MLR versus production history from the
reservoir simulation model for Producer 2..................................................................... 67
Figure 3. 19: Calculated values of fij of the CRM equation for Producer 1. .................. 69
Figure 3. 20: Calculated values of fij of the CRM equation for Producer 2. .................. 69
Figure 3. 21: Calculated values of fij of the CRM equation for Producer 3. ................. 70
x
Figure 3. 22: Measured values of time constant for each producer. ............................... 70
Figure 3. 23: Drainage volume of each producer calculated from reservoir simulator. . 71
Figure 3. 24: Calculated production rate from CRMP versus production history of
Producer 1. ...................................................................................................................... 72
Figure 3. 25: Calculated production rate from CRMP versus production history of
Producer 2. ...................................................................................................................... 72
Figure 3. 26: Calculated production rate from CRMP versus results production history
of Producer 3. .................................................................................................................. 73
Figure 3. 27: WC of each producer at the end of 20 years of production. ...................... 74
Figure 3. 28: Calculated LR of each producer at the end of 20 years of production. ..... 74
Figure 3. 29: Calculated OI of each producer at the end of 20 years of production. ...... 75
Figure 3. 30: Determined allocation factor for each injection well from different
statistical methods. .......................................................................................................... 76
Figure 3. 31: Comparison of improvement in oil production rate after WAM based on
different statistical techniques. ........................................................................................ 76
Figure 3. 32: Comparison of improvement in cumulative oil production after WAM

based on different statistical techniques.......................................................................... 77
Figure 3. 33: Comparison of reduction in cumulative water production after WAM

based on different statistical techniques.......................................................................... 77
Figure 4. 1: Structural model of an artificial neuron[7]. ................................................. 83
Figure 4. 2: An example of pattern classification by ANN [2]. ...................................... 84
Figure 4. 3: An example of clustering by ANN[2]. ........................................................ 85
xi
Figure 4. 4: An example of function approximation by ANN[2]. .................................. 85
Figure 4. 5: An example of forecasting by ANN [2]. ..................................................... 86
Figure 4. 6: Testing results for the trained network for producer 1 (L1 is the actual
liquid rate for producer 1 from reservoir simulation and L1 output is the estimated
values from the network) when the first 80% portion of the data is used for training... 91
Figure 4. 7: Testing results for the trained network for producer 1 (L1 is the actual
liquid rate for producer 1 from reservoir simulation and L1 output is the estimated
values from the network) when the last 80% portion of data is used for training. ......... 92
Figure 4.8: Testing results for the trained network for producer 1 (L1 is the actual liquid
rate for producer 1 from reservoir simulation and L1 output is the estimated values from
the network) when the first 40% portion and the last 40% portion of the data are used
for training....................................................................................................................... 92
Figure 4. 9: Histogram of input data (liquid production rate) of producer 1. ................. 93
Figure 4. 10: Test results of the optimum trained network, where L1 is the actual liquid
rate from reservoir simulation and L1 output is the estimated results from the network.
......................................................................................................................................... 94
Figure 4. 11: Mean absolute error of the estimated production rate obtained from
designed networks based on different combinations of injectors with producer 1. ........ 95
Figure 4. 12: Sensitivity analysis on the number of hidden layers shows that by
increasing number of hidden layers, error of estimation will increase too. .................... 96
Figure 4. 13: Sensitivity analysis on the number of processing elements. ..................... 96
Figure 4. 14: Sensitivity analysis on the number of iterations for training the network. 97
Figure 4. 15: Optimum network test results , where L1 is the actual liquid rate from
reservoir simulation and L1 output is the estimated result from the network................. 98
xii
Figure 4. 16: Inter-well connectivity analysis for producer 1 (colours represent the
change in liquid production of the producer by change in the rate of associated injector
and numbers are the injection well numbers). ................................................................ 98
Figure 4. 17: Inter-well connectivity analysis for producer 2 (colours represent the
change in liquid production of the producer by change in the rate of the associated
injector). .......................................................................................................................... 99
Figure 4. 18: Inter-well connectivity analysis for producer 3 (colors represent the
change in liquid production of the producer by change in the rate of associated injector).
......................................................................................................................................... 99
Figure 4. 19: Results of inter-well connectivity measurements for all producers by

FFBP. ............................................................................................................................ 100
Figure 4. 20: Results of connectivity measurement for all producers from CANFIS... 101
Figure 4. 21: Calculated allocation factors for both networks. ..................................... 101
Figure 4. 22: Cumulative oil production obtained from WAM, based on FFBP, CANFIS
and CRM. ...................................................................................................................... 102
Figure 4. 23: Comparison of oil production rate from WAM, based on FFBP, CANFIS
and CRM. ...................................................................................................................... 102
Figure 4. 24: Cumulative water production from WAM, based on FFBP, CANFIS and
CRM. ............................................................................................................................. 103
Figure 5. 1: Water cut of each producer at the end of each 5-year interval after
beginning of the water allocation management. ........................................................... 106
Figure 5. 2: Calculated water allocation factor for each injector at the end of each 5 year
period. ........................................................................................................................... 107
Figure 5. 3: Inter-well connectivity measurements determined from CRM for each

producer. ....................................................................................................................... 107
xiii
Figure 5. 4: Reservoir oil production rate for base case and the case with water
allocation management.................................................................................................. 108
Figure 5. 5: Cumulative oil production for base case and the case with water allocation
management. ................................................................................................................. 108
Figure 5. 6: Field water cut for base case and the case with water allocation
management. ................................................................................................................. 109
Figure 5. 7: Reservoir cumulative water production for base case and the case with
water allocation management........................................................................................ 109
Figure 5. 8: Cumulative water cut of each producer. .................................................... 111
Figure 5. 9: Calculated water allocation factor for different production periods.......... 111
Figure 5.10: Reservoir oil production rate for all three injection scenarios.................. 112
Figure 5.11: Reservoir cumulative oil production for all three injection scenarios. ..... 112
Figure 5. 12: Reservoir water cut for all three injection scenarios. .............................. 113
Figure 5.13: Field cumulative water production for all three injection scenarios. ....... 113
Figure 5. 14: Calculated water allocation factor based on OPR at the end of each 5-year
period. ........................................................................................................................... 116
Figure 5. 15: Measured OPR for each producer at the end of each 5-year period. ....... 117
Figure 5. 16: Comparison of daily oil production rate obtained from different allocation
management scenarios. ................................................................................................. 117
Figure 5. 17: Comparison of cumulative oil production obtained from different

allocation management scenarios.................................................................................. 118
Figure 5. 18: Comparison of field water cut obtained from different allocation
xiv
Figure 5. 19: Comparison of cumulative water production obtained from different
Figure 5. 20: Reservoir containing three production wells. .......................................... 120
Figure 5. 21: Measured OPI for reservoir and each production well. ........................... 121
Figure 5. 22: Calculated water allocation factor, based on OPI at the end of each 5-year
period. ........................................................................................................................... 121
Figure 5. 23: Comparison of daily oil production rate obtained from different allocation
Figure 5. 24: Comparison of cumulative oil production obtained from different

Figure 5. 25: Comparison of field water cut obtained from different allocation
Figure 5. 26: Waterflooding performance in terms of cumulative oil and water

production for different water allocation procedures. ................................................... 124
Figure 5. 27: Oil production rate obtained from different techniques of production
Figure 5. 28: Cumulative oil production obtained from different techniques of

production allocation management. .............................................................................. 129
Figure 5. 29: Produced water rate obtained from different techniques of production
Figure 5. 30: Cumulative water produced from different techniques of production

Figure 5.31: Comparison of oil production rate obtained from WAM, PAM and W&P
AM. ............................................................................................................................... 131
Figure 5. 32: Cumulative oil production from WAM, PAM and W&P AM. ............... 131
xv
Figure 5. 33: Water production rate from WAM, PAM and W&P AM. ...................... 132
Figure 5. 34: Cumulative water production obtained from WAM, PAM and W&P AM.
....................................................................................................................................... 132
Figure 5. 35: Bar chart of cumulative oil and water production from WAM, PAM and
W&P AM. ..................................................................................................................... 133
Figure 5. 1: Plot of oil recovery rate versus time for different WAM scenarios. ......... 152
Figure 5. 2: Plot of cumulative water production versus time for different WAM
scenarios. ....................................................................................................................... 153
Figure 6. 1: Oil production rate for base case and WAM by LP. ................................. 144
Figure 6. 2: Cumulative water production for base case scenario and WAM by LP. ... 145
Figure 6. 3: Oil recovery versus time for base case scenario and WAM by LP. .......... 145
Figure 6. 4: Water cut development of the 3 production wells for WAM scenario by LP.
....................................................................................................................................... 146
Figure 6. 5: Calculated water allocation factor at the end of each 5-year period. ........ 146
Figure 6. 6: Calculated allocation factor for each period based on optimizing cumulative
oil production as objective function. ............................................................................. 147
Figure 6. 7: Oil recovery versus time for base case, WAM by LP and WC and LP and
COC. ............................................................................................................................. 148
Figure 6. 8: Cumulative water production versus time for base case, WAM by LP and
WC and LP and COC. ................................................................................................... 149
Figure 6. 9: Change in producer’s water cut in case of WAM with LP and COC. ....... 149
Figure 6. 10: GA solution development over 50 iterations. .......................................... 151
xvi
Figure 6. 11: Optimum water allocation factor obtained from GA for each injection
well. ............................................................................................................................... 151
Figure 6. 12: Plot of oil production rate versus time for different WAM scenarios. .... 152
Figure 7. 1: Permeability variation in the reservoir model. .......................................... 157
Figure 7. 2: Porosity variation in the reservoir model. ................................................. 157
Figure 7.3: Well outflow sensitivity analysis shows that the minimum required reservoir
pressure is 2800 psi. ...................................................................................................... 159
Figure 7. 4: Calculated OPR for all production wells. .................................................. 160
Figure 7.5: Old and new values of PAF for each producer........................................... 160
Figure 7. 6: Calculated production rate from CRM versus production history of

Producer 1 from reservoir model. ................................................................................. 161


Figure 7.9: Calculated production rate from CRM versus production history of Producer
8 from reservoir model. ................................................................................................. 162
Figure 7. 10: Calculated coefficient of CRM for all production wells. ........................ 163
Figure 7. 11: New water allocation factor for injection wells obtained from LP-CRM
technique. ...................................................................................................................... 163
Figure 7. 12: Comparison of oil recovery obtained from different injection scenarios.
....................................................................................................................................... 164
xvii
Figure 7. 13: Comparison of oil production rate obtained from different injection
scenarios. ....................................................................................................................... 165
Figure 7. 14: Comparison of cumulative water production obtained from different

injection scenarios. ........................................................................................................ 165
Table 2. 1: The properties of the studied reservoir model. ............................................. 18
Table 2. 2: Allocated liquid production rate of each producer. ...................................... 19
Table 2. 3: Results of estimated future water cut for base case scenario from cut-cum
plot extrapolation for cumulative oil production of 2.23 × 108 STB. .............................. 24
Table 2. 4: Results of estimated future water cut for WAM injection scenario
extrapolated to cumulative oil production of 2.38 × 108 STB. ........................................ 26
Table 2. 5: Estimated future WOR for cumulative oil production of 2.23 × 108 STB for
base case, from WOR plot. ............................................................................................. 28
Table 2. 6: Estimated future WOR for cumulative oil production of 2.38 × 108 STB for
WAM injection scheme................................................................................................... 29
Table 2. 7: Estimated future oil recovery from extrapolation results of X-plot at X value
of 2.7 for base case. ......................................................................................................... 31
Table 2. 8: Estimated future oil recovery from extrapolation results of X-plot at X value
of 2.7 for WAM scenario. ............................................................................................... 31
Table 2. 9: Determination of a and b constants.............................................................. 32
Table 2. 10: Calculated injectivity index from Hall plots for each injection well. ......... 34
Table 2. 11: Properties of the studied reservoir model. .................................................. 36
Table 2. 12: Calculated tpm for each injection volume of water, in terms of different
VRs.................................................................................................................................. 40
xviii
Table 3. 1: Properties of the reservoir model………………….……………………….59
Table 3. 2: This table shows which injectors are connected to each producer (√ means
connected)………………………………….…………………………..,………………64
Table 4. 1 Initial properties of the FFBP network designed to estimate liquid production
rate of producer number one. .......................................................................................... 90
Table 4. 2: Results of connectivity analysis for producer 1. ........................................... 95
Table 4. 3: Optimum properties of the desired network for estimatiing liquid production
rate of producer 1. ........................................................................................................... 97
Table 4. 5: Properties of the CANFIS Network. ........................................................... 100
Table 7. 1: Reservoir model properties ......................................................................... 156
Table 7. 2: Allocated liquid production rate of each production well for base case
scenario. ........................................................................................................................ 158
Table 7. 3: Calculated tpm and Qipm for each VR. ...................................................... 159
xix
LIST OF ACRONYMS
AF allocation factor
ANNs Artificial Neural Networks
BHP bottom hole pressure
CANFIS Co-adoptive neuro-fuzzy inference system
COC Cumulative oil cut
COR Cumulative oil ratio
CRM capacitive resistive mode
CRMIP Applied CRM for drainage volume between each injector-producer pair
CRMP Applied CRM for drainage volume of each producer
CRMT Applied CRM for volume of the entire field or tank model
CWC Cumulative water cut
FFBP Feed-forward back propagation network
GA Genetic algorithm
ICV inflow control valves
IWC Intelligent Well Completion
LP Linear programming
LR liquid production ratio
MLR Multi-Linear Regression
OI oil production performance index
OOIP original oil in place
OPI oil production index
OPR Relative oil production ratio
OR Oil Ratio
PAM Production allocation management
RVM Reservoir voidage management
PE Processing elements
VR Voidage ratio
WAM water allocation management
WC water cut
WFM Waterflood management
WOR Water oil ratio
WR Water ratio
xx
NOMENCLATURE
βij MLR weighting factors

β0j unbalance constant for unbalanced MLR
Z transfer function
Vpr reservoir pore volume
Vp drainage pore volume
tpm time of starting pressure maintenance by water injection
t2 the duration time of injection with the second VR
t production time
Swc connate water saturation
Sw water saturation
Sav average water saturation
rw well-bore radius
rs Spearman rank correlation coefficient
re well influence zone radius
Qwr reservoir’s cumulative water production
Qwibt2 injected cumulative water volume before water break through by VR2
Qwibt1 injected cumulative water volume before water break through by VR1
Qwibt injected cumulative water volume before water break through
Qw cumulative water production
Qor reservoir’s cumulative oil production
Qo cumulative oil production
ql liquid production rate
ql liquid production rate
Ql cumulative liquid production
qi injection rate
qi injection rate
pwf flowing bottom hole pressure
Prmin minimum reservoir pressure
Pr initial reservoir pressure
Pi injection pressure
Pe reservoir pressure
PAFj production allocation factor for the producer j
xxi
ORj oil ratio for the producer j
OPIrj difference between producer OPIj and reservoir OPIr
n number of observations
N total number of producers
k formation permeability
J productivity index
I total number of injectors
II Injectivity index
H formation thickness
fw fractional flow of water
fij CRM inter-well connectivity between injector I and producer j
ER Oil recovery
dV change in reservoir volume
dP change in reservoir pressure
di difference between the rankings of the ith observations
Ct Reservoir total compressibility
C compressibility factor
Bo Oil formation volume factor
AFti total water allocation index for injector i
AFni normalized allocation factor for the injector i
a and b permeability ratio constants
µ injected fluid viscosity
krw
relative permeability ratio of water to oil
kro
µo
viscosity ratio of oil to water
µw
q j (t ) production rate of producer j
ii (t ) injection rate of injector i

−
qj average production rate
−
ii average injection rate
µj Lagrange multiplier
−
P average reservoir pressure
xxii
τ CRM time constant
I ijk injection rate of injector i at time interval k in CRM equation
∆Pwf( k,)j changes in BHP of producer j during the time interval t k-1 to t k
AFni+ production wells with positive AFni
AFni− production wells with negative AFni
OPR −j wells with negative OPR
OPR +j wells with Positive OPR
OPI rj+ wells with Positive OPIrj
OPI rj− wells with negative OPIrj
qoj− old allocated production rate correspond to wells with negative OPR or OPI
qnj− new allocated production rate correspond to wells with negative OPR or OPI
qoj+ old allocated production rate correspond to wells with positive OPR or OPI
qnj+ new allocated production rate correspond to wells with positive OPR or OPI
xxiii
Chapter 1– Introduction
Primary recovery methods using natural flow mechanisms, rock liquid and gas
expansion plus solution gas drive leave 80% or more of the original oil in place after the
pressure in the reservoir has been depleted. A vast amount of hydrocarbon thus remains
unrecovered in the world’s depleted reservoirs. Billions barrels of additional oil could
have been recovered through waterflooding, the most important method for improving
recovery from oil reservoirs [1].
Water flooding is often initiated after reservoir (pressure) depletion has occurred. It is
frequently the case that (free) gas saturation, or gas cap, forms in a solution gas drive
reservoir due to this pressure depletion. Initially, the increasing reservoir pressure due to
the injected water will force the free gas to redissolve into the oil. The response of the
oil production rate to the water injection will become most apparent after the water
injection has achieved fill-up of the gas space (Figure 1.1) [1].
The maximum oil production rate as the reservoir is re-pressurised depends upon the
reservoir characteristics and the injection rate. Continued water injection will eventually
lead to “breakthrough” of the injected water at the production wells. This peak oil
production rate will then start to decline as the water cut increases [1].
1.1 Water production
One of the most significant challenges in oil production operations relates to the
production of water. This excessive water production is recognised not only as a large
operational problem, but is also a significant economic and environmental one [2]. This
is particularly relevant when discussing mature fields, where water can be continuously
pumped into a reservoir with little oil being produced [3]. Many of these mature fields
in decline are currently producing at well above 80% water cut. A scenario such as this
has several negative implications, one of which being that the majority of the energy
supplied to the artificial lift system installed in the production wells will be spent on
lifting water, rather than oil [4]. More importantly, government regulations regarding
1
the treatment and disposal of this water are continually becoming more rigorous, and
trends in the development of legislation suggest that meeting legal requirements will
require increasingly costly water treatments [3, 5]. Estimates for global water
production vary and are generally hard to quantify. What can easily be identified
however is the growing trend in the water to oil ratio in terms of total global production.
In the mid-1990s, the water to oil ratio in North America was estimated at
approximately 7:1 [6], increasing to almost 10:1 ten years later [7]. Extrapolation of this
trend, suggests that the ratio of produced water to oil is currently well in excess of 10:1
[2]. The costs of treating this water have been estimated to be in excess of $40-50
billion per year, implying that development of new options to reduce this amount has
become a top priority within the industry [6].
Figure 1. 1: Waterflood performance at different periods [8].
2
There are, of course, other risks associated with water production that stretch beyond
the handling and processing costs. Scaling, corrosion and reservoir souring are all
significant challenges facing engineers around the world today [9]. These scenarios can
be only remediated by well workovers or remedial treatments, so these unnecessary
situations have significant impacts both technically, through lost performance, and also
financially, in lost revenue from decreased or halted production. Poor sweep efficiency
of water injection is another challenge that frequently requires attention. This
unfavourable situation can lead to bypassed oil, as well as unexpected and undesirable
early water breakthrough [5]. The problem has not been devoted to attention. Finding
approaches to dealing with these issues has been recognised as an effective way to
increase the overall recovery efficiency in a cost effective manner [2].
1.2 Waterflooding management
A review of the underlying physics of the water flood recovery processes indicates that
their efficiency is influenced by:
1. Timing of the initiation of the waterflood

2. Rate of water injection into and fluid production from the reservoir
3. Location of the water injection and distribution of water between injectors
Improper injection may cause early water breakthrough, higher water production and
lower recovery, leading to the unnecessary cost of handling produced water. Two
important questions need be answered for efficient waterflood performance.
1. What is the optimal total amount of water needed to be injected into the
reservoir?
2. How should be the water be allocated to the different injectors?
Water injection has two roles; (i) to maintain the reservoir pressure at a suitable level
and (ii) to push the oil toward the producers. The first question requires development of
the pressure maintenance plan, once the reservoir engineers defined the preferred
reservoir pressure. This pressure should ideally support the flow of fluid from the
reservoir to the producers and from bottom of the well to the surface, otherwise artificial
lift techniques will need to be installed in the production wells. However, although
injection of a large volume of water may keep the reservoir pressure constant, it will
3
also encourage early breakthrough of water in the producers, with the potential result of
reduction in the total field oil production. Therefore, the optimum total volume of water
should support the reservoir pressure, while being injected in such a manner that it
delays the breakthrough time as much as possible.
The answer to the second question involves the determination of the optimum injection
rate allocation process for the injection wells. The water injection allocation
management process should provide reduced support to production wells with the
greatest water cut, by injecting less water into those injection wells supporting these bad
producers.
1.3 Intelligent Well Completion (IWC) and waterflooding management
With a global decline in oil production, it becomes more important to improve and
increase the recovery of the remaining oil reserves. In recent years Intelligent Well
Completion (IWC) has been employed as a possible solution to some of these associated
challenges [2].
An Intelligent Well Completion (IWC) is essentially defined as any completion

containing at least one of the following elements: multiple downhole pressure and
temperature sensors, downhole flow measurement, fixed or adjustable inflow control
valves (ICV), or a multilateral well with inflow control of one or more laterals [10]. The
goal of IWCs has been to provide the user with one or more of the following elements:
well performance monitoring, zonal productivity control, and subsequent well
production optimisation (Figure 1.2) [11].
IWCs can be employed for closed loop reservoir management. The process of reservoir
management by intelligent completion (Figure 1.3) can be divided into two important
stages:
1) Monitoring:
a) Data gathering by different types of gauges and sensors inside the well.
b) Processing and interpreting the data to capture suitable desired information
about the performance of the production system.
2) Controlling:
a) Defining new production and injection strategies and scenarios.
4
b) Applying new schemes via control devices inside the well.
IWC can therefore be a good tool for monitoring the performance of waterflood, and
then controlling the injection and production schemes to improve the efficiency of the
water injection project.
Figure 1. 2: Different components of an intelligent well completion.
One of the most important tasks in IWC research is how to analyse and interpret the raw
data in order to capture useful information. As an example, new technique has been
developed to detect the time and the location of water influx into the well [12] or zonal
flow rates can be determined by analysing temperature sensor data[12, 13]. Another
important challenge is developing new algorithms based on this information to control
injection and production rate. This means that if we obtain good information describing
the performance of the system then how should we change the system parameters to
improve the performance of the system?
5
In this research we are interested in developing new techniques for obtaining the
necessary information from injection and production history from the gauges, in order
to understand waterflood performance and then, based on the captured information,
managing water and production rate to improve the flooding efficiency.
Controling
Data Controling
interpretation algorithm
Sensors
and IWC Control
devices
Gauges
Data Applying
gathering new scenario
Monitoring
Figure 1. 3: Closed loop reservoir management by IWC.
1.4 Research objectives
The overall objective of this research is to determine the optimum amount of injected
water into the reservoir and its optimal allocation between different injectors, by
employing injection and production history.
6
In Chapter 2 will review different waterflood performance monitoring techniques to
evaluate their effectiveness for monitoring the waterflood and determining performance
parameters. A new technique also is proposed to control total water injection volume
into the reservoir.
Two main approaches have been used in this study, to measure the optimum water
allocation factor. (i) Allocation management, based on inter-well connectivity between
injection wells and producers, together with performance evaluation of the production
wells and (ii) The optimum water allocation factor, determined by monitoring the effect
of change in the injection rate on future oil production from the production wells.
Chapters 3, 4 and 5 are concerned with the first methodology. Chapter 3, reviews
several statistical techniques used for inter-well connectivity measurement. The
application of some artificial intelligence techniques for determining the connection
between production and injection wells is examined in Chapter 4. New parameters are
introduced to describe production well performance in Chapter 5. In Chapter 6, a second
methodology for water allocation management, linear programming coupled with a
capacitive resistive model for water allocation management, is introduced and its results
compared with a combination of Genetic algorithm and reservoir simulation. Chapter 7
summarises the results of the previous chapters and discuss conclusions from the work
performed.
7
References
1. Thakur, G.C. and A. Satter, Integrated waterflood asset management. 1998,

Tulsa, Okla.: PennWell.
2. Tawse, C., Using dP Trending Analysis to Improve Waterflooding Performance,

in Institute of Petroleum Engineering. 2011, Heriot-Watt U/niversity:
Edinburgh.
3. Ferreira, E.A., et al., MANAGING WATER PRODUCTION PROBLEMS WITH

AN INTELLIGENT SYSTEM THAT USES PRODUCTION WELL AND
RESERVOIR DATA. 2002, World Petroleum Congress.
4. Nguyen, P.D., S.R. Ingram, and M. Gutierrez, Maximizing Well Productivity

Through Water and Sand Management - A Combined Treatment, in Production
and Operations Symposium. 2007, Society of Petroleum Engineers: Oklahoma
City, Oklahoma, U.S.A.
5. Kalli, C., Monetizing Reserves Faster by Managing Water Better, in Offshore

Technology Conference. 2006: Houston, Texas.
6. Wallace, E.D., et al., WATER MANAGEMENT IN THE OIL PRODUCTION

INDUSTRY. 2000, World Petroleum Congress.
7. Hecht-Nielsen, R., Neurocomputing. 1990, Reading, Mass.: Addison-Wesley

Pub. Co.
8. Thakur, G.C., Waterflood surveillance Techniques - A Reservoir Management

Approach. Journal of Petroleum Technology, 1991. 43(10): p. 1180-1188.
9. Souza, A.L.S., et al., Water Management In Petrobras: Developments And

Challenges, in Offshore Technology Conference. 2005: Houston, Texas.
10. Spearman, C., The proof and measurement of association between two things.
By C. Spearman, 1904. The American journal of psychology, 1987. 100(3-4): p.
441-471.
11. Schatzinger, R.A.I.R.C.T.C., Reservoir characterization : recent advances ;

[proceedings of the Fourth International Reservoir Characterization Technical
Conference, held March 2-4, 1997, in Houston, Texas]. 1999, Tulsa, Okla.:
American Association of Petroleum Geologists.
12. Zar, J., Significance Testing of the Spearman Rank Correlation Coefficient.
Journal of the American Statistical Association, 1972. 67(339): p. 578-580.
13. Khodaei, A., et al., A Mutual Information-Based Metric for Identification of

Nonlinear Injector Producer Relationships in Waterfloods, in SPE Western
Regional Meeting. 2009, Society of Petroleum Engineers: San Jose, California.
8
Chapter 2- Monitoring The Performance Of Waterflooding
2.1 Monitoring and surveillance of water injection projects
A billion barrels of additional reserves have been produced through waterflooding, one
of the most important methods of improving recovery from oil reservoirs; waterflood
projects account for over half of the Canadian and U.S. oil production [1]. With the
economic uncertainty of EOR techniques as a result of oil-price instability, optimization
and management of waterflood projects has become more important than ever [2].
A successful waterflood depends on the proper operation of individual wells in a

pattern, to maintain the balance between water injection and production over the entire
life of the field, and also on preventing well failures. Therefore it is necessary to design
a system of monitoring and control, which acquires and processes waterflood data, and
helps field engineers to make optimal decisions [3].
An essential element of a successful waterflooding project is a well-planned and well-

executed program of surveillance and monitoring [4]. Surveillance and monitoring
principles are the key to understanding reservoir performance and identifying
opportunities that will improve ultimate oil recovery. The key ingredients of any
surveillance program are planning and accurate data collection. Surveillance techniques
should always be a precursor to in-depth studies, including numerical simulation [5].
A fundamental geological/petrophysical analysis is a cornerstone of good reservoir

engineering analysis. However, geological studies alone do not conclusively quantify
the reserve and oil rate increases that can be achieved by optimizing the existing water-
floods. While a geological/ petrophysical study is key in understanding the initial
question: What is the OOIP?[6] , production and pressure surveillance data can
implicitly account for a useful scale of heterogeneity. Therefore, if used properly, this
data can be extremely useful in developing changes in operational strategy that can
maximize recovery [6].
Production, injection, water-supply, and water-disposal are four types of wells requiring
surveillance. Of these, production and injection wells require the most attention;
9
Monitoring well performance requires a program of selected well tests to be conducted
regularly [4]. The surveillance of production data is fundamental to good reservoir
management of waterfloods and miscible floods. This type of surveillance is useful to
understand flood performance to date and can highlight good versus poor recovery
areas. In particular, surveillance can identify areas of extreme water cycling, patterns
with poor sweep, or local voidage imbalances, providing “real-time” monitoring of a
flood without having to construct a detailed flow-simulation model [7].
Monitoring performance of waterflooding is an important issue in the management of

water injection. When sufficient data are available and production is declining, the past
production curve of individual wells field can be extended to indicate future
performance. In the following section we are going to review some of the common
techniques employed for monitoring the performance of water injection.
2.1.1 Semi-log plot of oil cut or water cut versus cumulative oil production (cut-cum
plot)
One of the commonly used performance curve analysis methods for a waterflood project
is plot of water cut or oil cut versus cumulative production (Qo) (Figure 2.1). This curve
is used when the economic production rate is dictated by the cost of water disposal. A
straight line extrapolation of log of water cut versus cumulative oil production may not
be reasonably done in the higher water cut level. On the other hand, if oil cut data are
used instead of water cut in the same levels, straight extrapolation of log of oil cut
versus cumulative oil production may deteriorate and lead to optimistic reserve
estimates [2, 8].
2.1.2 Semi-log plot of Water oil ratio (WOR) versus cumulative oil production
A common practice in petroleum engineering [9] has been to analyse the performance
of waterfloods by plotting log10(WOR) versus cumulative produced oil (Qo) for
individual wells or from field data.
10
Figure 2. 1: An example of the plot of water cut versus cumulative oil production [8].
krw
Defining the relative permeability ratio of water to oil ( ) with water saturation (Sw)
kro
by the following equation [10, 11]:
krw
= aebSw (Equation
kro
2.1)
Based on the Buckley-Leverett theory for a linear, immiscible displacement, the

logarithm of the water-oil ratio (WOR) is linearly proportional to the cumulative oil
production (Qo) .[12]
 b (1 − S wc )   a µo  1
L og=
10 (WOR )   Qo + L og10   + bS w − (Equation 2.2)
 OOIP   µw  Ln10
That is, a semi-log straight-line relationship exists between WOR and the cumulative
produced oil, Qo. The slope of the straight line is  ( wc ) b 1− S 

 and the intercept is
 OOIP 
 aµ  1 µo
L og10  o  + bS w − . In Equation 2.1
µw
is the viscosity ratio of oil to water;
 µw  Ln10
Swc is the connate water saturation; and OOIP is the original oil in place. The constants
a and b in Equation 2.2 are derived from defining the relative permeability ratio of
11
krw
water to oil ( ) with water saturation Sw, as defined by Equation 2.1. Therefore by
kro
plotting log10(WOR) versus cumulative produced oil (Qo) for individual wells or from
field data; the data generally form a straight line and this straight line can be
extrapolated to predict future performance and estimate ultimate oil recovery from
waterflooding (Figure 2.2). [12]
Figure 2. 2: Typical example of WOR plot for poor, average and good waterflood performances[13].
2.1.3 The X-Plot
Extrapolation of past performance on the “cut-cum” plot is a complicated task. The

difficulty arises mainly because curve fitting by simple polynomial approximation does
not result in satisfactory answers, in most cases.[11]
Ershaghi and Omorigie (1978) developed the X-Plot water flood analysis based on the
semi-log linear relative permeability ratio (Equation 2.1) for intermediate saturation
values, as follows:[14]
The fractional flow equation (after neglecting the capillary pressure and gravity terms)
can be written as [11];
1
fw = (Equation 2.3)
ko µ w
1+ ×
k w µo
By introducing Equation 2.1 into Equation 2.3, it can be written as[11]:
12
1
fw = (Equation 2.4)
1 + AebSw
According to Welge [15], the water saturation at the producing end could be expressed
as [11]:
1 − fw
S=
w Sav − (Equation 2.5)
f w'
'
where Sav is the average water saturation, and f w is the derivative of f w with respect to
Sw. Since [11]:
Sav = ER (1 − S wc ) + S wc (Equation 2.6)
then:
1 − fw
S w = ER (1 − S wc ) + S wc − (Equation 2.7)
f w'
In these equations, ER is the reservoir oil recovery, which is the volume of the oil
recovered divided by oil in place [11].
The first derivative of Equation 2.4 will be [11]:
− AbebSw
f = bSw 2 =
'
−bf w (1 − f w ) (Equation 2.8)
(1 + Ae )
w
On substitution, in Equation 2.5 will be [11]:
1
S w = ER (1 − S wc ) + S wc − (Equation 2.9)
b. f w
Putting Equation 2.9 into Equation 2.4 and solving it based on ER gives:[11]
1   1  1 1  1 
=ER ln  − 1 −  −  S wc + ln A  (Equation 2.10)
b(1 − S wc )   f w  f w  1 − S wc  b 
13
By defining [11]:
1
m= (Equation 2.11)
b(1 − S wc )
  1  1
=
X ln  − 1 −  (Equation
  fw  fw 
2.12)
1  1 
n=
−  S wc + ln A  (Equation 2.13)
1 − S wc  b 
Equation 2.10 can be expressed as [11]:
=
ER mX + n (Equation 2.14)
Therefore, a graph of fractional recovery (ER) versus X results in a straight line (Figure
2.3). The straight line may be extrapolated to any desired water cut to obtain the
corresponding recovery in future. But since the objective is always to project a
waterflood performance into the future, only water cuts higher than 50% should be used
in the linear regression model. The slope m and intercept n can be obtained from
production data, if the swept volume is known. Other than predicting waterflood
performance by extrapolating the linearity between X and ER , permeability ratio
constants a and b can also be computed, using the following equations [16]:
µo −b n(1− Swc )+ Swc 

a= e (Equation
µbw
2.14)
1
b= (Equation 2.15)
m(1 − S wc )
14
Figure 2. 3 An example of plot of X values versus Recovery obtained from Equation 2.2 [17].
2.1.4 The Hall Plot
A simple and efficient method of monitoring injection well performance at steady-state

flow was proposed by Hall in 1963[18, 19]. Hall devised a qualitative approach to
eliminate the complications of both pressure and injection rate variations[20]. This
technique, used to analyze injection-well data, is based on a plot of cumulative pressure
versus cumulative injection [21].
The Hall plot is a tool to analyze steady-state flow at an injection well. Originally, it
was based on the radial flow Pi model [22]. According to this model [23]:
µ r
P=i Pe + Ln e qi (Equation 2.16)
2π kH rw
where Pi and Pe are the downhole wellbore pressure (injection pressure) and reservoir
pressure, respectively, qi is the injection rate, µ is the injected fluid viscosity , k is
formation permeability and H is the reservoir thickness. re is the well influence zone
radius, which is the zone near the wellbore where the fluid pressure changes appreciably
due to the injection. And rw is the well-bore radius [23].
Equation 2.16 is based on several assumptions. The fluid is homogeneous and

incompressible. The reservoir is vertically confined and uniform, both with respect to
the permeability and the thickness. The reservoir is homogeneous, isotropic, and
horizontal and gravity does not affect the flow. Consequently, the flow is radial. During
15
the entire time of observations, the pressure at the distance equal to re is constant, and
this distance is constant as well. In practice, not all, if any, of these assumptions are
strictly satisfied. The injection interval usually covers multiple zones [23].
Equation 2.16 can be integrated with time to:
t t
µ  re 
∫ ( P − P ) dτ =
t0
i e∫ 2π k H ln  r
t0 w
 qi dτ
w 
(Equation
2.17)
Then the plot of the integral:
t
∏(t )= ∫ ( P − P ) dτ
t0
i e
(Equation 2.18)
versus the cumulative injection volume:
t
V (t ) = ∫ qi dτ (Equation 2.19)
t0
at a constant ambient reservoir pressure, pe , is a straight line whose slope is equal to

µ r 
ln  e  , which is the reciprocal of the injectivity index, (II) [23] (Figure 2.3).
2π k w H  rw 
Thus, a deviation from a straight line should signify alterations of the formation
properties [19].
The Hall Plot, is an alternative to the transient well test analysis approach. Technically,
it is very simple: just plot the time integral of the difference between the injection and
reservoir pressures versus cumulative injection. The integration automatically filters out
short-term fluctuations. The slope of lot is interpreted as an indicator of the average well
injectivity. As inputs, the Hall method only requires the regular collection of injection
rates and injection pressures that are a part of routine waterflood operations. At normal
conditions, the plot is a straight line and kinks on the plot should indicate changes of
injection conditions [23]. But it should be mentioned that if Hall s method is applied
without a priori knowledge of the effective ambient reservoir pressure, then the
16
conclusion that a kink in the plot is an indication of changes in the well injectivity can
be wrong [23].
A typical Hall plot at different conditions is illustrated in Figure 2. Early in the life of an
injection well the water-zone radius will increase with time, causing the slope to
concave upward, as shown by Segment ab in Figure 2.3. After fill-up, Line bA indicates
stable or normal injection. An increasing slope that is concave upward generally
indicates a positive skin or poor water quality (Line D). Similar slopes may occur if a
well treatment is designed to improve effective volumetric sweep. In this case, however,
the slope will first increase and then stay constant. Line B indicates a decreasing slope,
indicating negative skin or injection above parting pressure. The injection under the
latter condition can be verified by running step-rate tests. A very low slope value, as
shown by Line bC, is an indication of possible channelling or out-of-zone injection [2].
Figure 2. 4: Typical Hall plot for various injection well conditions[2].
2.2 Case study
In the next step in this chapter we will employ a small reservoir model to evaluate these
monitoring techniques. In this analysis we are interested to see the efficiency of those
methods in differentiating between normal and improved flooding performance. In other
words we want to know:
1. What is the accuracy of these techniques in forecasting waterflood performance?
17
2. What kind of information can be obtained from each method?
3. Are these methods able to determine the optimum parameters in order to
improve the waterflood performance?
4. What will be the advantages and disadvantages of each technique?
In order to reduce the complexity of the problem and focus more on the techniques
rather than the challenges of complex models, a small reservoir model is used to carry
out this analysis. The employed reservoir model is a standard example of the Reveal
reservoir simulator that is mainly used to study water injection management. Two
different injection schemes are defined for water injection.
In order to answer these questions all these methods are plotted for two injection
schemes. Then for each method, the difference between the plots is analysed.
2.2.1 Model description
The model has 3 producers and 4 injectors that support them. The position of the wells
has been shown in Figure 2.5. The model has 3 reservoir layers. The properties of the
reservoir are given in Table 2.1.
Table 2. 1: The properties of the studied reservoir model.
Number of the cells 20×15×4

Initial oil in place 5.7522×108 STB
Initial reservoir pressure 4000 psia
Temperature 100o F
Compressibility 3×106 1/psi
Permeability 220-320 mD
Porosity (fraction) 0.2-0.25
Oil density 35 API
Bubble point pressure 2054 psia
Gas oil ratio 500 Scf/STB
18
Figure 2. 5: Location of the wells in the reservoir (colours represent the reservoir porosity).
The model was set to produce for 40 years and the injection started from the beginning
of production, for both injection scenarios. The same production scenario was employed
for both injection schemes. The production wells were producing with constant surface
liquid production, as shown in Table 2.2.
Table 2. 2: Allocated liquid production rate of each producer.
Production Wells Liquid production

(STB/day)
Producer 1 9000
Producer 2 9000
Producer 3 12000
For injection two scenarios were studied: the base case scenario and the water allocation
management scenario (WAM).
19
2.2.1.1 Base case scenario
In this scenario the overall injection was controlled by voidage replacement by injecting
with total voidage ratio (VR) of 1 for the whole 40 years of production. This means that
we are injecting as much as we are producing from the reservoir. The total injection rate
was equally distributed between injectors, so that, each injection well is injecting with
0.25 VR.
2.2.1.2 WAM scenario
In the WAM scheme, the total injection rate of the reservoir was based on voidage
replacement with VR of 1, as in the base case scenario. For the first 20 years, the water
distributed equally between the injectors. But for the next 20 years, at 5 year intervals,
water was allocated to the injectors based on the inter-well connectivity between
injectors and producers and also the performance of the production wells.
In order to do that, a capacitive resistive mode (CRM) [24, 25] is used to quantify inter-
well connectivity between injectors and producers (the procedure briefly explained in
Chapter 3 sections 3.2.3 and 3.3.3). Then the oil production index (OPI) is defined to
describe the production well performance (briefly explained in Chapter 5 section 5.1.4).
Then more water was allocated to the injectors that were highly connected to the
producers with better OPI. This helped to improve the waterflood performance by
promoting more oil recovery (Figure 2. 6) and less water production (Figure 2.7).
2.2.2 Plot Analysis
An Excel macro was developed in this research to plot all these performance plots,
based on the results of the simulation imported from the reservoir model by the Reveal
fluid flow simulator. This macro was able to plot all the above mentioned methods.
In the following section we will review the result the plot analysis for both injection
schemes, for each surveillance technique.
20
Figure 2. 6: Plot of oil recovery versus time for the base injection scenario and the WAM injection scenario.
Figure 2. 7: Plot of water cut versus time for the base injection scenario and the WAM injection scenario.
21
2.2.2.1 Cut-Cum plot
Oil cut versus cumulative oil production was plotted on a semi-log scale, as shown in
Figures 2.8 and 2.9.
100
Water cut (percentage)
10
Base Case
WAM
1
0 50000000 100000000 150000000 200000000 250000000
Cumulative oil production (STB)
Base Case
WAM
40
120000000 140000000 160000000 180000000 200000000 220000000
Figure 2. 8: Semi log plot of water cut versus cumulative oil production plot for base injection scenario and the
WAM injection scenario.
As can be seen, there is a change in the trend of the WAM plot (see green circles in
Figure 2.8) while this trend is constant for the base case. This is the time when water
allocation management has been started. At the end of each plot it can be seen that more
cumulative oil is produced in the lower water cut in the WAM scheme (red and blue
22
arrows in Figure 2.8), which is a signature of improvement in the water injection
efficiency. By fitting the cut-cum curve with proper polynomial, the future performance
of waterflood in terms of future water cut or future cumulative oil production can be
estimated from the cut-cum plot. Three types of polynomial are used to fit the cut-cum
plot: linear, second order and third order polynomials. The results of extrapolation with
these three polynomials were compared with the result of simulation, to evaluate the
accuracy of this method. Table 2.3 shows the results of this analysis for the base case
scenario after 10 years.
100 100

y = 4E-07x + 9.0794 y = -7E-16x2 + 5E-07x + 2.572
10 10
Base Case Base Case

1 1
0 100000000 200000000 300000000 0 100000000 200000000 300000000
Cumulative oil production (STB) Cumulative oil production (STB)
100
10
y = -6E-25x3 - 5E-16x2 + 5E-07x + 2.9186
Base Case
1
0 100000000 200000000 300000000
Figure 2. 9: Extrapolation of cut-cum plot for estimating future performance by employing first, second and
third order polynomials.
23
Table 2. 3: Results of estimated future water cut for base case scenario from cut-cum plot extrapolation for
cumulative oil production of 2.23 × 108 STB.
Polynomials
linear 2nd order 3rd order Simulation result
Estimated Water cut 98.1 79.1 89.4 81.3
Estimation error 20.5 2.7 9.8 ---
The second order polynomial gave the best estimation compared to the simulation
forecast. Looking at the results of other polynomials shows that fitting the cut-cum plot
with a suitable polynomial has a significant effect on the result of extrapolation. We
also tried to fit the polynomial to the last portion of the plot, for the water cut values
greater than 70% (Figure 2.10). Using the second order polynomial gave a water cut of
79.5325%. This is not very different from the extrapolation resulted from all data points
(estimation error of 2.25 %).
100
y = -2E-15x2 + 1E-06x - 43.953
10
160000000 170000000 180000000 190000000 200000000 210000000
Figure 2. 10: Second order polynomial fitted for water cut values higher that 70%.
The second procedure was applied for extrapolation of the cut-cum plot for the WAM
injection plan but only the 2nd order polynomial was used to fit the curve. At first, the
polynomial was fitted to all data points. Comparing the results of extrapolation with the
simulation results shows an error of estimation of 8.14%, which is higher than the
24
results of base case plot (Figure 2.11). This is because of the change in the curve after
starting water allocation management. In order to obtain a more accurate extrapolation,
the curve should be fitted after the change in the injection plan. Doing this reduced the
error to 1.46% (Figure 2.12 and Table 2.4). So it is important not only to use a suitable
polynomial, but also to select the correct portion of the plot to perform the extrapolation
in the cut-cum plot.
100
y = -1E-15x2 + 6E-07x + 1.8276

10
WAM
1
0 50000000 100000000 150000000 200000000 250000000
Figure 2. 11: Curve fitting of cut-cum plot for the WAM scenario when all data points are used.
25
100
Water cut (percentage) y = 1E-15x2 - 3E-07x + 76.294
Series2
10
120000000 140000000 160000000 180000000 200000000 220000000
Figure 2. 12: Curve fitting of cut-cum plot for WAM scenario for data points after start of injection
management.
Table 2. 4: Results of estimated future water cut for WAM injection scenario extrapolated to cumulative oil
production of 2.38 × 108 STB.
All data points After WAM Simulation

Estimated Water cut (%) 87.9 61.5 72.6
Estimation error (%) 21.1 15.2 ---
2.2.2.2 WOR plot
Figure 2.13 shows the plot of WOR in logarithmic scale versus cumulative oil
production for both water injection scenarios.
26
10
Base Case
WAM
WOR (STB water/STB oil) 1

0 50000000 100000000 150000000 200000000 250000000
0.1
0.01
Cumulative oil produced (STB)
10
Base Case
WAM
WOR (STB water/STB oil)
1
120000000 140000000 160000000 180000000 200000000 220000000
Figure 2. 13: WOR plots for base case and WAM injection schemes.
The WOR plot of the base case is a straight line, except at the beginning of production,
but in the case of WAM there is a change in the slope of the plot, which represents the
change in the injection plan. Compared with the cut-cum plot, the difference between
the base case and the WAM plot is visually clearer. We can also observe that in the
WAM plot, more cumulative oil production is obtained with less WOR.
27
According to Equation 2.2 there is a linear relationship between the logarithm of WOR
and cumulative oil production, so a linear polynomial is employed to fit the curves.
Curve fitting is therefore simpler, compared to cut-cum plots.
Results of extrapolation for the base case scenario are shown in Figure 2.14 and Table
2.5. The extrapolation has been done for whole set of data points and also for WOR
values greater than 1.
10 10
y = 1E-08x - 0.2427
WAM

1
0 50000000 100000000 150000000 200000000 250000000
y = 2E-08x - 1.101
0.1
WAM
1
0.01
Cumulative oil produced (STB) 0 50000000 100000000 150000000 200000000 250000000
Figure 2. 14: WOR plot curve fitting for base case scenario when all data points are used and when WOR
values higher than of 1 are employed.
Table 2. 5: Estimated future WOR for cumulative oil production of 2.23 × 108 STB for base case, from WOR
plot.
All points WOR>1 Simulation

Estimated WOR (fraction) 5.1 4.6 4.3
Although we are limited to the use of a linear equation for the WOR plot curve fitting, it
still depends upon the portion of the curve we choose to do curve fitting. It is better to
use that part of the plot which becomes a straight line.
Curve fitting has also been done for the WOR plot of the WAM injection scheme.
28
10
y = 1E-08x - 0.2427
1
0 50000000 100000000 150000000 200000000 250000000
0.1
WAM
0.01
10
WAM
y = 2E-08x - 1.101
1
0 50000000 100000000 150000000 200000000 250000000
Figure 2. 15: WOR plot curve fitting for WAM injection scenario.
Table 2. 6: Estimated future WOR for cumulative oil production of 2.38 × 108 STB for WAM injection scheme.
All points After WAM Simulation

Estimated WOR (fraction) 2.6 3.6 4.3
Again, it can be concluded that the selection of data points is very important in curve
fitting and extrapolation. In order to get an accurate forecast we have to consider that
part of the curve which represents the last stable change in the water injection plan. One
29
of the important factors that we need to consider for data selection is change in the
operational conditions of the project, such as a change in the injection or production
plan or changes in the number of active production and injection wells. As we saw
when we used the data points after the start of water allocation management, the
estimation become more accurate.
2.2.2.3 X-plot
The X-plot has been plotted for both waterflood projects in Figure 2.16. The green
circle shows the change in the injection plan in the WAM scenario. After this point the
curve shifts upward, showing improvement in the water flood project as more recovery
is obtained. For example, as shown by the red and blue arrows for a specific X value
(representing specific fw), more oil recovery is achieved in the WAM plan.
Both plots have been extrapolated to obtain future reservoir recovery. Results of the X-
plot curve fitting for both cases are illustrated in Figures 2.17 and 2.18.
Tables 2.7 and 2.8 represent the calculated future oil recovery from the extrapolation of
X-plots for base case and WAM injection plans.
0.4
0.35
Oil Recovery (fraction)
0.3
0.25
Base Case
0.2
WAM
0.15
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6
X values
Figure 2. 16: Plot of oil recovery versus X values (X–plots) for base case and WAM injection schemes.
30
Looking to the extrapolation results of the X-plot shows that, although it has been
recommended to use the plot for values of fw more than 50% for recovery estimation, in
this example, using the plot for values above 65 % provided us with more accurate
estimation, especially in the case of WAM.
0.4 0.4
y = 0.2722x - 0.3343 y = 0.214x - 0.1957
0.35 0.35

0.3 0.3
0.25 0.25
Base Case Base Case

0.2 0.2
0.15 0.15
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.1 2.2 2.3 2.4 2.5 2.6
X values X values
Figure 2. 17: X-plot curve fitting for base case scenario for all data points and for fw higher than 65%.
Table 2. 7: Estimated future oil recovery from extrapolation results of X-plot at X value of 2.7 for base case.
All data points fw>65 Simulation

Oil recovery estimation (fraction) 0.40 0.38 0.38
0.45 0.39
y = 0.5158x - 0.8206
0.4 0.37
0.35
0.35
0.33
0.3 y = 0.2999x - 0.3367
0.31
0.25
0.29
WAM WAM
0.2 0.27
0.15 0.25
1.9 2 2.1 2.2 2.3 2.4 2.5 2.6 2.1 2.2 2.3 2.4 2.5
X values X values
Figure 2. 18: X-plot curve fitting for WAM scenario for all data points and for fw higher than 65%.
Table 2. 8: Estimated future oil recovery from extrapolation results of X-plot at X value of 2.7 for WAM
scenario.
All data points fw>65 (After WAM) Simulation

Oil recovery estimation (fraction) 0.50 0.43 0.41
31
Apart from the prediction of future oil recovery by X-plot, Equations 2.14 and 2.15 can
be employed to determine the a and b constants of Equation 2.1. Table 2.9 shows
calculated values of these constants for both water injection plans.
Table 2. 9: Determination of a and b constants.
Base case WAM Real Values

m 0.21 0.29 ---
n 0.19 0.33 ---
a 3.60 4.22 4.75
b 5.19 3.70 3.92
The values of a and b can be used to produce an effective field relative permeability
plot knowing the estimates of Swc and the viscosity ratio.
2.2.2.4 Hall plot
Unlike the previous methods that applied for the whole reservoir, the Hall plot can be
generated for each individual injector. The following Figures, 2.19 and 2.20, illustrate
the Hall plots for all the injection wells in both injections schemes.
20000000
18000000 Base Case I 1

Cumulative pressure (psi.day/STB)
16000000 Bace Case I 2
10000000
8000000
6000000
4000000
2000000
0
0 50000000 100000000
Cumulative water injected (STB)
Figure 2. 19: Determined Hall plots of each injection well for base case injection scenario.
32
20000000
18000000
WAM I 1

16000000
WAM I 2
14000000
WAM I 3
12000000
WAM I 4
10000000
8000000
6000000
4000000
2000000
0
0 50000000 100000000 150000000
Cumulative water injected (STB)
Figure 2. 20: Determined Hall plots of each injection well for WAM injection scenario.
Looking at the base case injection well Hall plots shows that the same amount of water
was injected from each injector but different cumulative pressure was obtained. The
slope of the Hall plot is the reciprocal of the injectivity index of the injection wells.
Those wells with more cumulative pressure have steeper slopes, showing these wells
have lower injectivity indexes. Since VR is 1, the reservoir pressure is almost constant,
therefore to inject the same amount of water, injection pressures need to be increased,
resulting in higher cumulative pressure for these wells (injectors 3 and 2).
33
4500000 12000000
4000000
WAM I 1 WAM I 2

10000000
3500000 Base Case I 1 Base Case I 2
3000000 8000000
2500000
6000000
2000000
1500000 4000000
1000000
2000000
500000
0 0
0 50000000 100000000 150000000 0 50000000 100000000 150000000 200000000
Cumulative water injected (STB) Cumulative water injected (STB)
20000000 4500000
18000000
Base Case I 4
4000000
WAM I 3

16000000
WAM I 4
Base Case I 3 3500000
14000000
3000000
12000000
2500000
10000000
2000000
8000000
1500000
6000000
4000000 1000000
2000000 500000
0 0
0 50000000 100000000 150000000 0 50000000 100000000
Cumulative water injected (STB) Cumulative water injected (STB)
Figure 2. 21: Comparison of the Hall plot of each injector in base case and WAM injection scenarios.
Comparing the Hall plot of each injector in both injection plans (Figures 2.21) shows
that the amount of water that injected from each injector has changed from the base case
injection scenario to the WAM injection scheme and therefore, different cumulative
pressures are obtained from each of them. This is the only difference that we can find
between these two injection plans from the Hall plot.
The slope of the Hall plots for each injection well in both scenarios remains constant.
Therefore, we can conclude that there is no sign of positive or negative skin or change
in injectivity index.
The injectivity index of each injection well has been calculated from the slope of their
Hall plots (Table 2.10). Since the input data came from reservoir simulation and there
was no change in the injectivity index of the wells, the results are in very good
agreement with the average injectivity index of the wells obtained from simulation.
Table 2. 10: Calculated injectivity index from Hall plots for each injection well.
I1 I2 I3 I4
Slope 0.03 0.07 0.15 0.03
II from Hall plot 32.36 13.51 6.25 28.40
II from simulation 32.38 13.48 6.25 28.65
34
2.3 Reservoir voidage management
This section aims to develop a new voidage replacement plan for the reservoir in order
to delay the breakthrough time as much as possible, maintain the reservoir pressure, and
decrease water production. Again, in order to reduce the complexity of the problem and
focus more on the techniques rather than the challenges of complex models a simple
horizontal reservoir is employed in this analysis. The reservoir produces with constant
liquid production rate. A sensitivity analysis was carried out to monitor the effect of
different injected volumes of water on pressure maintenance and oil recovery from the
reservoir. The result of this sensitivity analysis was then used to define a suitable
reservoir voidage strategy.
2.3.1 Model description
Figure 2.22 shows the one layer, homogenous and horizontal reservoir model which is
used in this study. This model contains one production well and one injection well. The
properties of this model are given in Table 2.11.
Figure 2. 22: Schematic of the reservoir model (colours representing oil saturation).
35
Table 2. 11: Properties of the studied reservoir model.
Dimension Depth Pressure Temperature Permeability Porosity Net to gross
450×450×25 ft3 5000 ft 5000 psig 200 oF 500 mD 30% 100%
2.3.2 Methodology for reservoir voidage management
This section describes the steps in setting up the model and defining the voidage
replacement plan:
1. The reservoir was set to produce with constant liquid production of 250
STB/day for three years. The voidage replacement strategy was applied for
water injection and water injected into the reservoir with constant voidage ratio
(VR) from beginning of the oil production until the end of the production time. A
sensitivity study was done to see the effect of different injection volumes of
water on breakthrough time and final oil recovery. Figures 2.23 to 2.25 show the
results of injection with different VRs from 0.1 to 1.0.
Figure 2. 23: Oil production rate versus the time for injection with different VRs.
36
Figure 2. 24: Oil recovery versus time, obtained from injection with different VRs.
Figure 2. 25: Cumulative water production from injection with different VRs.
As expected, this analysis shows that:

a) More injection of water will cause early break through, less oil recovery
with more water production but better pressure maintenance.
37
b) Less injection will delay breakthrough time and cause more oil to be
swept but cannot maintain reservoir pressure and, as we can see, with
VRs of less than 0.6, at some point, reservoir pressure (Figure 2. 25) is
not enough to support production any more.
Figure 2. 26: Change in the reservoir pressure versus the time during injection with
different VRs.
Even with a VR of 0.6, the reservoir pressure is not enough to deliver the
produced fluid to the surface. We therefore need to define suitable
reservoir pressure, in other word the minimum reservoir pressure
required to produce from the reservoir at the planned production rate.
2. In order to define the suitable reservoir pressure, a well model is designed for
the production well. This well is employed to run a sensitivity analysis on the
reservoir pressure (Figure 2.27) by analysing the well model. According to this
sensitivity study, the minimum reservoir pressure (Prmin) in order to produce
from the production well is 2400 psi. But since the bubble point pressure is 2630
psi and we do not want to produce below the bubble point pressure, 2750 psi
selected as the minimum reservoir pressure.
38
Figure 2. 27: Sensitivity analysis on the well outflow performance to determine the minimum required
reservoir pressure.
3. After determining the minimum reservoir pressure, the following procedure is

proposed to define the reservoir voidage replacement plan:
1) The reservoir pressure is allowed to decline to decline the suitable
reservoir pressure before the breakthrough time, by starting to inject with
VR less than 1.
2) As soon as the pressure falls to the minimum pressure, injection of VR 1
is commenced to maintain the pressure at the minimum level.
4. The starting VR for injection should be selected in a way to delay breakthrough
time as long as possible. Sensitivity analysis on injection with different VRs
shows the breakthrough will happen after injection of 45000 STB cumulative
water volume (Qwibt) into the reservoir. Therefore, based on this strategy it can
be said:
Q=
wibt Qwibt1 + Qwibt 2 (Equation 2.20)
in which Qwibt1 and Qwibt2 are the cumulative water injected by the first and
second VRs, before breakthrough time. The first important parameter is the time
of starting the injection with the second VR (which is 1 in this procedure). Let us
call this time the time of starting pressure maintenance (tpm). tpm should be the
39
time in which reservoir pressure declines to the minimum required reservoir
pressure (Prmin). According to the definition of compressibility we can say:
dV =−C × V pv × dP (Equation 2.21)
In which dV is the depleted volume of the reservoir, C is the reservoir total

compressibility, Vpr is the reservoir pore volume and dP is the change in the
reservoir pressure. dV is the produced fluid from the reservoir minus the injected
volume of the water into the reservoir:
dV =
t × Bo ql (1 − VR) (Equation 2.22)
in which the t is the duration time of the production. Therefore tpm can be
estimated by the following equation:
−C × V pv × ( Pr − Pr min )
t pm = (Equation 2.23)
Bo × ql (1 − VR)
in which Pr is the initial reservoir pressure and ql is the liquid production rate of
the reservoir. Table 2.12 shows the estimated tpm for different VRs.
Table 2. 12: Calculated tpm for each injection volume of water, in terms of different VRs.
VR tpm (days)
0.9 330
0.8 180
0.7 120
0.6 85
0.5 70
5. The important question is what VR should be selected to start the injection. The
governing parameter is the time of breakthrough. Since the voidage replacement
strategy is used for injection, the injection rate is:
q=i VR × Bo ql (Equation 2.21)
In which qi and ql are injection rate and liquid production rate respectively and
Bo is the oil formation volume factor. According to Equation the 2.21, we can
say:
Qwibt = t pm × VR1 × ql + t2 × VR2 × ql (Equation 2.22)
40
in which t2 is the duration time of injection with the second VR (which is 1 in
this case) and it will be:
Qwibt − t pm × VR1 × ql
t2 = (Equation 2.23)
ql
and the breakthrough time will be:
t=
bt t pm + t2 (Equation 2.24)
Figure 2.28 shows the calculated breakthrough time for each combination of
VRs.
185
180
175
Breakthrough time (days)
170
165
160
155
150
145
140
135
0.9 & 1 0.8 & 1 0.7 & 1 0.6 & 1 0.5 & 1
VRs Combination
Figure 2. 28: Calculated breakthrough time for different starting injection VRs.
As can be seen, the maximum breakthrough time obtained is when we start to

inject with a VR of 0.7.
Therefore, according to the proposed methodology, the voidage replacement
plan for this reservoir will be to start to inject with VR of 0.7 for 120 days then
increase the VR to 1 for the rest of the production.
The following figures compare the results of new pressure maintenance strategy
with the base case plan in which water was injected with VR of 1.
41
Figure 2. 29: Oil production rate versus time for base case and voidage management injection scenarios.
Figure 2. 30: Plot of oil recovery versus time for base case and voidage management injection scenarios.
42
Figure 2. 31: Plot of cumulative water production versus time for base case and voidage management injection
scenarios.
Figure 2. 32: Cumulative injected water versus time for base case and voidage management injection
scenarios.
43
2.4 Results and discussion
In general, the techniques reviewed can be classified in to two groups. The first group
comprises those methods that addressed the determination of future oil reserves or
future oil recovery based on past production data. The second group of techniques, that
analyse the injection well’s history, are used to monitor the performance of the
injection well.
The first group includes the Cut-cum, WOR and the X-plot. To some extent, and by
selecting the most appropriate section of the plots, they were accurate enough to predict
future cumulative oil production (the Cut-Cum & WOR plots) or future oil recovery (the
X-plot). This study indicated that Cut-Cum plot predicted better than the WOR plot but
the results were very sensitive to the polynomial fitted to the curve. By contrast, a
simple linear polynomial was employed for extrapolating the WOR plot. The most
accurate estimation was obtained from X-plot. This can be used to generate an effective
permeability plot that, unlike the laboratory curves, is a composite curve that includes
the reservoir geometry, heterogeneity, and operational conditions of the field, along
with the displacement characteristics of the fluids [17]. Note that, in order to generate
the X-plot, fw should be transformed to X values by Equation 2.12 and the extrapolation
is only valid for fw higher than 50%. The proposed technique in the X-plot is based on
actual performance of a waterflood project.
It should be noticed that one major assumption in the techniques used for estimating
future performance of the waterflood was that the operating method will remain
relatively unchanged. Any variations in operational procedure such as change in the
injection or production strategy, infill drilling or shutting down wells, will result in a
shift of actual performance, and should be considered in updating the plots [11]. In this
study, we used the reservoir simulation injection and production history. Data
preparation for forecasting is another important factor for good estimation. The input
data set should be cleaned from any noise and they should represent the current
condition of the waterflood system.
The only technique that belongs to the second group is the Hall plot. In this study, the
Hall plot provided good information about the performance of the injection well
performance such as injectivity index or change in the skin, but it could not give any
44
information about the overall performance of waterflood, in terms of improvements in
the oil recovery or a decrease in water production.
One of the main questions that this study sets out to answer is: are these techniques able
to determine the optimum injection parameters in order to improve waterflood
performance?
Unfortunately, the answer is no. Those methods in the first group showed that WAM
improved the oil recovery from the reservoir but there is no way to find how this
improvement was achieved. In the case of the Hall plot, it shows the change in the
injection parameters from base case to WAM but we could not determine the influence
of this change on waterflood efficiency.
The only way that we can use these methods for defining an algorithm for water flood
management, such as determining the total amount of water injection or water allocation
between injectors, is to use the plot obtained from a previous successful flood project as
a type curve, in order to control injection for a new project that has the same flood
characteristics as the oil project. In this case, the Hall plot can be a good candidate, as it
directly addresses the waterflood parameters, such as injection rate or injection pressure.
For example, if an X-plot of a water injection project shows a good oil recovery, the
Hall plot of the associated injectors can be employed for the injection control of another
reservoir with the same properties as the current one.
The illustrated workflow for the reservoir voidage management shows a significant
increase in oil recovery. Not only did oil production increase, water production reduced
and this happened with less injection volume of water. High volume injection of water
can maintain reservoir pressure but it also causes early water breakthrough. This can
cause a poor well outflow performance and reduces the efficiency of the water/oil
displacement process. Proper reservoir voidage management can avoid excessive water
production and increase oil recovery. This example illustrates that it is not always
necessary to maintain the reservoir pressure at the initial reservoir pressure. Defining
the suitable reservoir pressure, based on well outflow performance and oil recovery, is a
simple approach to define a voidage management strategy. Water breakthrough time is
another important parameter that can help to develop a suitable voidage plan for the
reservoir.
45
2.5 References
1. Willhite, G.P. and A. Society of Petroleum Engineers of, Waterflooding. 1986,

Richardson, TX: Society of Petroleum Engineers.
2. Thakur, G.C., Waterflood surveillance Techniques - A Reservoir Management

Approach. Journal of Petroleum Technology, 1991. 43(10): p. 1180-1188.
3. De, A., D.B. Silin, and T.W. Patzek, Waterflood Surveillance and Supervisory
Control, in SPE/DOE Improved Oil Recovery Symposium. 2000, Copyright
2000, Society of Petroleum Engineers Inc.: Tulsa, Oklahoma.
4. Talash, A.W., An Overview of Waterflood Surveillance and Monitoring. Journal

of Petroleum Technology, 1988. 40(12): p. 1539-1543.
5. Terrado, R.M., S. Yudono, and G.C. Thakur, Waterflooding Surveillance and

Monitoring: Putting Principles Into Practice. SPE Reservoir Evaluation &
Engineering, 2007. 10(5): p. pp. 552-562.
6. Baker, R., Reservoir Management For Waterfloods. Journal of Canadian

Petroleum Technology, 1997. 36(4).
7. Batycky, R.P., et al., Revisiting Reservoir Flood-Surveillance Methods Using

Streamlines. SPE Reservoir Evaluation & Engineering, 2008. 11(2): p. 387-394.
8. Thakur, G.C. and A. Satter, Integrated waterflood asset management. 1998,

Tulsa, Okla.: PennWell.
9. Craig, F.F., The reservoir engineering aspects of waterflooding. 1971: H. L.

Doherty Memorial Fund of AIME.
10. Calhoun, J.C., Fundamentals of reservoir engineering. 1953, Norman:

University of Oklahoma Press.
11. Ershaghi, I. and O. Omorigie, A Method for Extrapolation of Cut vs Recovery

Curves. 1978.
12. Lo, K.K., H.R. Warner Jr., and J.B. Johnson, A Study of the Post-Breakthrough
Characteristics of Waterflood, in SPE California Regional Meeting. 1990, 1990
Copyright 1990, Society of Petroleum Engineers Inc.: Ventura, California.
13. Baker, R., Reservoir Management For Waterfloods-Part II. Journal of Canadian
Petroleum Technology, 1998. 37(1).
14. Yang, Z., A New Diagnostic Analysis Method for Waterflood Performance. SPE
Reservoir Evaluation & Engineering, 2009. 12(2): p. pp. 341-351.
15. Welge, H.J., A Simplified Method for Computing Oil Recovery by Gas or Water
Drive. 1952.
46
16. Yang, Z., A New Diagnostic Analysis Method for Waterflood Performance, in
SPE Western Regional and Pacific Section AAPG Joint Meeting. 2008, Society
of Petroleum Engineers: Bakersfield, California, USA.
17. Ershaghi, I. and D. Abdassah, A Prediction Technique for Immiscible Processes

Using Field Performance Data (includes associated papers 13392, 13793,
15146 and 19506 ). Journal of Petroleum Technology, 1984. 36(4): p. 664-670.
18. Hall, H.N., How to Analyze Waterflood Injection Well Performance. World Oil,
1963: p. 129-130.
19. Silin, D.B., et al., Waterflood Surveillance and Control: Incorporating Hall Plot
and Slope Analysis, in SPE Annual Technical Conference and Exhibition. 2005,
Society of Petroleum Engineers: Dallas, Texas.
20. Ojukwu, K.I. and P.J.v.d. Hoek, A New Way to Diagnose Injectivity Decline
During Fractured Water Injection By Modifying Conventional Hall Analysis, in
SPE/DOE Symposium on Improved Oil Recovery. 2004, Society of Petroleum
Engineers: Tulsa, Oklahoma.
21. Thakur, G.C., The Role of Reservoir Management in Carbonate Waterfloods, in

SPE India Oil and Gas Conference and Exhibition. 1998, Society of Petroleum
Engineers: New Delhi, India.
22. MUSKAT, M., The Flow of Homogeneous Fluids Through Porous Media. Soil
Science, 1938. 46(2): p. 169.
23. Silin, D.B., R. Holtzman, and T.W. Patzek, Monitoring Waterflood Operations:
Hall Method Revisited, in SPE Western Regional Meeting. 2005, Society of
Petroleum Engineers: Irvine, California.
24. Lake, L.W., et al., Optimization of Oil Production Based on a Capacitance

Model of Production and Injection Rates, in Hydrocarbon Economics and
Evaluation Symposium. 2007, Society of Petroleum Engineers: Dallas, Texas,
U.S.A.
25. Sayarpour, M., et al., The Use of Capacitance-Resistive Models for Rapid
Estimation of Waterflood Performance, in SPE Annual Technical Conference
and Exhibition. 2007, Society of Petroleum Engineers: Anaheim, California,
U.S.A.
47
Chapter 3– Producer-Injector Inter-well Connectivity Measurement;
Statistical Approach
3.1 Producer-injector inter-well connectivity
Numerous studies have concluded that connectivity is one of the most important factors
controlling the success of improved oil recovery processes [1]. Inter-well connectivity
evaluation determines how effectively two wells are connected to each other (Figure
3.1). This can provide useful information on reservoir heterogeneity, identify flow
barriers and conduits and provide tools for reservoir management and production
optimization thus it leads to better waterflood management [2].
Figure 3. 1: An example of calculated inter-well connectivity between injectors (blue circles with arrows) and
producers (black circles) [3].
Typically, reservoir description and characterization, together with observation of

injection and production rates, are used to determine the influence of each injector on
producers. Analysis of injection and production data can be particularly useful for
determining connectivity between well pairs, since flow rates are among the most
commonly made measurements made during field life [4].
48
Appropriate determination of this parameter will lead to efficient design of the injection
scheme and infill drilling program [5]. Inter-well connectivity measurement can be
coupled with the performance of the producers, to manage the allocation of water
between injection wells. This will lead to supporting good oil producers and decreasing
the water production in producers with high water cut.
3.2 Inter-connectivity measurement
Modelling the trend of fluid flow between wells in petroleum reservoirs for proper
reservoir management and rate allocation is a complex problem. This is because of the
non-linear nature of the interaction between parameters such as pressure, temperature,
chemical composition of the fluid and reservoir heterogeneity [6].
Probably the most widely available source of data for waterflood management is the
monthly welltest production and injection rates. Useful and valuable information can be
obtained by proper analysis of such data. In general, reservoir description combined
with observation of production and injection history is frequently used to determine the
connection between producers and injectors [7].
Different techniques introduced to quantify communication between wells in the

reservoir. Most of these methods are based on the analysis of injection and production
history data. In this chapter, we will review and evaluate the common statistical
methods that have been developed for analysing the injection and production data.
1. Spearman Rank Correlation

2. Multi-Linear Regression (MLR)
3. Capacitance Resistive Model (CRM)
3.2.1 Spearman rank correlation coefficient
The Spearman rank correlation coefficient [8] (rs) is a function of the sum of the square
of the difference of the two rankings for each observation and the number of the
observations [9]:
n
6
rs = 1 −
n( n 2
− 1)
∑d
i =1
i
2
(Equation 3.1)
49
where di is difference between the rankings of the ith observations and n is the number
of observations. Because rs is a correlation coefficient, it has a value between -1 and +1.
If there is a perfect positive correlation, all the differences will be zero and rs equals +1.
If there is a perfect negative correlation, in which the low-ranking observation in one
classification corresponds to the high-ranking observation in the other, value of the rs
will be -1. If the two ranking sets are independent, this correlation coefficient will be
zero.
The Spearman rank correlation coefficient is a quick, simple, and powerful test of the
existence of the association between variables, regardless of the population distribution
from which the samples are drawn [10].
The Spearman rank correlation coefficient is thus an ideal tool for investigating the
correlation between production and injection rates in a reservoir to determine the
communication between injectors and producers. An increased correlation indicates a
greater connection between an injector and a producer [11].
3.2.2 Multi-linear regression (MLR)
Multi-linear Regression views the reservoir as a system that processes a stimulus

(injection) and returns a response (production). In a waterflood system with multiple
injectors and producers, the effect of this input/output on the reservoir will depend on
the orientation and location of each well. This technique uses different statistical
approaches, based on constrained multi-linear regression, to infer connectivity [7].
The liquid (oil and water) production rates and the water injection rates for every well
that makes up the waterflood system are the main input data. The gas production rate is
not included in the analysis; hence periods with no significant free gas production must
be selected for analysis with this technique.
Production and injection rates in reservoir volumes are used as input to derive the
coefficients for the equation used to estimate the production rate. Two different systems
will be considered [4]:
1. An unbalanced system: injection rate and production rate are different:

I
j 0j ∑
q (t ) = β + β i (t )
i =1
ij i
50
(j = 1,2,…, N) (Equation 3.3)
where q j (t ) is the production rate, ii (t ) is the injection rate, N is the total number
of producers and I is the number of injectors. This equation states that, at any
time, the total production rate at well j is a linear combination of the rates of
every injector plus a constant term, β0j. The factors βij are the weighting factors
and the constant term β0j accounts for the unbalance [4].
Jensen et al. present the solution procedure for the MLR problem. This involves
minimising the variance between the actual production rates and the estimated
one ( q ^j ). This will lead to the following I linear equations [4]:
𝛽1𝑗 2
2
𝜎11 ⋯ 2
𝜎1𝐼 𝜎1𝑗
� ⋮ ⋱ ⋮ �×� ⋮ �=� ⋮ � (Equation 3.4)
2
𝜎𝐼1 ⋯ 𝜎𝐼𝐼2 𝛽𝐼𝑗 𝜎𝐼𝑗2
This can be solved by standard means. The constant β0j is given by:
𝑖=𝐼
𝛽0𝑗 = 𝑞�𝑗 − ∑𝑖=1 𝛽𝑖𝑗 𝚤̅𝑖 (Equation 3.5)
in which 𝑞�𝑗 and 𝚤̅𝑖 are the average values of production and injection rates.
2. A balanced system (BMLR): in which the total field injection rate

approximately equals the total field production rate. The system can be
described by [4]:
I
q j (t ) = ∑ β ij ii (t ) (j= 1,2,… N) (Equation 3.6)
i =1
Again the production rate of producer j is a linear combination of the injection

rates of its associated injectors. In this case, there is no constant term and the
average balanced condition can be given by
− I −
q j (t ) = ∑ βij i i (t ) (j= 1,2,… N) (Equation 3.7)
i =1
Inclusion of the constraints of the BLMR condition, can be achieved by

employing the Lagrange multiplier ( µ j ) in a similar manner to kriging. Then the
following term should be minimised [4]:

 I

Var (q ^j − q j ) − 2 µ j q −j − ∑ λij ii−  (Equation 3.8)
 i =1  51
resulting in the final set of the equation:
2
𝜎112
… 𝜎𝐼12
𝚤̅1 𝛽1𝑗 𝜎1𝑗
⋮ ⋱ ⋮ ⋮ ⎛ ⋮ ⎞
� 2 2 � × �𝛽 � = ⎜𝜎2 ⎟ (Equation 3.9)
𝜎𝐼1 … 𝜎𝐼𝐼 𝚤̅𝐼 𝐼𝑗 𝐼𝑗
𝚤�1 … ��
𝚤𝐼1 0 𝜇 𝑗 𝑞�𝑗
⎝ ⎠
Equation 3.9 can be solved for 𝛽𝐼𝑗 by standard means. Sets of I+1 equations
with I+1 unknowns must be solved for each producer, 𝛽𝑖𝑗 , providing a
quantitative expression of the connectivity between injector and each producer
[4].
3.2.3 Capacitive resistive model (CRM)
The MLR technique does not address the time lag between injection wells and
production wells directly. This should be done before MLR analysis by filtering the
input data. It also neglects the effect of production wells on each other. Yousef et al.
(2005) [12-14] introduced new analytical method called the capacitive resistive model
(CRM) that can quantify inter-well connectivity and the degree of fluid storage available
between injection and production wells [15].
Like MLR, CRM is a non-linear multivariate-regression analysis in which [16], the

reservoir is considered as a system that converts inputs (injection rates) into outputs
(production rates) [17]. Two coefficients are determined for each injector-producer pair:
one parameter (weight) quantifies the connectivity and the other (time constant)
quantifies the degree of fluid storage between the wells [18]. It can therefore provide
good information about the preferential transmissibility trends present within a
waterflood as well as the presence of flow barriers [12].
Mathematical development: The CRM considers the total mass balance of the injection
and produced fluid along with compressibility. A single injector-producer well pair in a
drainage volume is the simplest case. The governing material balance differential
equation at reservoir conditions is introduced by the following equation [13]:
52
−
dP
CtV p = i (t ) − q (t ) (Equation 3.10)
dt
−
In which Ct is the total compressibility; Vp is the drainage pore volume; P is the
average pressure in Vp; i(t) is the total injection rate and q(t) is the total production rate.
According to this equation, at any time, the total rate of mass depletion from the
drainage volume is related to the rate of change of the average pressure within the
volume [13].
In Equation 3.10, it has been assumed that the total compressibility of the reservoir is
small and constant and there is no fluid transfer out of or into the volume Vp. It can also
be derived from a spatial integration of the diffusivity equation, under the same
assumption. With i(t) = 0 this equation is used by Walsh and Lake to describe primary
depletion. The equation is also used to describe the flow of electrical current in a
resistor-capacitance network, which has the same form, hence the term capacitance in
the description [14].
Equation 3.11 introduces the linear productivity index model that helps to describe the
the system, based entirely on the rates [13]:
− 
=
q (t ) J  P r − Pwf  (Equation 3.11)
 
where pwf and J are the flowing bottom hole pressure (BHP) and productivity index of
the producer, respectively. Equation 3.11 assumes stabilized flow, which is unlikely to
be accurate in conditions where rates are constantly changing. However, the
productivity index (and equivalent alternative definitions) is almost universally applied
in describing well performance, in practice [14].
Eliminating the average pressure between Equations 3.10 and 3.12 give:
dq dPwf
τ + q (t ) =i (t ) − τ J (Equation 3.12)
dt dt
where τ is the "time constant" of the drainage volume, and is defined by:
53
CtV p
τ= (Equation 3.13)
J
The differential equation (Equation 3.13) being derived from the linear equation of
productivity index and material balance equation, respects the following assumptions
[19]:
1. Slightly compressible fluid

2. Instantaneous equilibrium
3. Immiscible phase
4. Constant temperature
5. Constant productivity
The solution to the above equation is [13]:
 t −t  ξ ξ
−t −t
− 0  t t
1 dpwf
q (t ) =
q (t0 )e  τ 
+e
τ ∫ eτ
t0
τ
i (ξ )d ξ + e
τ ∫ eτ Ji(ξ )
t0
dξ
(Equation 3.14)
This means that the production rate is affected by [14]:
1. Primary production
2. Contribution from injection
3. Change in bottomhole flowing pressure
Sayarpour et al. [20] introduced analytical solutions for the fundamental differential
equation of the CRM, based on superposition in time and presented these solutions for
three different reservoir control volumes: 1) drainage volume of each producer (CRMP),
2) volume of the entire field or tank model (CRMT), and 3) drainage volume between
each injector-producer pair (CRMIP) [15].
1. One time constant for each producer (CRMP): For a pattern of I number of
injectors and N number of producers, Figure 3.2 represents the in-situ volumetric
balance over the effective pore volume of a producer. Sayarpour et al. derived
analytical solutions for two cases: a linear variation of bottom-hole flowing
54
pressure (BHP), but with stepwise changes in injection rate, and a linear
variation of both injection rate and BHP during consecutive time intervals [19].
a) For a case of fixed injection rate of ii (∆tk ) = I (k)

i , and a linear BHP
variation during time intervals ∆tk , (k=1,2,…,n), by assuming a constant
productivity index at time tn , the total production rate of producer j can be
written as [19]
− ( t n − t0 )
n I ∆P ( k ) 
−∆tk
 −(tτn −tk ) 
+ ∑  ∑ fij I ijk − J jτ j wf , j
τj τj
q j (tn ) = q j (t0 )e  1 − e e j 
=  i 1
k 1= ∆tk   
 
 
(Equation 3.15)
where I ijk and ∆Pwf( k,)j represent the injection rate of injector i and the changes in
BHP of producer j during the time interval t k-1 to t k , respectively. Equation 3.15
effectively assumes that the variation of injection rates is stepwise.
Figure 3. 2: Schematic representation of control volume for producer j, CRMP [21].
55
b) If there is a linear change between two consecutive injection rates and BHP
during time intervals t k-1 to t k, at time tn, the total production rate of producer j
can be written as [19]
− ( t n − t0 )
I   − ( t n − t0 )

+ ∑ fij ii (tn ) − e
  ii (t0 )   −
τj τj
q j (tn ) = q j (t0 )e
i =1   
  
(Equation 3.16)
 I
n ∆Pwf( k,)j 
−∆tk
 −(tτn −tk ) 
∑  ∑ f ij I ij − J jτ j τj
k
 1 − e τ j e j 
=  i 1
k 1= ∆tk   
 
 
2. One time constant for field (CRMT): A reservoir may be represented by a

single producer and a single injector, as a tank. This is achieved by combining
all production and injection rates (Figure 3.3). Hence, the CRMP solution can be
applied by equating fij to unity to arrive at the tank solution (designated as
CRMT). Therefore [19]:
n  
− ( t n − t0 ) − n k  
 t −t  −∆tk
=
qF (tk ) qF (t0 )e + ∑  I Fk e   1 − e τ F  (Equation 3.17)
τF k =1 
 
  
in which I F (t ) and qF (t k ) are the total field injection and production rates. τf is
the field-time constant and, hence, represents the field average properties [19].
Figure 3. 3: Schematic representation of a field with one injector and one producer, CRMT [21].
56
CRMT does not account for the variation of bottom-hole pressures in individual
wells, therefore only two system parameters, initial production rate and field
time constant, are sought [19].
3. One time constant for each injector-producer (CRMIP): The volumetric

balance in reservoir conditions over the affected pore volume injector/producer
pair, ij, is shown in Figure 3.4. Similar boundary conditions as those discussed
above for CRMP, may be used to obtain analytical solutions, by assuming a
linear variation of a) BHP, while the injection rate is changing in steps, and b)
both injection rate and BHP during consecutive time intervals [20].
a) The total production rate of producer j for the case of fixed injection rate and
a linear BHP variation during time intervals, at time tn, can be written as [20]:
I
− ( t n − t0 )
I  n  I ∆Pwfj  
−∆tk
 −(tτn −tk )  
∑q + ∑ ∑  ∑ fij I i − J ijτ ij
  e ij  
τ ij τj
q j (tn )=  1 − e
k
(t0 )e
ij
 k 1 = ∆tk   
   
=i 1 =i 1 = i 1

(Equation 3.18)
b) For a case of linear change between two consecutive injection rates and
producer’s BHP during time interval ∆tk , the total production rate of producer j
at time tn can be written as [20]:
− ( t n − t0 )
− ( t n − t0 )
I  I 
∑ qij (t0 )e + ∑ f ij ii (tn ) − e
τ ij τ ij
q j (t=
n)
 i0 (t0 ) 
=i 1 =i 1  
 
(Equation 3.19)
I  n  f ∆i k ∆Pwfjk    −(tτn −tk )  
−∆tk
−∑ τ ij ∑ 
  e ij  
τ ij
 ij i
−J  1 − e
   ∆ t ∆ t    
=i 1 =

k 1
 k k   
57
Figure 3. 4: Schematic representation of control volume between each injector/producer pair [21].
3.3 Case study
The same model as used for evaluating the performance of waterflooding monitoring
techniques in Chapter 2 of was used to study the application of these techniques for
inter-well connectivity measurements. The model has been described in section 2.2.1
(Figure 3.5 and Table 3.1).
Figure 3. 5: Reservoir model (colours represent the porosity).

58
Table 3. 3: Properties of the reservoir model.
Size of the cells 20*15*4

Initial pressure 4000 psia
Temperature 100o F
Permeability 220-320 mD
Porosity (fraction) 0.2-0.25
Oil density 35 API
The reservoir has been producing for 20 years with the pressure support supplied by
water injection from the beginning of production. The producers were controlled by a
constant surface liquid production rate and injectors set to inject by constant voidage
replacement ratio (VR). The sum of the all VRs is unity i.e. total volume of injected
water equals the total produced fluid volume.
Downhole injection and production rates were used as input for all these methods to
quantify the connection between wells (Figures 3.6 and 3.7). Then the results for
connectivity were combined with the performance of each producer to define the
injection allocation factor for each injector.
9200
Injector 1
9000
Injection rate (bbl/day)
Injector 2
8800 Injector 3
8600 Injector 4
8400
8200
8000
7800
0 2000 4000 6000 8000
Time (days)
Figure 3. 6: 20-year down-hole injection rate history of all the injectors from the reservoir model.
59
15000
Production rate (bbl/day)

14000
13000
Producer 1
12000
Producer 2
11000
Producer 3
10000
9000
0 2000 4000 6000 8000
Time (days)
Figure 3. 7: 20-year down-hole production rate history of all the producers from the reservoir model.
3.3.1 Spear Rank correlation coefficient
Data used for this analysis are the injection and production rates in reservoir barrel per
day. SPSS software is used to calculate the Spearman Rank Correlation coefficient
between the each injector/producer pair, to establish the dominant communication
trends in the reservoir.
The calculations were based on the liquid production rates of both oil and water. The
rates were converted to ranks, and the Spearman rank correlation coefficient was
calculated for pairs composed of each injection well and its adjacent production wells.
The results are shown in Figures 3.7 and 3.8.
60
0.6 Producer 1
Producer 2
0.5
Spearman rank correlation Producer 3
coefficient 0.4
0.3
0.2
0.1
0
Injector 1 Injector 2 Injector 3 Injector 4
Figure 3. 8: Calculated Spearman correlation coefficients for each injector and its associated producers.
Analysis shows that producer 2 has the highest connectivity with the injectors while
producer 3 has the lowest value. And also it can be seen that injector number three has
the highest correlation with each producer while injector 2 has the minimum correlation,
although the difference is not significant.
0.6 Injector 1
Injector 2
0.5 Injector 3
Spearman rank correlation
Injector 4
0.4
Coefficient
0.3
0.2
0.1
0
Producer 1 Producer 2 Producer 3
Figure 3. 9: Calculated Spearman correlation coefficients for each producer and its associated injectors.
61
3.3.2 Multi-linear regression
Balanced MLR can also be used to determine the well connectivity, since the injection
started at the start of production from the reservoir, while the voidage replacement ratio
was one. Equation 3.6 implies that the rate of production from each producer will be a
linear combination of the rate of the injectors. Therefore, our reservoir will have 3
producers and 4 injectors; for each producer, the rate of liquid production is:
q j (t ) = β1 j i1 (t ) + β 2 j i2 (t ) + β 3 j i3 (t ) + β 4 j i4 (t ) (Equation 3.19)
The above equation illustrates one of the most important challenges in managing
waterflood performance. Is the producer connected to all injectors or just connected to
some of them? The answer to this question has a significant effect on the calculated
coefficient of Equation 3.19.
For example, let us assume two cases for Producer 1. Case (1) supposes that Producer
1 is connected to all the injectors, while in Case (2) two we assume that it is connected
to Injectors 1, 2 and 3 only. Figure 3.10 shows the results of calculated connectivity
measurement. It can be seen that there is a significant difference between the calculated
coefficients for Injectors 1 and 2, which will affect the measurement of the injection
allocation factor.
The superposition approach is proposed to determine which injector is connected to a

specified producer. The methodology is as follows:
1. The two injectors closest to the producer are assigned to that producer. The
coefficients of the MLR equation are then calculated along with a measurement
of the estimation error.
2. The next nearest injector is then assigned to the system of one producer and the
two injectors. The MLR coefficients and the error calculation are repeated.
3. An injector will be assigned to the producer if and only if the calculated error in
step 2 is less than in step 1, otherwise it will be considered that the injector is not
connected to the producer.
62
0.7
Producer connected to all injectors
Calculated Inter-well connectivity from MLR

0.6 Producer connected to injector 1,2&3
0.5
0.4
0.3
0.2
0.1
0
I1 I2 I3 I4
Injectors
Figure 3. 10: Calculated inter-well connectivity for Producer 1, assuming two different injector-producer
combinations.
Figure 3.11 represents the application of this procedure for Producer 1, where the
nearest injectors to this producer are Injectors 1& 2.
0.4315
0.431
Error of estimation (%)
0.4305
0.43
0.4295
0.429
0.4285
0.428
Injector 1&2 Injector 1,2&3 Injector 1,2,3&4
Figure 3. 11: Calculated error of estimation of production rate of Producer 1 for different combinations of
injectors.
Adding Injector 3 to the system improves the production rate estimation, while the
further addition of Injector 4 increases the error. This implies that Producer 1 is
connected to Injectors 1, 2 and 3 but not Injector 4.
63
The above procedure was applied for all producers. The results are summarised in
Figures 3.12 and 3.13 and Table 3.2.
0.395
0.39
0.385
0.38
0.375
0.37
0.365
0.36
0.355
0.35
Injector 4&3 Injector 4,3&1 Injector 4,3&2
Figure 3. 12: Calculated estimation error of the production rate of Producer 2 for different combinations of
injectors.
0.44
0.43
0.42
0.41
0.4
0.39
0.38
0.37
Injector 2&3 Injector 1,2&3 Injector 2,3&4
Figure 3. 13: Calculated estimation error of the production rate of Producer 3 for different combinations of
injectors.
Table 3. 4: This table shows which injectors are connected to each producer (√ means connected).

Producer 1 √ √ √ ×
Producer 2 × × √ √
64
Table 3.2 shows that Producer 1 and Producer 2 are connected to Injectors 1,2 and 3,
while Producer 3 is supported by Injectors 3 and 4. The calculated MLR connectivity
coefficients for each producer are shown in Figures 3.14, 3.15 and 3.16. These
coefficients represent the inter-well connectivity between injectors and producer.
Inter-well connectivity measurement based on MLR shows Injector 1 has more

connection to Producers 1 and 3 compared to Producers 3 and 2, while Producer 2 is
highly affected by Injector 4.
Injector 3 is the only injector that is connected to all the producers but its connection to
the production wells is not high, compared to the other injectors.
Producer 1
0.7
0.6
0.5
MLR coefficients
0.4
0.3
0.2
0.1
0
65
Producer 2
0.9
0.8
0.7
MLR coefficients 0.6
0.5
0.4
0.3
0.2
0.1
0
Producer 3
0.45
0.4
0.35
0.3
MLR coefficients
0.25
0.2
0.15
0.1
0.05
0
The following figures (Figure 3.17 to 3.19) represent the estimated production rate
using the MLR technique for each producer versus the production history obtained from
the 20-year reservoir simulation model over the 20-year period. As can be seen, there is
a good match between these estimates and the production rate from the simulation.
66
Producer 1
10800
Production history
10600
Estimated production rate by MLR
10400
10200
10000
9800
9600
9400
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3. 17: Calculated production rate from MLR versus production history from the reservoir simulation
model for Producer 1.
Producer 2
10800
Production history
10600
10400 Estimated production rate by MLR

10200
10000
9800
9600
9400
9200
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3. 18: Calculated production rate from MLR versus production history from the reservoir simulation
67
Producer 3
14800
Production history
14600
Estimated production rate by MLR
14400
14200
14000
13800
13600
13400
13200
13000
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3.19: Calculated production rate from MLR versus production history from the reservoir simulation
3.3.3 Capacitive resistive model
History matching and optimization are two important steps of determining parameters
of CRM equation. Minimization of the average-absolute error over the total production
history is a straightforward way to evaluate model parameters. Since waterflood
management by reallocating injected water can be obtained from CRMP (Equation
3.18), CRMP have been used in this study.
CRMP requires (I +3) number of model parameters for each producer: f1j , f2j ,…, fI,j , j ,
qj (to ) and Jj ; hence, application of CRMP requires a minimum of P× (I +3) injection
and production data points. The average absolute of the error for each of the producers
can be evaluated and the sum of these errors becomes the objective function [13].
An optimization procedure is required to determine the optimum solution of fij and τj. In
the optimization, τj s are set to be the free parameters and the objective function is to
minimize the squared errors between measured production rates and those generated by
Equation 3.18. For a given set of τj s Equation 3.18 becomes linear and therefore, MLR
68
is used to determine fij. After iterating in τj s, the optimum set of fij is obtained. The
optimum τj and fij are therefore obtained at the end of this procedure. Relying as it
does on linear regression, this procedure, allows us to use the error estimates of the
weights based on MLR[13].
The procedure employed to find which injector is connected to which producer similar
to that was used for the MLR problem. Table 3.2 shows that the CRMP connectivity
results are similar to those obtained for MLR. Figures 3.20 to 3.22 present the calculated
fij for each producer.
Producer 1
0.7
0.6
CRM injection well coefficient
0.5
0.4
0.3
0.2
0.1
0
I1 I2 I3 I4
Figure 3. 19: Calculated values of fij of the CRM equation for Producer 1.
Producer 2
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
I1 I2 I3 I4
69
Producer 3
0.5
0.45

0.4
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0
I1 I2 I3 I4
The determined values of the time constant for each producer can be seen in Figure
3.23. Based on values of τj, we can say Producer 3 has the highest drainage volume
compared to the other producers and Producer 2 has the lowest one. Figure 3.24 that
shows the top view of the drainage volume of each producer qualitatively confirms the
analysis of the τj values.
180
160
140
CRM time constant
120
100
80 Ƭ
60
40
20
0
P1 P2 P3
Figure 3. 22: Measured values of time constant for each producer.
70
Figure 3. 23: Drainage volume of each producer calculated from reservoir simulator.
The estimated values of the liquid production rate from CRMP versus the values from
simulation are shown in Figures 3.25 to 3.27. As can be seen, there is an acceptable
match between estimated production and production history.
71
Producer 1
10800
Production history
10600 Estimated Production rate by CRM
10400
10200
10000
9800
9600
9400
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3. 24: Calculated production rate from CRMP versus production history of Producer 1.
Producer 2
10800 Production history
10600 Estimated Production rate by CRM

10400
10200
10000
9800
9600
9400
9200
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3. 25: Calculated production rate from CRMP versus production history of Producer 2.
72
Producer 3
14800
Production history
14600
Estimated Production rate by CRM
14400
14200
14000
13800
13600
13400
13200
13000
0 1000 2000 3000 4000 5000 6000 7000 8000
Time (days)
Figure 3. 26: Calculated production rate from CRMP versus results production history of Producer 3.
3.4 Allocation management
In this section a simple new approach is introduced, in which the calculated inter-well
connectivity results are used to determine the allocation factor for each injector. The
aim is to allocate more water to those injectors that are supporting the better producers
i.e. production wells with the highest oil production performance. This requires
evaluating the performance of each producer and combining the results of the
connectivity measurement in order to optimise the distribution of water between the
injectors. So in this new methodology, connectivity results have been coupled with the
parameters describing the production well performance, to manage the water allocation.
We will start with one of the most common parameters in petroleum engineering for
describing the performance of the production wells, which is the water cut (WC).
However, WC should be considered in conjugation with the liquid production ratio (LR)
of the producer (Equation 3.20) to compare the performance of each individual producer
with other production wells.
qlj
LR = (Equation 3.20)
qlf
73
in which qlj is the liquid production of the well and q lf is the total liquid production of
the reservoir. Figure 3.28 shows the WC of each producer after 20 years of production
and Figure 3.29 will shows the LR of each producer.
0.8
0.7
0.6
Water cut (fraction)
0.5
0.4
0.3
0.2
0.1
0
P1 P2 P3
Figure 3. 27: WC of each producer at the end of 20 years of production.
0.45
0.4
0.35
Liquid ratio (fraction)
0.3
0.25
0.2
0.15
0.1
0.05
0
P1 P2 P3
Figure 3. 28: Calculated LR of each producer at the end of 20 years of production.
Taking LR into account, the oil production performance index of each producer (OI) can
be defined as:
OI =−
(1 WC ) × LR (Equation 3.21)
74
The calculated OI for each producer is shown in Figure 3.29.
0.25
0.2
0.15
Oil index
0.1
0.05
0
P1 P2 P3
Figure 3. 29: Calculated OI of each producer at the end of 20 years of production.
OI is the parameter that describes the performance of each producer. A higher OI

means a better oil producer. In this example, Producer 3 has the best OI, while Producer
2 has the lowest one. For each injector, the total water allocation index (AFt) is defined
as:
P
=
AFti ∑ (OI
j =i
j × Cij ) (i=1,2,…,I) (Equation 3.22)
In the above equation AFti is the total water allocation index for injector i. OIj is the oil
production performance index of producer j and Cij is the calculated inter-well
connectivity from previous sections for a paired producer j and injector i. The final
water allocation factor (AF) for each injector will be:
AFti
AFi = I
(Equation 3.24)
∑ AF
i =i
ti
The calculated AFs for each injector, based on the determined connectivity
measurement from different techniques, are shown in Figure 3.30.
75
0.45
0.4
0.35
Water allocation factor

0.3
I1
0.25
I2
0.2
I3
0.15
I4
0.1
0.05
0
SRC MLR CRM
Figure 3. 30: Determined allocation factor for each injection well from different statistical methods.
In the final stage of this chapter, the calculated AFs are used to manage the allocation of
water between injectors for the next 20 years of production. The following figures show
the results of each technique.
Figure 3. 31: Comparison of improvement in oil production rate after WAM based on different statistical
techniques.
76
Figure 3. 32: Comparison of improvement in cumulative oil production after WAM based on different
statistical techniques.
Figure 3. 33: Comparison of reduction in cumulative water production after WAM based on different
statistical techniques.
77
The results of simulation after 40 years show improvement in waterflood performance
after water allocation management. A better daily oil production profile was achieved
(Figure 3.31), more cumulative oil produced (Figure 3.32) and water production
decreased (Figure 3.33). All these were obtained while the same amount of water was
injected into the reservoir. This means that water allocation management based on inter-
well connectivity measurement and production well performance evaluation will
improve the sweep efficiency in a water injection project.
However, only a very small improvement was obtained from the Spearman rank
correlation. As discussed earlier in this chapter, in section 3.3.1, this technique cannot
properly predict the connection between individual injectors and producers. It will only
give an idea of how the total injection system is working. For example the highest
values of this coefficient are related to Producer 2 (Figures 3.8 and 9), the production
well that is producing more water than the others.
The results of MLR and CRM were almost the same. The proposed technique for finding
the injectors connected to the producer was effective and significant improvement was
achieved in overall performance of water injection. CRM works a little better, as it has a
time constant that will take into account the lag time between producers and injectors.
Although a new technique has been developed for water allocation, it needs to be
pointed out that, in reality, there are other parameters that put constraints on the
allocated injection rate. For example the injection rate should match the outflow
performance of the injection well, compressor and pump capacity and in most cases the
injection pressure should be less than the matrix fracture pressure.
OI was a good parameter to describe the performance of the producer but we are still
interested in finding another parameter that can differentiate more clearly between a
good and bad producer. In addition, it should be a good representative of the production
history. This will be briefly discussed in the Chapter 5.
78
3.6 References
1. Lee, K.-H., et al., A Multivariate Autoregressive Model for Characterizing

Producer-producer Relationships in Waterfloods from Injection/Production
Rate Fluctuations, in SPE Western Regional Meeting. 2010, Society of
Petroleum Engineers: Anaheim, California, USA.
2. Aggrey, G.H. and D.R. Davies, Real-Time Water Detection and Flow Rate
Tracking in Vertical and Deviated Intelligent Wells with Pressure Sensors, in
Europec/EAGE Conference and Exhibition. 2008, Society of Petroleum
Engineers: Rome, Italy.
3. Honarpour, M.M. and L. Tomutsa, Injection/Production Monitoring: An

Effective, Method for Reservoir Characterization, in SPE/DOE Enhanced Oil
Recovery Symposium. 1990, 1990: Tulsa, Oklahoma.
4. Albertoni, A. and L.W. Lake, Inferring Interwell Connectivity From Well-Rate

Fluctuations in Waterfloods, in SPE/DOE Improved Oil Recovery Symposium.
2002, Copyright 2002, Society of Petroleum Engineers Inc.: Tulsa, Oklahoma.
5. Zhai, D. and J.M. Mendel, Robust Production-Rate Interpolation in Waterflood

Management, in SPE Western Regional Meeting. 2012, Society of Petroleum
Engineers: Bakersfield, California, USA.
6. Kaviani, D., et al., Estimation of Interwell Connectivity in the Case of

Fluctuating Bottomhole Pressures, in Abu Dhabi International Petroleum
Exhibition and Conference. 2008, Society of Petroleum Engineers: Abu Dhabi,
UAE.
7. Albertoni, A. and L.W. Lake, Inferring Interwell Connectivity Only From Well-
Rate Fluctuations in Waterfloods. SPE Reservoir Evaluation & Engineering,
2003. 6(1): p. 6-16.
8. Sunbul, A.H., et al., The Evolution of Advanced Well Completions To Enhance

Well Productivity and Recovery in Saudi Aramco's Offshore Fields, in SPE
Middle East Oil and Gas Show and Conference. 2007, Society of Petroleum
Engineers: Kingdom of Bahrain.
9. Muradov, K.M. and D.R. Davies, Application of Distributed Temperature

Measurements to Estimate Zonal Flow Rate and Pressure, in International
Petroleum Technology Conference. 2011, International Petroleum Technology
Conference: Bangkok, Thailand.
10. Aggrey, G.H., D.R. Davies, and L.T. Skarsholt, A Novel Approach of Detecting
Water Influx Time in Multizone and Multilateral Completions Using Real-Time
Downhole Pressure Data, in SPE Middle East Oil and Gas Show and
Conference. 2007, Society of Petroleum Engineers: Kingdom of Bahrain.
79
11. Muradov, K.M. and D.R. Davies, Zonal Rate Allocation in Intelligent Wells, in
EUROPEC/EAGE Conference and Exhibition. 2009, Society of Petroleum
Engineers: Amsterdam, The Netherlands.
12. Yousef, A.A., L.W. Lake, and J.L. Jensen, Analysis and Interpretation of
Interwell Connectivity From Production and Injection Rate Fluctuations Using
a Capacitance Model, in SPE/DOE Symposium on Improved Oil Recovery.
2006, Not subject to copyright. This document was prepared by government
employees or with government funding that places it in the public domain.:
Tulsa, Oklahoma, USA.
13. Yousef, A.A., et al., A Capacitance Model To Infer Interwell Connectivity From
Production and Injection Rate Fluctuations, in SPE Annual Technical
Conference and Exhibition. 2005, Society of Petroleum Engineers: Dallas,
Texas.
Production- and Injection-Rate Fluctuations. SPE Reservoir Evaluation &
Engineering, 2006. 9(6): p. pp. 630-646.
15. Nguyen, A.P., et al., Integrated Capacitance Resistive Model for Reservoir
Characterization in Primary and Secondary Recovery, in SPE Annual Technical
Conference and Exhibition. 2011, Society of Petroleum Engineers: Denver,
Colorado, USA.
16. Izgec, O. and C.S. Kabir, Understanding reservoir connectivity in waterfloods

before breakthrough. Journal of Petroleum Science and Engineering, 2010.
75(1–2): p. 1-12.
17. Liang, X., A simple model to infer interwell connectivity only from well-rate
fluctuations in waterfloods. Journal of Petroleum Science and Engineering,
2010. 70(1–2): p. 35-43.

U.S.A.
19. Sayarpour, M., et al., The use of capacitance–resistance models for rapid
estimation of waterflood performance and optimization. Journal of Petroleum
Science and Engineering, 2009. 69(3–4): p. 227-238.
20. Sayarpour, M., et al., The Use of Capacitance-Resistive Models for Rapid
Estimation of Waterflood Performance, in SPE Annual Technical Conference
and Exhibition. 2007, Society of Petroleum Engineers: Anaheim, California,
U.S.A.
21. Sayarpour, M., C.S. Kabir, and L.W. Lake, Field Applications of Capacitance-
Resistance Models in Waterfloods. SPE Reservoir Evaluation & Engineering,
2009. 12(6): p. pp. 853-864.
80
Chapter 4– Producer-Injector Inter-well Connectivity Measurement;
Artificial Intelligence Techniques
Neural networks, non-algorithmic, non-digital, intensely parallel and distributive

information processing systems, are being used more and more every day. The main
interest in neural networks is rooted in the recognition that the human brain processes
information in a different manner to conventional digital computers. Artificial neural
networks (NNs) are specialized techniques that generate strategies to map input to
output data. Artificial neural networks are routinely used in complex time series
prediction. A typical ANN involves processing patterns that evolve over time [1].
In this chapter, first a brief discussion of the artificial neural networks is presented
followed by strategies to design a network for fluid flow simulation and prediction of
well interaction in heterogeneous permeable media. The results are then presented and
conclusions drawn based on these results.
4.1 Artificial neural networks (ANNs)
Artificial Neural Networks (ANNs) are computational modelling tools that have recently
emerged and found extensive acceptance in many disciplines for modelling complex
real-world problems [2]. However, if a problem is solvable by conventional methods,
neural networks (or any other virtual-intelligence technique) should not be used to solve
it. Although there is academic value to solving simple problems, such as polynomials
and differential equations, with neural networks to show their capabilities, they should
be used mainly to solve problems that otherwise are very time-consuming or simply
impossible to solve by conventional methods [3].
ANNs may be defined as structures comprised of densely interconnected adaptive

simple processing elements (called artificial neurons or nodes) that are capable of
performing massively parallel computations for data processing and knowledge
representation [4]. These networks are physical cellular systems that can acquire, store,
and use experiential knowledge. The knowledge is in the form of stable states or
81
mapping embedded in networks that can be recalled in response to the presentation of
cues [5].
Much of the interest in neural networks arises from their ability to discover the
underlying system by developing a function between input and output vectors on the
basis of historical data. Neural networks accumulate the knowledge implicitly in
connection weights between the layers. As a consequence, the knowledge can be
modified by changing the weights through back-propagation. With the back-
propagation rule, also referred as delta learning rule, the network first uses the input
vector to produce an output, and compares this output to the desired output. In the case
there is a difference, the weights are modified between the layers to further decrease the
difference. This continues until the minimum desired error rate is obtained between the
network produced and actual output [6].
4.1.1 Structure of ANNs
Artificial neural networks are information-processing systems that are a simplified

simulation of human biological process and have the same performance characteristics
as those of biological neural networks inside human body. The development of ANNs
as generalizations of mathematical models of human cognition or neural biology is
based on the these assumptions [3].
1. Information processing occurs in many neurons or simple elements that are

sometimes called processing elements (PE) (Figure 4.1) [3].
2. Signals are passed between neurons over connecting links [3].
3. Each connecting link has an associated weight, which, in a typical neural
network, multiplies the signal being transmitted [3].
4. The output signal of each neuron is determined by an activation function
(usually nonlinear) which applied to its net input [3].
82
Figure 4. 1: Structural model of an artificial neuron[7].
4.1.2 Classification of ANNs
ANNs can be classified in many different ways, according to one their relevant features.
In general, classification of ANNs may be based on (i) the function of the ANN (e.g.,
pattern association, clustering), (ii) the degree (partial/full) of connectivity of the
neurons in the network, (iii) flow direction of information within the network (recurrent
and non-recurrent), with recurrent networks, which are dynamic systems in which the
state at any given time is dependent on previous states, (iv) the learning algorithm type,
which represents a set of systematic equations that employ the outputs obtained from
the network along with an arbitrary performance measure to update the internal
structure of the ANN, (v) the learning rule (the driving engine of the learning algorithm),
and (vi) the degree of learning supervision needed for training the ANN. Supervised
learning will train an ANN when the correct answers (i.e. target outputs) are provided
for every example, and the solution of the ANN is compared to the corresponding target
values to determine the required amount by which each weight should be adjusted.
Reinforcement learning is supervised; however, the ANN is provided with a critique on
correctness of output rather than the correct answer itself. Unsupervised learning does
not require a correct answer for training, however the network, through exploring the
underlying structure in the data and the correlation between the various examples,
organizes the examples into clusters (categories) based on their similarity or
dissimilarity. Finally, the hybrid learning procedure combines supervised and
unsupervised learning [2].
83
4.1.3 Application of ANN
ANNs have been utilized in a variety of applications, such as modelling, classification,

pattern recognition, and multivariate data analysis.
4.1.3.1 Pattern classification
Pattern classification will use supervised learning to assign an unknown input pattern,
to one of several pre-specified classes based on one or more properties that characterize
a given class, as shown in Figure 4.2 [2].
Figure 4. 2: An example of pattern classification by ANN [2].
4.1.3.2 Clustering
Clustering can be performed by unsupervised learning in which the clusters (classes) are
formed by exploring the similarities or dissimilarities between the input patterns
according to their inter-correlations (Figure 4.3). The network will assign ‘similar’
patterns to the same cluster [2].
84
Figure 4. 3: An example of clustering by ANN[2].
4.1.3.3 Function approximation (modelling)
Function approximation (modelling) concerns training the ANN on input-output data in

order to explore the underlying rules relating the inputs to the outputs (Figure 4.4).
Multilayer ANNs are considered as universal estimators that can approximate any
arbitrary function to any degree of accuracy [8], and therefore are normally employed in
this application. This type of ANN is applied to problems (i) where there is no
theoretical model, i.e. data obtained from experiments or observations are utilized, or
(ii) to substitute theoretical models that are hard to compute analytically by utilizing
data obtained from such models [2].
Figure 4. 4: An example of function approximation by ANN[2].
85
4.1.3.4 Forecasting
Forecasting includes training of an ANN on samples from a time series representing a

certain phenomenon at a given scenario and then using it for other scenarios to predict
(forecast) the behaviour at subsequent times (Figure 4.5). That is, the network will
predict Y(t + 1) from one or more previously known historical observations [e.g., Y(t -
2), Y(t - 1), and Y(t), where t is the time step [2].
Figure 4. 5: An example of forecasting by ANN [2].
4.1.4 Application of ANN in petroleum engineering
Applications of ANNs in petroleum engineering can be divided into two categories:

those that use neural networks to analyse formation lithology from well logs and those
that use neural networks to pick a reservoir model to be used in conventional well test
interpretation studies. These tasks are usually done by log analysts and reservoir
engineers, and their automation using a fault-tolerant process may prove valuable[5].
Neural networks can help engineers and researchers by addressing some fundamental
petroleum engineering problems as well as specific ones that conventional computing
has been unable to solve. Petroleum engineers may benefit from neural networks on
occasions when engineering data for design and interpretations are less than adequate
[5].
Neural networks have proved to be valuable pattern-recognition tools. They are capable
of finding highly complex patterns within large amounts of data. A relevant example is
86
well log interpretation. It is generally accepted that there is more information embedded
in well logs than meets the eye. Thus, determination, prediction, or estimation of
formation permeability without actual laboratory measurement of the cores or
interruption in production for well test data collection has been a fundamental problem
for petroleum engineers [5].
Neural networks have shown great potential for generating accurate analysis and results
from large historical databases, the kind of data that engineers may not consider
valuable or relevant in conventional modelling and analysis processes. Neural networks
should be used in cases where mathematical modelling is not a practical option. This
may be because all the parameters involved in a particular process are not known and/or
the interrelation of the parameters is too complicated for mathematical modelling of the
system. In such cases, a neural network can be constructed to observe the system
behaviour (what type of output is produced as a result of certain set of inputs) and try to
mimic its functionality and behaviour [3].
4.2 ANN and inter-well connectivity measurement
Consider a pair of injection-production wells in a homogeneous cross section where qw

is the injection rate of water and ql is the liquid production rate. For this simple system
we can express ql as:
ql = Z(qw) (Equation 4.1)
where Z is the transfer function which models the fractional flow characteristics of the
medium. For a simple system, as in Equation 4.1, an ANN can easily be designed by
using the injection history as the input data to the network and production rates as the
output. [1] Therefore, an ANN can be designed to forecast the liquid production of a
production well by inputting the injection rate of the surrounding injectors. By
sequentially varying the injection rates of the injectors around a target well one can
determine the relative influence of each of the injectors surrounding a target well [1].
In this section we will give a short introduction to two types of ANNs used in this
research.
87
4.2.1 Feed-forward back propagation (FFBP)
Feed-forward networks are a generalization of the multi-layer perceptron (MLP), such

that connections can jump over one or more layers. In theory, an MLP can solve any
problem that a generalized feed-forward network can solve. In practice, however,
generalized feed-forward networks often solve the problem much more efficiently. A
classic example of this is the two spiral problem. A standard MLP requires several
hundred times more training steps than a generalized feed-forward network with the
same number of processing elements.
4.2.2 Fuzzy logic network (Co-adoptive neuro-fuzzy inference system CANFIS)
The CANFIS (Co-Active Neuro-Fuzzy Inference System) model integrates adaptable

fuzzy inputs with a modular neural network to rapidly and accurately approximate
complex functions [9]. Fuzzy inference systems are also valuable as they combine the
explanatory nature of rules (membership functions) with the power of "black box"
neural networks.
4.3 Case Study
The same reservoir model as that used in Chapter 2 (Section 2.2.1) has been employed
in this analysis.
As previously described, inter-well connectivity is determined based on 20 years of

injection and production rate history (Figures 3.6 and 7). The injection water is then re-
allocated to the injectors for the next 20 years of production, based on the results of the
connectivity calculation.
Both types of ANNs represented above (FFBP and CANFIS) were employed.
The statistical techniques described in Chapter 3 allowed direct calculation of inter-well

connectivity coefficients. However, these coefficients cannot be determined directly
when employing ANNs. This difficulty was overcome by:
1. An ANN is designed for each producer in order to estimate the rate of liquid
production based on the injection rates of the connected injectors.
88
2. The designed ANN is used to run a sensitivity analysis to monitor the effect of
the change in the injection rates of the connected injectors on the production
rate of the producer. The results of this sensitivity analysis will show the inter-
well connectivity between the producer and its surrounding injectors.
3. At the end, an allocation factor will be determined for each injector, based on its
connection to the producers and the producing water cut of the production wells
4.3.1 Feed-forward back propagation (FFBP)
The process employed in the application of FFBP will be described in detail since it is
the first ANN technique being employed.
4.3.1.1 Designing the optimum network
The main steps in designing and developing an efficient neural network for each
producer are:
1. Defining the base design

2. Preparing a data set for training the base case
3. Carrying out super-position analysis to identify which injector is connected to
each of the production wells.
A. Base case design
The initial base case is designed based on the number of inputs and outputs and the
volume of data available for each parameter. The network has one output, the liquid
production rate of the producer. The number of inputs depends on how many injectors
are connected to each producer. For the base case, it has been assumed that the producer
is only supported by the nearest two injectors. The properties of the base case network
for each producer are given in Table 4. 1.
89
Table 4. 1 Initial properties of the FFBP network designed to estimate liquid production rate of producer
number one.
Number of Hidden layers 1
Number of processing elements (PE) 4
Hidden layer transform function Sigmoid
Output layer transform function Sigmoid
Number of epochs for training 1000
The transfer (activation) function [10] is necessary to transform the weighted sum of all
signals impinging onto a neuron so as to determine its firing intensity. Most ANNs
utilizing back propagation (BP) employ a sigmoid function, which possesses the
distinctive properties of continuity and differentiability, essential requirements in BP
learning [2].
The number of neurons in the hidden layers is very important, since it affects the
training time and generalization property neural networks. On the one hand, too many
neurons may cause the network to memorize (over-fitting) as opposed to generalize; on
the other hand too few neurons would require more training time in finding the optimal
representation or generally result in under-fitting. We adjusted the number of neurons in
the hidden layer experimentally One rule of thumb in the neural network literature
indicates that the number of neurons in a hidden layer should be 2/3 of the number of
input neurons plus the number of output neurons [6].
B. Training the network
The most critical aspect of a successful ANN design is the selection of an input vector
that is general enough for the network to train on efficiently. A poorly chosen input
vector will yield an ill-formed transfer function or weight matrix that will not converge
during the training or will yield inaccurate results during the prediction phase. The
selection of the input vector becomes more complicated as the nature of the process
under consideration becomes more complex. If the window of the input vector is sub-
optimal the network will fail to generalize, which will result in poor learning and
inaccurate prediction.
90
The following figures (Figures 4.6 to 4.8) show how different selection of data for
training will affect the network performance (these figures are related to the producer 1
and injectors 1 and 3). This is a very important step in designing an ANN. Improper data
allocation will result in an accurate or unreliable network. The training and testing data
set should be selected in a way that respects all different ranges of samples in the inputs
and outputs. In all these selections, 80% of the data were used for training and 20
percent are used for testing.
10800
10700 L1
10600 L1 Output
Liquid rate (bbl/day)
10500
10400
10300
10200
10100
10000
9900
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Sample number
Figure 4. 6: Testing results for the trained network for producer 1 (L1 is the actual liquid rate for producer 1
from reservoir simulation and L1 output is the estimated values from the network) when the first 80% portion
of the data is used for training.
91
9750
L1
9700
L1 Output
Liquid rate (bbl/day) 9650
9600
9550
9500
9450
9400
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Sample number
Figure 4. 7: Testing results for the trained network for producer 1 (L1 is the actual liquid rate for producer 1
from reservoir simulation and L1 output is the estimated values from the network) when the last 80% portion
of data is used for training.
10050
10000 L1
L1 Output
9950
9900
9850
9800
9750
9700
9650
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Sample number
Figure 4.8: Testing results for the trained network for producer 1 (L1 is the actual liquid rate for producer 1
from reservoir simulation and L1 output is the estimated values from the network) when the first 40% portion
and the last 40% portion of the data are used for training.
92
This analysis shows that the data selection for training and testing the network has a
significant effect on the performance of the network. In order to make the network
representative of all the data in this research, the histogram of the data is plotted (Figure
4.9). Then 80% percent of data in each interval are assigned to training and the rest are
used for testing. This helped us to respect the whole range of the available data in the
training of the network. Figure 4.10 shows the results of the actual and estimated data
from the new trained network. As can be seen, there is a good agreement between them.
Figure 4. 9: Histogram of input data (liquid production rate) of producer 1.
93
10800
10600
L1
10400
Liquid rate (bbl/day) L1 Output
10200
10000
9800
9600
9400
9200
9000
8800
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Sample number
Figure 4. 10: Test results of the optimum trained network, where L1 is the actual liquid rate from reservoir
simulation and L1 output is the estimated results from the network.
C. Superposition analysis
For each producer, we assume that it is connected to the two nearest injectors; this is
because we do not have any information about the geology and formation structure of
the reservoir, so it is more likely that the nearest wells are connected to each other. We
will thus define a base case network that determines the production rate of the producer
by using the injection rate of the two injectors. After that, we will try to find the
optimum parameters of the network. This optimum designed ANN will be used to
estimate the rate of liquid production. Then we add the nearest injector as a new input
to the network. We will compare the error of the estimation for both cases. The third
injector will be added to the network if there is decrease in the error of estimation. This
process will be continued, to see which injector is connected to the selected producer.
Again, since the number of inputs has changed, the number of parameters in the
network should be optimized.
94
6
Mean absolute error of estimation (%)

5
0
1&2 1,2&3 1,2&4 1,2,3&4
Different combinations of injectors
Figure 4. 11: Mean absolute error of the estimated production rate obtained from designed networks based on
different combinations of injectors with producer 1.
Table 4. 2: Results of connectivity analysis for producer 1.

Producer 2 × × √ √
Figure 4.11 represents the superposition analysis for producer number one: as can be
seen, the best liquid production estimation came from a combination of this producer
and injectors 1, 2 and 3. Table 4.2 shows the results of this analysis for all producers.
Now the number of inputs for each producer’s network is determined, the next step will
be optimizing the network parameters.
For each parameter, we changed the number of elements then we monitored the effects
of this change on the absolute error of the estimation. The optimum point is the point
with the lowest error. The results of this sensitivity study for producer 1 can be seen in
the Figures 4.12 to 4.14.
95
25
20
Mean Absolute Error (%)

15
10
0
0 0.5 1 1.5 2 2.5 3 3.5
Number of hidden layers
Figure 4. 12: Sensitivity analysis on the number of hidden layers shows that by increasing number of hidden
layers, error of estimation will increase too.
5.9
5.8
Mean Absolute Error (%)
5.7
5.6
5.5
5.4
5.3
5.2
0 1 2 3 4 5 6 7
Number of Elements
Figure 4. 13: Sensitivity analysis on the number of processing elements.
96
6.4
6.2
Mean Absolute Error
6
5.8
5.6
5.4
5.2
0 1000 2000 3000 4000 5000 6000 7000
Number of epochs
Figure 4. 14: Sensitivity analysis on the number of iterations for training the network.
Table 4.3 shows the optimum properties for the network designed for estimating the
liquid production rate of the producer 1 from the injection rate of the injectors.
Table 4. 3: Optimum properties of the desired network for estimatiing liquid production rate of producer 1.
Number of hidden layers 1
Number of PE 4
Hidden layer transform function Sigmoid
Output layer transform function Sigmoid
Number of epochs in training 5000
The optimum network is designed to estimate the rate of liquid production based on the
injection rate of the connected injector to producer 1. Figure 4.15 shows the results of
the testing. This network can be used to forecast the future production rate of producer 1
by inputting the injection rate of the injectors. Also, it can be used to carry out
connectivity analysis.
4.3.1.2 Inter-well connectivity analysis
To determine how an injector is supporting the producer, the injection rate of that
injector is changed while the rate of other connected injectors kept constant and the rate
of liquid production is calculated. Figure 4.16 shows the results of the connectivity
97
study for producer 1. According to this figure, it can be concluded that injector 1 has the
highest impact on producer 1 while injector 3 provides the lowest support.
10800
10600 L1
10400 L1 Output
10200
10000
9800
9600
9400
9200
9000
8800
1 2 3 4 5 6 7 8 9 10 11 12 13 14
Sample number
Figure 4. 15: Optimum network test results , where L1 is the actual liquid rate from reservoir simulation and
L1 output is the estimated result from the network.
10500
10400
10300
Production Rate (bbl/day)
10200
1
10100
2
3
10000
9900
9800
8000 8100 8200 8300 8400 8500 8600 8700 8800 8900 9000 9100
Injection Rate (bbl/Day)
Figure 4. 16: Inter-well connectivity analysis for producer 1 (colours represent the change in liquid production
of the producer by change in the rate of associated injector and numbers are the injection well numbers).
98
The same procedure was applied for the rest of the producers; Figures 4.17 and 4.18
represent the connectivity analysis for producers 2 and 3, and the calculated inter-well
connectivity for all producers can be seen in Figure 4.19.
10600
10400
10200
10000 3
4
9800
9600
9400
8000 8200 8400 8600 8800 9000 9200
Figure 4. 17: Inter-well connectivity analysis for producer 2 (colours represent the change in liquid production
of the producer by change in the rate of the associated injector).
14300
14200
14100
14000 1
2
13900
3
13800
13700
13600
8000 8200 8400 8600 8800 9000 9200
Figure 4. 18: Inter-well connectivity analysis for producer 3 (colors represent the change in liquid production
of the producer by change in the rate of associated injector).
99
0.9
0.8
Inter-well connectivity coefficient

0.7
0.6
I1
0.5
I2
0.4
I3
0.3 I4
0.2
0.1
0
P1 P2 P3
Figure 4. 19: Results of inter-well connectivity measurements for all producers by FFBP.
4.3.2 Co-Active Neuro-Fuzzy Inference System (CANFIS)
The second network that has been employed is CANFIS. The same procedures as have
been applied for FFBP are employed to design the optimum network for each producer,
to run the superposition analysis and to determine the inter-well connectivity. Table 4.5
will show the properties of the optimized network. Results of calculated inter-well
connectivity are given in figure 4.21.
Table 4. 4: Properties of the CANFIS Network.
Transform Function Bias Axon
Number of PE 5
Number of epochs in training 1000
100
1.2
Inter-well connectivity coefficient

1
0.8
I1
0.6 I2
I3
0.4 I4
0.2
0
P1 P2 P3
Figure 4. 20: Results of connectivity measurement for all producers from CANFIS.
4.3.3 Determination of water allocation factor
The procedure described in Chapter 3, Section 3.4, for determining the injection
allocation factor was applied to the connectivity results of both networks to determine
the water allocation factor .
Water allocation factor (fraction)
0.4 I1
I2
0.35
I3
0.3 I4
0.25
0.2
0.15
0.1
FFBP CANFIS
Figure 4. 21: Calculated allocation factors for both networks.
101
In the final stage of this chapter the calculated AFs are used to manage the allocation of
water between injectors for the next 20 years of production. The following Figures
show the results of each technique.
Figure 4. 22: Cumulative oil production obtained from WAM, based on FFBP, CANFIS and CRM.
Figure 4. 23: Comparison of oil production rate from WAM, based on FFBP, CANFIS and CRM.
102
Figure 4. 24: Cumulative water production from WAM, based on FFBP, CANFIS and CRM.
As can be seen from the results of the simulation, including daily oil production rate,
cumulative oil production and cumulative water production, the best results were
obtained from Feed-forward back propagation. CANFIS-based WAM also improved the
flooding efficiency. It seems that BP can work better than fuzzy logic for determining
the inter-well connectivity. Comparing them with CRM shows that CRM is the winner
in terms of more improvement in waterflood performance. And it indicates that CRM
works better in determining the inter-well connectivity, as it contains more parameters
related to the production and injection well connection. And also it is much easier to use
CRM to measure connectivity. Firstly, there is no need to run a sensitivity analysis and
second, it is simpler to apply. Therefore for the rest of this study we decided to work
based on the connectivity results obtained from CRM.
103
4.5 References
1. Panda, M.N. and A.K. Chopra, An Integrated Approach to Estimate Well

Interactions, in SPE India Oil and Gas Conference and Exhibition. 1998,
Society of Petroleum Engineers: New Delhi, India.
2. Basheer, I.A. and M. Hajmeer, Artificial neural networks: fundamentals,

computing, design, and application. Journal of Microbiological Methods, 2000.
43(1): p. 3-31.
3. Mohaghegh, S., Virtual-Intelligence Applications in Petroleum Engineering:

Part 1—Artificial Neural Networks. Journal of Petroleum Technology, 2000.
52(9): p. 64-73.
2010. 70(1–2): p. 35-43.
5. Mohaghegh, S., Neural Network: What It Can Do for Petroleum Engineers.

1995.
6. Demiryurek, U., et al., Neural-Network Based Sensitivity Analysis for Injector-

Producer Relationship Identification, in Intelligent Energy Conference and
Exhibition. 2008, Society of Petroleum Engineers: Amsterdam, The
Netherlands.
7. Agatonovic-Kustrin, S. and R. Beresford, Basic concepts of artificial neural

network (ANN) modeling and its application in pharmaceutical research.
Journal of Pharmaceutical and Biomedical Analysis, 2000. 22(5): p. 717-727.
8. Dell, J., S. Meakin, and J. Cramwinckel, Sustainable Water Management in the

Oil and Gas Industry: Use of the WBCSD Global Water Tool to Map Risks, in
SPE International Conference on Health, Safety, and Environment in Oil and
Gas Exploration and Production. 2008, Society of Petroleum Engineers: Nice,
France.

Part 3—Fuzzy Logic. Journal of Petroleum Technology, 2000. 52(11): p. 82-87.
10. Sayarpour, M., et al., Probabilistic history matching with the capacitance–
resistance model in waterfloods: A precursor to numerical modeling. Journal of
Petroleum Science and Engineering, 2011. 78(1): p. 96-108.
104
Chapter 5– New Allocation Management Methodologies
Although several statistical [1-7] and artificial intelligent techniques [8, 9]have been
introduced to measure the connection between wells in previous studies, there has been
less effort directed towards using the inter-well connectivity information for managing
and improving the allocation of the water between injectors. In section 3.4 of Chapter 3
a simple new approach is introduced, in which the calculated inter-well connectivity
results combined with water cut to determine the allocation factor for each injector.
Water allocation management aims to inject the water in a manner that increases the
total oil recovery for a given volume of water. The “good” injectors are thus those
which support the “good” producers. The previous two chapters used the water cut as a
parameter for describing the performance of the producers. In this chapter I will try to
find a better way to quantify a production well’s performance. Therefore further new
techniques are developed: first WC methodology is extended to cumulative water cut
(CWC) and then new parameters are defined for better description of production well
performance. New procedures are then defined for water allocation management, based
on these new parameters and inter-well connectivity measurements.
The reservoir model used in Chapters 3 and 4 will be employed again, with the inter-
well connectivity measurements obtained from the CRM.
5.1 Water allocation management (WAM)

5.1.1 Water Cut
In the previous work in this research inter-well connectivity was measured based on
twenty years of injection and production history. Section 3.4 explained how the
producer’s water cut could be combined with the results of the inter-well connectivity
measurements derived from the CRM technique to determine the water allocation factor
for the next 20 years. The water cut is the parameter which changes during the well’s
105
production life, after water breakthrough at the production wells. An alternation in the
injection scenario by applying a new allocation factor to the injectors will change the
water cut in the producers; hence the analysis in Section 3.4 has been extended by
updating the injection allocation factor every 5 years, taking into account the history of
change in the water cut after each 5-year interval.
Figure 3.1 shows the results of the new analysis, representing the water cut for each
producer at the end of each 5 year interval.
100
90
80
70
60
50 P1
40 P2
30 P3
20
10
0
20 25 30 35 40
Production time (years)
Figure 5. 1: Water cut of each producer at the end of each 5-year interval after beginning of the water
allocation management.
Water cut is increasing in producers 1 and 3 while the water cut of producer 2 shows a
decrease in the first 5 year period, then starts to increase with a similar trend as that of
the other producers in the second period but again, in the third period, the water cut
showed a smaller rate of increase compared with other producers. The water allocation
factor for each of the four injectors, calculated from the results of CRM inter-well
connectivity measurement and water cut of each producer (according to the procedure
defined in section 3.4) is shown in Figure 3.2.
Figure 3.2 shows that injector 1 is consistently allocated the greatest volume of water
while injector 3 receives the minimum fraction of available water. This is because
producer 3, the best producer, receives most support from injector 1 (Figure 5.3).
106
Producer number 2 had the highest water cut at the beginning of the 20 year period of
water allocation management. After 5 years its performance improved due to the
efficiency of the water allocation management. This resulted in an increase in the
volume of water allocated to the injectors connected to this producer during the second
period. This can be observed in Figure 5.4, where more water allocated to the injector 4.
0.45
0.4
0.35
0.3
0.25 I1
0.2 I2
0.15 I3
0.1 I4
0.05
0
20 25 30 35
Time (years)
Figure 5. 2: Calculated water allocation factor for each injector at the end of each 5 year period.
0.9
0.8
Inter-well connectivity (fraction)
0.7
0.6
0.5 I1
0.4 I2
0.3 I3
0.2 I4
0.1
0
P1 P2 P3
Producers
Figure 5. 3: Inter-well connectivity measurements determined from CRM for each producer.
107
Figure 5. 4: Reservoir oil production rate for base case and the case with water allocation management.
Figure 5. 5: Cumulative oil production for base case and the case with water allocation management.
Figure 5.4 compares the daily oil production profile between the base case, without
allocation management, and the one with allocation management. It shows that the
allocation increases oil production by 5%, resulting in an increase in the field
cumulative oil production of 1.5% (Figure 5.5). Figures 5.6 and 5.7 illustrate how water
production also decreased after water injection management was implemented.
108
Figure 5. 6: Field water cut for base case and the case with water allocation management.
Figure 5. 7: Reservoir cumulative water production for base case and the case with water allocation
management.
109
5.1.2 Cumulative Water Cut (CWC)
The previous analysis showed that the water cut can change significantly in a
production well which is well-connected to the injectors, if the water allocation is
changed without respecting the previous history of the production. This was illustrated
by the experiment on producer number 2, in the previous section, when the water
allocation factor based on the water cut was changed. It was concluded that this was not
a suitable parameter for describing well performance when deciding long-term water
allocation management. We therefore tested whether the cumulative water cut fraction
(CWC) is a better parameter for water allocation management. CWC is defined as:
𝑄𝑤
𝐶𝑊𝐶 = (Equation 5.1)
𝑄𝑙
in which Qw and Ql are the cumulative water production and the cumulative liquid
production from a production well. Therefore Equation 3.21 in the water allocation
management procedure described in Section 3.4 of Chapter 3will change to:
OI =
(1 − CWC ) × LR (Equation 5.2)
The remainder of the procedure described in Chapter 3 remains unchanged. The

following figure shows the change in CWC for each producer after carrying out water
allocation management at 5-year intervals. Unlike the previous analysis, in this case,
producer 2 is still the worst producer and therefore the calculated water allocation factor
has not changed considerably (Figure 5.9).
Figures 5.10 to 5.13 report the improved results when applying CWC for allocation
management rather than WC: a 10% increase in daily oil production, 2% increase in
cumulative oil production and a 1.6% decrease in cumulative water production resulted.
More oil was initially produced when water allocation was based on WC but CWC gave
better results during the subsequent fifteen-year period. (Figures 5.10 to 5.13). It was
concluded that WC is more suitable for short-term optimization of the oil production,
while CWC is to be preferred for long term optimization.
110
70
Cumulative water cut (fraction)

60
50
40
P1
30
P2
20 P3
10
0
20 25 30 35
Time (years)
Figure 5. 8: Cumulative water cut of each producer.
0.45
0.4
0.35
0.3
0.25 I1
0.2 I2
0.15 I3
0.1 I4
0.05
0
20 25 30 35
Time (year)
Figure 5. 9: Calculated water allocation factor for different production periods.
111
Figure 5.10: Reservoir oil production rate for all three injection scenarios.
Figure 5.11: Reservoir cumulative oil production for all three injection scenarios.
112
Figure 5. 12: Reservoir water cut for all three injection scenarios.
Figure 5.13: Field cumulative water production for all three injection scenarios.
113
5.1.3 Relative Oil Production Ratio (OPR)
The previous section showed that the CWC-based water allocation management gave
better performance than a WC-based procedure, for long term optimisation. The CWC
parameter has a value between 0 and 1, i.e., the worst CWC value, one that corresponds
to a well with 100% water cut, is 0. Ideally, we would like to assign a negative value for
a production well that produces more water than oil. A new parameter, the Cumulative
Oil Production Ratio (OPR), to describe the production well’s performance, has been
defined as:
𝑂𝑃𝑅 = 𝑂𝑅 − 𝑊𝑅 (Equation 5.3)
In which OR and WR, the production well’s Oil Ratio and Water Ratio respectively, are
defined by:
𝑄𝑜
𝑂𝑅 = (Equation 5.4)
𝑄𝑜𝑟
𝑄𝑤
𝑊𝑅 = (Equation 5.5)
𝑄𝑤𝑟
where Qo and Qw are previously defined as the well’s cumulative oil and cumulative
water production and the new parameters, Qor and Qwr are the reservoir’s cumulative oil
production and cumulative water production, respectively. Equation 5.3 shows that the
OPR can have a negative value when a production well is producing more water than
oil.
The oil index (OI) parameter used for water allocation management (Equation 3.21)
now becomes:
OI = OPR (Equation 5.6)
Note that the LR parameter used in equations 3.21 and 5.2 may now be omitted, since
the OI clearly includes a comparison of the individual well’s production with that of the
reservoir as whole. A consequence of the OI having a negative value is that the total
water allocation index (AFt), defined by equation 3.22, also has a negative value. The
final allocation factor (AF) can be calculated from the following three combinations of
AFt values:
114
5.1.3.1 All AFt have positive values
The final allocation factor, as defined by equation 3.24, may be used.
5.1.3.2 All AFt have negative values
The final allocation factor is now:
1−𝐴𝐹𝑛𝑖
𝐴𝐹𝑖 = ∑𝐼 (Equation 5.7)
𝑖=1(1−𝐴𝐹𝑛𝑖 )
In which AFni is the normalized allocation factor for the injector i:
AFti
AFi = I
∑ AF
i =i
ti
(Equation 5.8)
5.1.3.3 Both positive (𝑨𝑭+ −

𝒕 ) and negative (𝑨𝑭𝒕 ) values of AFt are present
AFn is now calculated separately for wells with positive and with negative AFt values,
and the final allocation factor will be:
1. For wells with positive AFt:

+1
𝐴𝐹 = (1 + 𝐴𝐹𝑛𝑖 ) (Equation 5.9)
𝐼
2. For the wells with negative AFt:

1
−)
𝐴𝐹 = (1 − 𝐴𝐹𝑛𝑖 (Equation 5.10)
𝐼
in which I is the number of injectors. The calculated allocation factor using the OPR
approach for the four injectors at each of the four different periods is given in Figure
5.14. Figure 5.14 shows a significantly different water allocation factor to that
calculated by the earlier approaches. Further, it does respond to the change in the
production performance, especially towards the later time periods, when injector 2
shows a consistent increase in its water allocation factor.
Figure 5.15 is the OPR value for each producer. Two wells have a negative OPR value,
resulting in a reduced volume of water being assigned to the injectors connected to
those wells (Figure 5.14).
115
Applying this technique for water allocation management significantly improves the
performance of water flooding compared with WC and CWC (see Figures 5.16 to 5.20).
The main advantage of the OPR method is that it differentiates more clearly between
good and bad producers|, being better able to “punish” injectors supporting producers
with a high water cut.
0.4
0.35
0.3
0.25
I1
0.2
I2
0.15 I3
I4
0.1
0.05
0
20 25 30 35
Time (years)
Figure 5. 14: Calculated water allocation factor based on OPR at the end of each 5-year period.
116
0.25
0.2
0.15
OPR (fraction) 0.1
0.05 P1
0 P2
20 25 30 35 P3
-0.05
-0.1
-0.15
-0.2
Time (years)
Figure 5. 15: Measured OPR for each producer at the end of each 5-year period.
Figure 5. 16: Comparison of daily oil production rate obtained from different allocation management
scenarios.
117
Figure 5. 17: Comparison of cumulative oil production obtained from different allocation management
scenarios.
Figure 5. 18: Comparison of field water cut obtained from different allocation management scenarios.
118
Figure 5. 19: Comparison of cumulative water production obtained from different allocation management
scenarios.
5.1.4 Oil Production Index (OPI)
The range of OPR is between -1 and 1 but it cannot be exactly -1 or +1. For example,
look at the example illustrated in Figure 5.18. This figure represents a reservoir
consisting of 3 producers. For a certain period of time, one producer (P1) produced only
oil, the second producer produced just water and the last producer produced with 75%
CWC. Therefore, calculated OPR for these production wells will be -0.571, 0.8 and -
0.228 respectively.
In this study the aim is to assign -1 to a producer that is producing only water and 1 to
the producer that producing only oil. This will help to improve the water allocation
management. Therefore the new parameter is defined as the cumulative oil production
index (OPI),
𝑄𝑜 −𝑄𝑤
𝑂𝑃𝐼 = (Equation 5.11)
𝑄𝑙
119
Figure 5. 20: Reservoir containing three production wells.
where Qo and Qw are previously defined as the well’s cumulative oil and cumulative
water production and the new parameter Ql is cumulative liquid production. The OPI
values for production wells in this example (Figure 20) will be +1 for P1, -1 for P2 and
-0.5 for P3.
Like OPR, OPI could be a negative value. Therefore, the allocation management
procedure will be the same as for OPR. The only difference is the OI values for each
producer. In this case. the Equation 3.21 will be like this:
=
OI OPI r − OPI (Equation 5.12)
where OPIr and OPI are determined OPI values for reservoir and production well
respectively. The calculated OPI for wells and reservoir in our case study is shown in
Figure 5.21.
120
1.2
PI Field
1
PI 1
Oil production index
0.8 PI 2
PI 3
0.6
0.4
0.2
0
0 1000 2000 3000 4000 5000 6000 7000 8000
-0.2
Time (Day)
Figure 5. 21: Measured OPI for reservoir and each production well.
Figure 5.22 shows the determined water allocation factor for each period.
0.45
0.4
0.35
0.3
0.25 I1
0.2 I2
0.15 I3
0.1 I4
0.05
0
20 25 30 35
Time (years)
Figure 5. 22: Calculated water allocation factor, based on OPI at the end of each 5-year period.
Applying this technique for water allocation management slightly improves the
performance of water flooding, when compared with OPR (see Figures 5.23 to 5.25).
121
Figure 5. 23: Comparison of daily oil production rate obtained from different allocation management
scenarios.
Figure 5. 24: Comparison of cumulative oil production obtained from different allocation management
scenarios.
122
Figure 5. 25: Comparison of field water cut obtained from different allocation management scenarios.
Results of the allocation management with the water cut shows that water allocation
factors may change considerably by changing the water allocation in the producers
highly affected by injectors and also it did not respect the previous performance of the
well. Better results were obtained from CWC. Although at the beginning WC worked
better, CWC took over the WC at the final production intervals. Also the allocation
factor obtained from CWC showed less tendency to change during different intervals by
change in the performance of the production wells. Significantly better improvement
was obtained from OPR and OPI compared to CWC (Figure 5.26).
Comparison of WC with CWC indicated that WC is good for short-term improvement

while CWC can be applied for longer-term management. However, since OPR and OPI
had more power to separate good and bad production wells, water allocation based on
them helped to send more water to the injectors connected to the producers with higher
oil production, while injecting less in the injector supporting production wells with
higher water production. Because of that, significant improvement was achieved with
allocation management with OPR and OPI, and OPI was found to work slightly better
than OPR.
123
2.50E+08
Cumulative oil and water production

Cumulative oil production
2.40E+08
Cumulative water production
2.30E+08
2.20E+08
(STB) 2.10E+08
2.00E+08
1.90E+08
1.80E+08
Base case WC CWC OPR OPI
Figure 5. 26: Waterflooding performance in terms of cumulative oil and water production for different water
allocation procedures.
5.2 Production Allocation Management (PAM)
In the previous section, we explained the different parameters we employed to describe

the performance of the production wells for water allocation management. In this
section we examine how the various parameters can be used to manage the allocation of
production rate between the production wells, followed by evaluation of the
improvement in the waterflood project performance in terms of more oil recovery and
less water production. At first, we studied only production management with equal
distribution of water between the injectors (no WAM). In the following sections, the
parameters that have been used in this study will be reviewed.
5.2.1 Oil ratio (OR)
The OR is the first parameter to be employed for production allocation management.

OR, defined by Equation 5.4, is the ratio of the cumulative oil production of a producer
to the total cumulative oil production of the reservoir. Therefore, wells with higher OR
have a larger share of the cumulative oil production from the reservoir. As a result,
124
higher production rate should be allocated to the wells with higher OR. Equation 5.13
shows the calculated production allocation factor from OR.
OR j
PAFj = P
(Equation 5.13)
∑ (OR )
j =1
j
In which PAFj and ORj are the production allocation factor and oil ratio for the
producer, j, and P is the number of production wells.
5.2.2 Relative oil production ratio (OPR):
Normally a well will be producing oil but will also produce significant volumes of
water as well. The relative oil production ratio (OPR) can be employed to define a
production rate allocation factor that also takes into account the water production from
the well. This technique will favour production from a well that produces more oil and
less water.
Calculation of the OPR for each production well, using Equation 5.3, results in one of
these three possibilities:
5.2.2.1 All OPR values are positive
In this case, the following equation can be used to determine the PAF for each producer:
OPR j
PAFj = P
∑ (OPR )
j =1
j
(Equation 5.14)
in which OPRj is the relative oil production ratio of the producer, j.
5.2.2.2 All OPR values are negatives
Production wells with negative values are producing more water than oil. In such a
condition, Equation 5.15 shows how PAF can be calculated.
125
−
(1 − OPR −j )
PAF = j P−
∑ (OPR
j =1
−
j )
(Equation 5.15)
where PAFj− and OPR −j are the production allocation factor and relative oil production
ratio of the producer, j (the minus sign refers to the well having negative OPR value). P-
is the number of production wells with OPR less than 0.
5.2.2.3 Both positive (𝑶𝑷𝑹𝒋+ ) and negative (𝑶𝑷𝑹𝒋𝒕

−
) values of OPR are present
PAF will be calculated separately for wells with OPR −j 𝑎𝑛𝑑 𝑤𝑒𝑙𝑙𝑠 𝑤𝑖𝑡ℎ OPR +j .
−
OPR −j
PAF = j P−
∑ (OPR
j =1
−
j )
(Equation 5.16)
+
OPR +j
PAF = j P+
∑ (OPR
j =1
−
j )
(Equation 5.17)
where PAFj+ and OPR +j are the production allocation factor and relative oil production
ratio of the producer, j. P+ is the number of production wells with OPR more than 0. In
the next step the production rates of the negative wells will be decreased and at the same
time the production rate of producers with positive OPR will be increased. The
following equations introduce the new production rates of each type of well.
qnj− =−
qoj− ( PAFj− × qoj− )
(Equation 5.18)
qnj+ =+
qoj+ ( PAFj+ × qoj+ )
(Equation 5.19)
qoj− qnj−
and are the old and new production rate of the production well, j, that has a
qoj+ qnj+
negative value for OPR . and correspond to the old and new production rates of
producers having positive OPR.
126
5.2.3 Oil production index (OPI)
OPI, which has been previously defined by Equation 5.11, is final parameter tested for
production allocation management. In water allocation management we found that OPI
is the best parameter for describing production well performance, as it differentiates
clearly between a good producer and a bad producer. For determining the production
allocation for each production well, the OPI of each individual is compared with the
OPI of the reservoir.
=
OPI OPI j − OPI r
rj
(Equation 5.20)
where OPIrj is the difference between producer OPIj and reservoir OPIr . Therefore,
there will be two types of the wells: wells with OPIrj >0 and producers with OPIrj <0.
The PAF for well wells with OPIrj less than 0 will be calculated as:
−
OPI rj−
PAF = j P−
(Equation 5.21)
∑ (OPI
j =1
−
rj )
The production rate of the producers having a negative effect on the reservoir OPI will
then be reduced, based on Equation 5.21.
qnj− =−
qoj− ( PAFj− × qoj− )
(Equation 5.21)
The production allocation factor ( PAFj+ ) of positive OPIrj ( OPI rj+ ) is calculated using
equation 5.22.
OPI rj+
PAFj+ = P+
(Equation 5.22)
∑ (OPI
j =1
+
rj )
The amount of production rate subtracted ( qs− ) from producers with a negative effect on
the reservoir OPI ( OPI rj− ) will be calculated by Equation 5.23.
127
P− P−
−
=
q
s ∑q −∑q
−
oj
=j 1 =j 1
−
nj
(Equation 5.23)
+
qs− will then be added to the production rate of the producers with OPI rj , according to
the following equation.
qnj+ =+
qoj+ ( PAFj+ × qs_ )
(Equation 5.24)
Figures 15.27 to 15.29 represent the results of the production allocation management
from all three methodologies. The studied model is the same as the one used for water
allocation management. The first 20 years of production history are used to define new
a production allocation for each producer for the next 20 years. The process of
production allocation management is updated every 5 years. Minimum and maximum
liquid production is introduced for each production well, based on the sensitivity
analysis on the well models.
Figure 5. 27: Oil production rate obtained from different techniques of production allocation management.
128
Figure 5. 28: Cumulative oil production obtained from different techniques of production allocation
management.
Figure 5. 29: Produced water rate obtained from different techniques of production allocation management.
129
Figure 5. 30: Cumulative water produced from different techniques of production allocation management.
As expected, PAM caused more oil and less water to be produced. Significant
improvement was obtained from OPR and OPI. As discussed in WAM, these two
methods, although they not only tend to increase oil production from producers that are
producing oil but they also apply a penalty if the well also producing water. Since OPI
introduces a greater difference between a good producer and bad producer, more
improvement is achieved from it.
As both WAM and PAM increased the efficiency of the studied waterflooding project, it
was therefore decided to apply both types of allocation management simultaneously.
5.3 Production and injection allocation management (W&P AM)
CRM results of inter-well connectivity measurement and OPI are employed, based on
the procedures explained in sections 5.1.4 and 5.2.3 to define the water and production
allocation factor for the injectors and producers at the same time. The results of this
analysis are shown in the following figures.
130
Figure 5.31: Comparison of oil production rate obtained from WAM, PAM and W&P AM.
Figure 5. 32: Cumulative oil production from WAM, PAM and W&P AM.
131
Figure 5. 33: Water production rate from WAM, PAM and W&P AM.
Figure 5. 34: Cumulative water production obtained from WAM, PAM and W&P AM.
132
2.50E+08
Cumulative oil & water production

2.40E+08
2.30E+08
Base Case
(STB)
2.20E+08 WAM
PAM
2.10E+08
W&P AM
2.00E+08
1.90E+08
Cum waterProduced CumOilProduced
Figure 5. 35: Bar chart of cumulative oil and water production from WAM, PAM and W&P AM.
The comparison of PAM and WAM shows that WAM is more effective in improving
waterflood efficiency than PAM. Better oil production was obtained at the beginning of
PAM, while the improvement built up slowly in the case of WAM. This is because there
will be a lag time in seeing the change in the production rate as a result of change in the
injection scheme. The reduction in water production was bigger than the increase in oil
production when changing from WAM to PAM, showing that a combination of inter-
well connectivity and production performance monitoring were good for determining
the water allocation factor.
W&P AM significantly improved the performance of the project. However, it was much
more effective in increasing cumulative oil production (the amount of improvement was
twice that of WAM and PAM) than in decreasing water production (Figure 5.33). This is
because the both WAM and PAM are good at improving the oil production rate, while
PAM is not very good at decreasing water production. This is because PAM increases
the liquid production rate of the good producers. So due to increase in the liquid
production of these wells, water production also will increase.
Although new techniques have been developed for production allocation management
but in a reservoir, liquid and oil production are often constrained by the reservoir
conditions, outflow performance of the wells, flow characteristics of the pipeline
network, fluid-handling capacity of surface facilities, safety and economic
considerations, or a combination of these factors. While adjusting well production rates
133
and allocating water-injected rates can control production, implementation of these
controls in an optimal manner is also very important and is not easy. Determination of
the optimal operational settings at a given time, subject to all constraints, to achieve
certain operational goals is the objective of dynamic production optimization, which
requires simultaneous consideration of the interactions between the reservoir, the wells,
and the surface facilities [10].
134
5.4 References
1. Yousef, A.A., L.W. Lake, and J.L. Jensen, Analysis and Interpretation of
Interwell Connectivity From Production and Injection Rate Fluctuations Using
a Capacitance Model, in SPE/DOE Symposium on Improved Oil Recovery.
2006, Not subject to copyright. This document was prepared by government
employees or with government funding that places it in the public domain.:
Tulsa, Oklahoma, USA.

Production- and Injection-Rate Fluctuations. SPE Reservoir Evaluation &
Engineering, 2006. 9(6): p. pp. 630-646.

Texas.
2010. 70(1–2): p. 35-43.
5. Izgec, O. and C.S. Kabir, Understanding reservoir connectivity in waterfloods

before breakthrough. Journal of Petroleum Science and Engineering, 2010.
75(1–2): p. 1-12.
2003. 6(1): p. 6-16.

8. Panda, M.N. and A.K. Chopra, An Integrated Approach to Estimate Well

Interactions, in SPE India Oil and Gas Conference and Exhibition. 1998,
Society of Petroleum Engineers: New Delhi, India.
9. Demiryurek, U., et al., Neural-Network Based Sensitivity Analysis for Injector-

Producer Relationship Identification, in Intelligent Energy Conference and
Exhibition. 2008, Society of Petroleum Engineers: Amsterdam, The
Netherlands.

U.S.A.
135
Chapter 6– Water Allocation Optimisation
6.1 Introduction
In the previous chapters it was shown that water allocation was managed by combining
the results of connectivity measurements with performance evaluation of production
wells. In that methodology, less water was allocated to the injectors that were well-
connected to the high water cut producers.
This chapter will propose another approach for water allocation management. This
involves monitoring the effect of a change in the injection rate on the produced oil in the
producers. We will then employ an optimisation engine to find the best injection rates
for maximising the oil production rates.
We can forecast production rates by determining the injection rates of injection wells,
for example by multi-linear regression (MLR), capacitive resistive model (CRM) and the
reservoir simulator. In the next section MLR and CRM will be reviewed again in order
to find a suitable optimiser tool that can be connected to them for defining best
allocation factors for injection wells.
6.2 MLR review
Let the liquid production rate 𝑞𝑗 from the well 𝑗 be described by the following linear
model in terms of the injection rates:
𝑞𝑗 = ∑𝐼𝑗=1 𝛽𝑖𝑗 𝑖𝑖 , (Equation 6.1)
where the coefficients 𝛽𝑖𝑗 are determined from the observed data at the modelling stage.
By construction, the coefficients have a property that
∑𝐼𝑖=1 𝛽𝑖𝑗 = 1, 𝑗 = 1, 𝑁𝑖𝑛𝑗 . (Equation 6.2)
136
This ensures that the rate of the total injected water is equal to the rate of the total
produced liquid [1, 2]:
∑𝐼𝑗=1 𝑖𝑖 = ∑𝑃𝑖=1 𝑞𝑗 . (Equation 6.3)
The purpose of water allocation management is to find such injection rates (𝑖𝑖 ), 𝑗 = 1, 𝐼,
that maximise the total oil production rate from the reservoir.
If the water cut of the well 𝑗 is constant, the oil production rate from this well is
expressed in terms of the injection rates as follows:
𝑞𝑜,𝑖 = (1 − 𝑊𝐶𝑖 )𝑞𝑖 = (1 − 𝑊𝐶𝑖 ) ∑𝐼𝑗=1 𝛽𝑖𝑗 𝑖𝑖 . (Equation 6.4)
Then the objective function can be rewritten in terms of the injection rates as
P  P I  I P 
∑ qo ,i =∑ (1 − WCi ) ∑βij ii  =
=i 1 
∑  ∑ (1 − WCi ) βij  i j → max.
i 1 
=i 1 =j 1  =j 1 =
(Equation 6.5)
There are also constraints. One constraint states that the total injection rate is equal to a
certain value i t :
i
∑i
i =1
i = iit (Equation 6.6)
There are also two sets of constrains involving the maximum daily amounts of injected
water and produced liquid:
¯
ii ≤ iimax , j =
1, I (Equation 6.7)
And
I ¯
∑β i ≤ q
i =1
ij i
max
j , i=
1, P (Equation 6.8)
137
We need to ensure that ii is equal or greater than zero i.e. negative values are not
allowed. Thus:
𝑖𝑖 ≥ 0, 𝑖 = 1, 𝐼, (Equation 6.9)
We can thus formulate the following linear programming (LP) problem for the balanced
MLR in terms of the injection rates:
∑𝐼𝑖=1�∑𝑃𝑗=1(1 − 𝑊𝐶𝑖 )𝛽𝑖𝑗 �𝑖𝑖 → max (Equation 6.10)
Subject to:
∑𝐼𝑗=1 𝑖𝑖 = iit
I ¯
∑βij ii ≤ q maj x , i =
i =1
1, P
0 ≤ 𝑖𝑖 ≤ iimax 𝑗 = 1, 𝐼 (Equation 6.11)
Let us rewrite the LP problem in the vector notation, because this is more compact and
more convenient to code in Matlab or Excel. Firstly, denote the rates and the water cuts
as column vectors
iimax ��
𝑖𝑖,1 𝑞𝑗,1
q max
j
⎛ ,1 ⎞ ⎛ ,1 ⎞
𝐢𝑖 = � ⋮ � , ii = ⎜ ⋮ ⎟ , 𝐪𝑗 = � ⋮ � , q j = ⎜ ⋮ ⎟
max max
𝑖𝑖,𝐼 �� 𝑞𝑙,𝑃

iimax q max
j
⎝ ,𝐼 ⎠ ⎝ ,𝑃 ⎠
𝑞𝑜,1 𝑊𝐶1
𝐪𝑜 = � ⋮ �, 𝐰 = � ⋮ �
𝑞𝑜,𝑃 𝑊𝐶𝑃
and the coefficients of the MLR model as the 𝑃 × 𝐼 matrix:
𝛽11 ⋯ 𝛽1,𝐼
𝐁=� ⋮ ⋱ ⋮ �
𝛽𝑃1 ⋯ 𝛽𝑃,𝐼
138
Secondly, remember some basic linear algebra notation. The vector 𝟏𝒏 is the column 𝑛-
vector whose entries are all ones, the matrix 𝑰 is the identity matrix of the required
dimensionality (which is clear from the context), 𝑑𝑖𝑎𝑔(𝒙) is a matrix with the elements
of the 𝑛-vector 𝐱 on the main diagonal and zeroes elsewhere:
T 1 ⋯ 0 𝑥1 ⋯ 0
𝟏𝑛 = �1,
��… ,1� , 𝐈 = � ⋮ ⋱ ⋮ �, diag(𝐱) = � ⋮ ⋱ ⋮�
𝑛 0 ⋯ 1 0 ⋯ 𝑥𝑛
Then
( I − diag ( w ) ) q j =
qo = ( I − diag ( w ) ) Bii (Equation 6.12)
and the objective function takes the form:
𝟏𝑃 T 𝐪𝑜 = �𝟏𝑃 T �𝐈 − diag(𝐰)�𝐁�𝐢𝐪𝑖 → max (Equation 6.13)
The Equation 6.6 can be rewritten as
𝟏𝐼 T 𝐢𝑖 = iit
and Equation 6.7 as
𝐁𝐢𝑖 ≤ q tj (Equation 6.14)
Then the LP problem in terms of the injection rates is reformulated as follows:
�𝟏𝐼 T �𝐈 − diag(𝐰)�𝐁�𝐢𝑖 → max (Equation 6.15)
Subject to:
𝟏𝐼 T 𝐢𝑖 = iit (Equation 6.16)
𝐁𝐢𝑖 ≤ q max
j (Equation 6.17)
139
0 ≤ 𝐢𝑖 ≤ iimax (Equation 6.18)
This problem can be solved using Matlab or Excel (if the number of the injection wells
is relatively small).
6.3 CRM Review
The CRM at the (discrete) time step 𝑡𝑛 , 𝑛 ≥ 0, takes into account the injection rates and
change in bottom hole pressures (BHP) at the previous time steps as well as the primary
production (Yousef et al., 2005; Sayarpour et al., 2009) [3, 4]. Let the liquid production
rate 𝑞𝑗 from the well 𝑗 be described by the CRM:
− ( t n − t0 )
n I ∆P ( k ) 
−∆tk
 −(tτn −tk ) 
+ ∑  ∑ f ij iik − J jτ j wf , j
τj τj
q j (tn ) = q j (t0 )e  1 − e e j  (Equation 6.19)
=  i 1
k 1= ∆tk   
 
 
Where I ijk and ∆Pwf( k,)j represent the injection rate of injector i and the changes in BHP of
producer j during time interval t k-1 to t k respectively.; ∆𝑡𝑘 is the length of the time step
𝑘; 𝜏𝑖𝑗 , 𝑓𝑖𝑗 and 𝐽𝑗 are the time constant, the fraction of injection rate contributing in
production rate and productivity index associated with the volume between the injector
𝑖 and the producer 𝑗 pair.
Let us change the summation order in the second term of the Equation 6.19:
 t −t   I  
∆t
 − tnτ−tk   
 
I − n 0 
 τj 
n
  ∆Pwf( k,)i − k
q j ( tn )= ∑q ( t ) e  
+ ∑ ∑  fij i j − J jτ j
(k )

 1 − e
τj
 e  j  

 i 1  ∆tk    
j 0
=i 1 = k 1=
  
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(Equation 6.20)
and split it in three parts:
140
 t −t     − tnτ−tk   
 
∆Pwf( k,)i  
∆t
I − n 0 
 τ ij   I  ( k )
n −1 − k
q j ( tn )= ∑q ( t ) e  
+ ∑ ∑  fij ii − J jτ j  1 − e
τj
 e  j  
 ∑ tk   
 i 1  
j 0
=i 1 = k 1=
   
 − n 
∆t
∆Pwf( n,)i I I 
∆t
− n 
∑ 1 − e  J jτ j + ∑ 1 − e τ ij
τj
−  fij ii( n )

∆tn i 1 =  i 1 
=
   
(Equation 6.21)
It is easy to see that:
(𝑛) 𝐼
1. 𝑞𝑗 (𝑡𝑛 ) linearly depends on the �𝑖𝑖 � .
𝑖=1
2. The sum of the first three terms is constant at the time step 𝑡𝑛 ,, and, hence, does
not affect the objective function gradient. This means that the gradient depends
neither on the initial production nor on the prehistory of injection rates and
BHP’s.
(𝑛) 𝐼
3. If the time step length ∆𝑡𝑛 is the same for all 𝑛, then the coefficients at �𝑞𝑖 �
𝑖=1
are the same for all the time steps. This means that the objective function
gradient is the same all the time and only needs to be calculated once, at the
beginning.
4. Any constraints on the production rate depend on the initial production, the
prehistory of injection rates and BHPs, as well as the current value of BHP for
each production well (we presume that the BHP can be measured).
If the production wells’ BHPs are constant, Equation 6.21 can be simplified:
 t −t 
− n 0   I 
n −1
∆t  
 − tnτ−tk   I  ∆t

I
 τj    − k
(k )  
− n
q j ( tn ) ∑q j ( t0 ) e + ∑ ∑  1 − e j fij ii  + ∑ 1 − e j
τ τ
=  
e  ij 
 fij ii( n )
=i 1 =  i 1 
k 1= 

  =i 1 
 



(Equation 6.22)
The structure of the model is identical to that of the unbalanced MLR: a free term plus a
scalar product of a vector of coefficients and the vector of the injection rates to be
found. Let us denote
141
𝑡 −𝑡 (𝑘)
(𝑛) −� 𝑛𝜏 0 � (𝑘) ∆𝑃𝑤𝑓,𝑖
𝛽𝑖0 = ∑𝐼𝑖=1 𝑞𝑗 (𝑡0 )𝑒 𝑗 + ∑𝑛−1 𝐼
𝑘=1 �∑𝑖=1 ��𝑓𝑖𝑗 𝑖𝑖 − 𝐽𝑗 𝜏𝑗 � �1 −
∆𝑡𝑘
𝑒−∆𝑡𝑘𝜏𝑗𝑒−𝑡𝑛−𝑡𝑘𝜏𝑗−∆𝑃𝑤𝑓,𝑖(𝑛)∆𝑡𝑛𝑗=1𝐼𝐼1−𝑒−∆𝑡𝑛𝜏𝑗𝐽𝑗𝜏𝑗
(Equation 6.23)
∆𝑡
(𝑛) − 𝜏𝑛
𝛽𝑖𝑗 = �1 − 𝑒 𝑗 � 𝑓𝑖𝑗 (Equation 6.24)
where 𝑖 = 1, 𝑃 and 𝑗 = 1, 𝐼. Let us introduce the following vectors and matrix:
(𝑛) (𝑛) (𝑛) (𝑛)

𝑖𝑖,1 𝛽10 𝛽11 ⋯ 𝛽1,𝐼
(𝑛) (𝑛)
𝐢𝑖 = � ⋮ �, 𝐛0 = � ⋮ �, 𝐁(𝑛) =� ⋮ ⋱ ⋮ �.
(𝑛) (𝑛) (𝑛) (𝑛)
𝑖𝑖,𝐼 𝛽𝑃,0 𝛽𝑃,1 ⋯ 𝛽𝑃,𝐼
The water cuts can be taken from the previous time step:
(𝑛−1)
𝑊𝐶1
𝐰 (𝑛−1) =� ⋮ �
(𝑛−1)
𝑊𝐶𝑃
If, as above, iimax and q max

j mean the upper limits of the injection and liquid rates for all
time steps, the following LP problem can be formulated at each time step 𝑡𝑛 :
(𝑛)
�𝟏𝑃 T �𝐈 − diag�𝐰 (𝑛−1) 𝐰�� 𝐁 (𝑛) � 𝐢𝑖 → max
(Equation 6.25)
subject to:
(𝑛)
𝟏𝑁𝑖𝑛𝑗 T 𝐢𝑖 ≤ iit (Equation 6.26)
(𝑛) (𝑛)
𝐁 (𝑛) 𝐢𝑖 ≤ q max
j − 𝐛0 (Equation 6.27)
(𝑛)
0 ≤ 𝐢𝑖 ≤ q max
j (Equation 6.28)
142
This LP problem should be solved at each time step. If the upper rate limits and the total
amount of water iit change in time, this can be easily accommodated in the LP problem.
6.4 Linear Programming (LP)
As discussed in the previous sections, MLR and CRM are both linear, so their
constraints and objective functions are linear as well; therefore, Linear Programming
can be used as an optimiser to combine with these two estimation methods in order to
optimise the water allocation.
Linear programming is the name of a branch of applied mathematics that deals with
solving optimization problems of a particular form. Linear programming problems
consist of a linear cost function (consisting of a certain number of variables) which is to
be minimized or maximized, subject to a certain number of constraints [5]. The
constraints are linear inequalities of the variables used in the cost function. The cost
function is also sometimes called the objective function. Linear programming is closely
related to linear algebra; the most noticeable difference is that linear programming often
uses inequalities in the problem statement rather than equalities [6]. LP has been used
by petroleum engineers for range of optimising problems in the petroleum industry [6-
10].
6.5 Case study
The example used in this chapter is the same model studied in previous chapters.
1. MLR and CRM coefficients are calculated based on 20 years production and
injection history. This was already done in Chapter 3.
2. Developed MLR and CRM models are used for forecasting future oil production.
3. LP is applied periodically each 5 years for determining optimum injection rates
for each injector in order to maximise daily oil production rate for the next 20
years of production. LP is solved by a third party product; in this case, Excel
Solver, which was sufficient for solving the LP in this study.
4. Minimum and maximum limits are applied for injection rates. The total injection
rate is defined to be the same as the total rate of production (VR=1).
143
The results of calculated allocation factor for injectors are shown in Figures 6.1 to 6.3,
for both cases (in both cases the allocation results were the same).
As can be seen, the overall waterflooding performance improved: there is an increase in

oil production rate and decrease in water production. However, in both cases (MLR and
CRM), after changing the injection rate for the first 5 years, there was a sudden change
in water cut of producer 2 (Figure 6.4). This caused significant change in the calculated
optimum allocation factor for each period (Figure 6.5). As a result, we can see
fluctuation in the production rate (Figure 6.1). The same problem happened when the
water cut and the connectivity measurement were used together for water allocation
management, as reported in the previous chapter. In order to solve this, we changed the
objective function from maximising daily oil production to maximising cumulative oil
production.
Figure 6. 1: Oil production rate for base case and WAM by LP.
144
Figure 6. 2: Cumulative water production for base case scenario and WAM by LP.
Figure 6. 3: Oil recovery versus time for base case scenario and WAM by LP.
145
Figure 6. 4: Water cut development of the 3 production wells for WAM scenario by LP.
0.6
0.5
0.4
I1
0.3
I2
0.2 I3
I4
0.1
0
25 30 35 40
Time (years)
Figure 6. 5: Calculated water allocation factor at the end of each 5-year period.
The next approach to be studied is based on maximising the cumulative oil production.
Cumulative liquid production from producer i is:
Q=
l ,i ql ,i × t (Equation 6.29)
146
in which t is the production period. The cumulative oil production from producer i is:
=
Qo ,i COCi × Ql ,i (Equation 6.30)
In equation 6.27 COCi is the cumulative oil production cut of producer i, which is
defined by:
𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝑂𝑖𝑙 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑟 𝑗

𝐶𝑂𝐶𝑗 = (Equation 6.31)
𝐶𝑢𝑚𝑢𝑙𝑎𝑡𝑖𝑣𝑒 𝐿𝑖𝑞𝑢𝑖𝑑 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑟 𝑗
The objective function is to maximise the total cumulative oil production of the
reservoir:
N prod
=
Qo ,res ∑Q
i =1
o ,i → Max (Equation 6.32)
Since change in COR is not significant and it will respect the previous production
performance of the producer in comparing with water cut, there is no fluctuation in
calculated optimum allocation factor (Figure 6.6).
0.6
0.5
0.4
I1
0.3
I2
0.2 I3
I4
0.1
0
25 30 35 40
Time (years)
Figure 6. 6: Calculated allocation factor for each period based on optimizing cumulative oil production as
objective function.
147
6. 1: Comparison of improvement in oil production rate between base case, WAM by LP and WC and LP and
COC.
Figure 6. 7: Oil recovery versus time for base case, WAM by LP and WC and LP and COC.
As a result, stable daily oil production is obtained from this method and the cumulative
oil production has been improved (Figures 6.7 to 6.11).
148
Application of LP combined with MLR and the CRM model improves waterflooding
efficiency in terms of more oil recovery and less water production. LP problems are
sensitive to constraints (including technological limits). Incompatible constraints result
in infeasible problems (i.e., with no feasible solution).Better results are obtained from
optimisation based on COR, compared with those based on WC, for both CRM and MLR
cases.
Figure 6. 8: Cumulative water production versus time for base case, WAM by LP and WC and LP and COC.
Figure 6. 9: Change in producer’s water cut in case of WAM with LP and COC.
149
6.6 New optimisation using Genetic Algorithm (GA) and reservoir simulator
There are two important issues concerning the combination of LP with MLR or CRM to
optimise water allocation. Firstly, CRM or MLR involve a very simplified representation
of the reservoir, compared with a reservoir simulator. Secondly, LP is a gradient based
optimiser. This kind of optimisation method is very good at finding local optima , but
they may fail to obtain the global optimum. A new optimisation is proposed, to
overcome these limitations; a combination of genetic algorithm and reservoir simulator.
In general, this new approach is the same as the previous one, but the production rate
estimators (MLR and CRM techniques) are replaced by the reservoir simulator and the
optimization algorithm is changed from LP to GA.
Genetic Algorithms (GAs) are adaptive heuristic search algorithms premised on the
evolutionary ideas of natural selection and genetics [11]. The basic concept of GAs is
designed to simulate processes in a natural system necessary for evolution, specifically
those that follow the principles first laid down by Charles Darwin of survival of the
fittest. As such they represent an intelligent exploitation of a random search within a
defined search space to solve a problem. There have been several application of GAs in
petroleum engineering [10-13].
Fortunately, the Reveal reservoir simulator has a feature that allows us to control the
simulation by defining a batch file. A script is developed that can change the control
variable in Reveal, run the simulation and read the results and calculate the objective
function. This script is also equipped with an in-house GA optimiser that maximises the
objective function by changing defined control variables.
The objective function was maximizing the oil recovery by changing the injection rate
of the injection wells. Producers were producing with constant liquid production rate.
Injection was controlled by voidage replacement. There were maximum and minimum
limits for injection rates. As in the previous work the simulation is run for 20 years
without allocation management and then a GA is used to find the optimum injection rate
for the next 20 years.
The optimum solution was obtained after nearly 50 iterations, with 10 simulations run
in each iteration (Figure 6.11and 6.12). The sequential run time was 16 hours with one
150
CPU; of course it could be reduced by parallel calculation. Figures 6.12 to 6.14
compare the results of these two approaches for water allocation management.
0.372
0.37
0.368
Recovery factor (fraction)
0.366
0.364
0.362
0.36
0.358
0.356
0.354
0.352
0 10 20 30 40 50 60
Number of iterations
Figure 6. 10: GA solution development over 50 iterations.
0.6
0.5
0.4
0.3
0.2
0.1
0
I1 I2 I3 I4
Injectors
Figure 6. 11: Optimum water allocation factor obtained from GA for each injection well.
151
Figure 6. 12: Plot of oil production rate versus time for different WAM scenarios.
Figure 5. 36: Plot of oil recovery rate versus time for different WAM scenarios.
152
Figure 5. 37: Plot of cumulative water production versus time for different WAM scenarios.
Waterflooding performance was improved in terms of increased oil recovery and

less water production by GA-simulation technique. Comparing the results of OPI-
CRM water allocation management with LP-CRM indicates better improvement in
water injection performance by LP-CRM. This shows that optimisation based on
future forecast of production works better than water allocation management based
on previous history of production.
Comparison of water allocation optimisation with GA-simulation and LP-CRM with

COC shows that only small improvement obtained from GA-simulation. But the
main disadvantage of GA-simulation was the substantial increase in computation
time. Therefore in very large and more complex reservoir LP-CRM is an effective
tool for water allocation management. It is easy to apply and very fast to determine
the optimum solution. And if sophisticated reservoir simulators with powerful
computers are available then GA-simulation can be employed to determine optimum
water allocation factor. In this case LP-CRM can be used to determine the initial
guess for GA-simulation technique in order to reduce the computational time.
153
6.7 References

2003. 6(1): p. 6-16.

Texas.
4. Sayarpour, M., C.S. Kabir, and L.W. Lake, Field Applications of Capacitance-
Resistance Models in Waterfloods. SPE Reservoir Evaluation & Engineering,
2009. 12(6): p. pp. 853-864.
5. Ferris, M.C., O.L. Mangasarian, and S.J. Wright, Linear programming with
MATLAB. 2007, Philadelphia: Society for Industrial and Applied Mathematics :
Mathematical Programming Society.
6. ARONOFSKY, J.S., Linear Programming A Problem-Solving Tool for

Petroleum Industry Management. 1962.
7. Lo, K.K., G.P. Starley, and C.W. Holden, Application of Linear Programming to
Reservoir Development Evaluations. SPE Reservoir Engineering, 1995. 10(1): p.
52-58.
8. Eeg, O.S. and T. Herring, Combining Linear Programming and Reservoir

Simulation to Optimize Asset Value, in SPE Production Operations Symposium.
1997, 1997 Copyright 1997, Society of Petroleum Engineers, Inc.: Oklahoma
City, Oklahoma.
9. Lang, Z.X. and R.N. Horne, Optimum Production Scheduling Using Reservoir
Simulators: A Comparison of Linear Programming and Dynamic Programming
Techniques, in SPE Annual Technical Conference and Exhibition. 1983, 1983
Copyright 1983 Society of Petroleum Engineers of AIME: San Francisco,
California.
10. Maschio, C., L. Nakajima, and D.J. Schiozer, Production Strategy Optimization
Using Genetic Algorithm and Quality Map, in Europec/EAGE Conference and
Exhibition. 2008, Society of Petroleum Engineers: Rome, Italy.

Part 2—Evolutionary Computing. Journal of Petroleum Technology, 2000.
52(10): p. 40-46.
154
12. Romero, C.E., et al., A Modified Genetic Algorithm for Reservoir
Characterisation, in International Oil and Gas Conference and Exhibition in
China. 2000, Copyright 2000, Society of Petroleum Engineers Inc.: Beijing,
China.
13. Montes, G., P. Bartolome, and A.L. Udias, The Use of Genetic Algorithms in
Well Placement optimization, in SPE Latin American and Caribbean Petroleum
Engineering Conference. 2001, 2001,. Society of Petroleum Engineers Inc.:
Buenos Aires, Argentina.
155
Chapter 7– Waterflooding Management: New Case Study
In the previous chapters, new workflows have been developed in order to improve
waterflooding performance based on the analysis of the injection and production
history. It has been realised that waterflooding management can be performed by:
reservoir voidage management (RVM), water allocation management (WAM) and
production allocation management (PAM). The proposed techniques were applied to a
small reservoir model and the results showed significant improvement in the
performance of the water injection.
In this chapter we show how we applied those techniques to a real field reservoir model.
This study aims to evaluate the added value of these new methodologies on a more
complex reservoir model.
7.1 Model description
The reservoir model has 8 producers supported with 4 injectors. The properties of the
model are given in Table 7.1 and Figures 7.1 and 7.2.
Unfortunately there was no information available about the geology and the formation
structure of the model.
Table 7. 1: Reservoir model properties
Number of cells 41×80×21

Initial reservoir pressure 5692 psia
Temperature 208o F
Oil density 40 API
156
Figure 7. 1: Permeability variation in the reservoir model.
Figure 7. 2: Porosity variation in the reservoir model.
157
7.2 Reservoir voidage management (RVM)
The semi-analytical method developed in Chapter 2 is applied for RVM. The base case
scenario is run for 30 years with the following injection and production rates schemes:
water has been injected with total voidage replacement ratio of 1; this water is equally
allocated to the injectors; production is controlled by constant liquid production. Table
7.2 shows the allocated liquid production rate of each production wells. Simulation
results showed that water breakthrough will happen after cumulative injection of
25,130,960 STB water.
At first the minimum required reservoir pressure was determined by running a

sensitivity analysis on well outflow performance (Figure 7.3). According to this
analysis, the minimum required reservoir pressure for this reservoir to support
production in producers is 2800 psi.
Table 7. 2: Allocated liquid production rate of each production well for base case scenario.
Production Liquid production rate

well (STB/day)
P1 16000
P10 14000
P3 7000
P4 12000
P5 16000
P7 5000
P8 10000
P9 13000
Equation 2.23 is used to determine the time of the starting pressure maintenance
(injection with VR of 1) (tpm). Table 7.4 shows the calculated tpm and the cumulative
injected water before reaching the Prmin (Qipm)for each VR. Calculated Qipm for each VR
is bigger than the cumulative injected water before water breakthrough. Therefore,
water breakthrough will occur before the start of pressure maintenance for all of the
VRs. In order to maximise the breakthrough time we need to inject as little as we can.
The proposed voidage replacement plan for this field is to inject with VR 0.1 until the
reservoir pressure declines to minimum level (2587 days) then start to inject with VR of
1 for the rest of the production period.
158
Figure 7.3: Well outflow sensitivity analysis shows that the minimum required reservoir pressure is 2800 psi.
Table 7. 3: Calculated tpm and Qipm for each VR.
VR tpm Qipm
0.9 23288.82 2971700607
0.8 11644.41 1320755826
0.7 7762.942 770440898
0.6 5822.206 495283435
0.5 4657.765 330188956
0.4 3881.471 220125971
0.3 3326.975 141509553
0.2 2911.103 82547239.1
0.1 2587.647 36687661.8
159
7.3 Production allocation management (PAM)
The downhole production rate of the producers is used to define new production
allocation factor for production wells. PAM has been done based on the oil relative
production ratio (OPR) and was applied after starting of the pressure maintenance.
Table 7.4 shows the calculated OPR for each of the production wells. The new and old
production allocation factors are shown in Figure 7.4.
0.25
0.2
0.15
Oil relative production ratio
0.1
0.05
0
-0.05 P1 P10 P3 P4 P5 P7 P8 P9
-0.1
-0.15
-0.2
-0.25
-0.3
production wells
Figure 7. 4: Calculated OPR for all production wells.
0.25
Old PAF
Production allocation factor
0.2
New PAF
0.15
0.1
0.05
0
P1 P10 P3 P4 P5 P7 P8 P9
Production wells
Figure 7.5: Old and new values of PAF for each producer.
160
7.4 Water allocation management (WAM)
WAM was employed by LP-CRM method. 15 years of injection and production history
were used to determine the coefficients of CRM for each individual production well.
The following Figures 7.6 to 7.9 show the CRM estimated production rate versus the
original production rate of some of the producers.
21000
P1 simulation
20500
Liquid production rate (bbl/day)
P1 CRM
20000
19500
19000
18500
18000
17500
17000
0 1000 2000 3000 4000 5000 6000
Time (days)
Figure 7. 6: Calculated production rate from CRM versus production history of Producer 1 from reservoir
model.
9600
P3 simulation
9400
P3 CRM
Liquid production (bbl/day)
9200
9000
8800
8600
8400
8200
8000
7800
0 1000 2000 3000 4000 5000 6000
Time (days)
model.
161
16000
P4 simulation
15500
Liquid production rate (bbl/day)

P4 CRM
15000
14500
14000
13500
13000
12500
0 1000 2000 3000 4000 5000 6000
Time (days)
model.
13500
P8 simulation
13000
P8 CRM
Liquid production (bbl/day)
12500
12000
11500
11000
10500
10000
0 1000 2000 3000 4000 5000 6000
Time (days)
Figure 7.9: Calculated production rate from CRM versus production history of Producer 8 from reservoir
model.
Figure 7.10 shows the calculated coefficients of CRM for each production well. LP was
then performed to determine the optimum water allocation factor for the injection wells.
The objective function was to maximise the cumulative oil production. The new
allocation injection allocation factor is shown in Figure 7.11.
162
1.2
0.8
I1
0.6 I2
I3
0.4 I4
0.2
0
P1 P10 P3 P4 P5 P7 P8 P9
Figure 7. 10: Calculated coefficient of CRM for all production wells.
0.8
0.7
0.6
WAF (fraction)
0.5
0.4
0.3
0.2
0.1
0
I1 I2 I3 I4
Figure 7. 11: New water allocation factor for injection wells obtained from LP-CRM technique.
7.5 Water flooding management (WFM)
At the end of this study all the proposed techniques for waterflood management (RVM,
WAM and PAM) have been employed simultaneously. This scenario was called
waterflooding management (WFM).
7.6 Results and Discussions
The following Figures compare the results of each individual management workflow
together.
163
Injection efficiency improved a little by PAM, and much better improvement was
obtained from WAM but the best performance was achieved by RVM.
PAM aims to increase the liquid production from the wells with good oil production and
less water production. But if we increase the production from these wells we also
increase the water production too. And because of this we cannot get significant
improvement from PAM, as it will also increase water production.
Figure 7. 12: Comparison of oil recovery obtained from different injection scenarios.
164
Figure 7. 13: Comparison of oil production rate obtained from different injection scenarios.
Figure 7. 14: Comparison of cumulative water production obtained from different injection scenarios.
WAM and RVM will affect the displacement efficiency of the water injection. This will
cause more oil be pushed toward the producers without changing their production rate.
165
So, more oil will be produced and water production will reduce. Both WAM and RVM
are improving sweep efficiency but they will act differently. RVM will improve
waterflood movement, helping to leave less oil behind, whereas; WAM guides the water
front in the right direction. And because of this, when we apply WAM and RVM
together, we will get a very significant improvement in injection efficiency. RVM is like
a “good driver” in a racing car and WAM is a “good navigator”. Finally, this study
shows how effectively these simple workflows (WFM) can be employed to obtain a
successful waterflood project in real field models.
166
Chapter 8– Conclusions and Recommendations
8.1 Research story
This thesis has evaluated a wide range of techniques to mitigate one of the big
challenges of petroleum industry, water production. The techniques discussed for
waterflooding management, all aim to reduce excessive water production. The injection
and production history at a well and field level are the most common available data in
any oil field, especially when nowadays we can have these data in real time with the
implementation of the digital oil field and the intelligent well completion. This research
aims to understand the strength and weaknesses of the existing techniques and
“repackage” them to provide an optimum combination for more effective waterflood
management by analysing injection and production data history.
The first part of this research reviewed, tested and compared the analytical techniques
that have been previously used for analysing the injection and production. The methods
studied fell in to two distinct classes: those that monitor the waterflood performance
secondly, and those that determine the inter-well connectivity.
The second part of this thesis showed that an improved workflow, using the captured
information from the phase one methods, could be combined to give a more effective
waterflood management via combination of reservoir voidage management (RVM),
water allocation management (WAM) and production allocation management (PAM).
Finally, a semi-analytical method was introduced in this thesis for performing RVM.
Two approaches were defined for WAM and new techniques developed for PAM, all of
which employed only the production and injection history. The results from these
techniques were compared with the more advanced reservoir simulation methodologies
such as gradient free optimisation. This comparison showed the reliability of the
proposed techniques.
167
8.2 Monitoring the performance of waterflooding
8.2.1 Summary
Different techniques for monitoring the waterflood have been reviewed. These
techniques were the Cut-Cum plot, the WOR plot, the X-plot and Hall plot. The first
three can be used to forecast future oil reserve by knowing the amount of the water
injected to the reservoir, while the Hall plot can be employed to monitor the
performance of each individual injection wells. Finally a semi-analytical methodology
was proposed for reservoir voidage management (RVM).
8.2.2 Conclusions
1. The X-plot provided the most accurate estimation.

The results and accuracy of all three techniques were judged to be acceptable but
the best future predictions were obtained from the X-plot. The X-plot has the
additional advantage that it can also be used to generate an effective
permeability ratio plot too. This study thus recommends X-plot for preparation
of future oil recovery forecast. However, it must be mentioned that this method
is only reliable for water fraction values of higher than 50%. Also, only that
portion of data related to the new scheme should be used for the analysis, so that
there is a significant change in the injection strategy.
It was noted that the Cut-Cum produced better predictions than WOR plot, but
its practical application was found to be dependant to the form of the polynomial
employed to fit the data.
2. Hall plot can be employed only to monitor the change in the conditions of
injection wells. It will not give any information about the future oil reserve
or recovery.
This study showed that Hall plot provided good information about the
performance of the injection well such as injectivity index. But it could not give
any information about the overall performance of the waterflood in terms of
improving the oil recovery or decreasing the water production.
3. None of the studied techniques were able to determine the optimum value
for key parameters affecting the water injection efficiency.
The first three techniques were very good at differentiating between the
performance of base case injection plan and the improved performance whan
168
water allocation management (WAM) was implemented. They clearly showed
that the performance of the reservoir or the injection wells had changed,
improved or deteriorated. However, they were not capable of advising which
change should be made to the parameters governing the injection efficiency in
order to improve the overall efficiency of the injection scheme.
4. Determination of the minimum required reservoir pressure and the time of
breakthrough were showed to be two important within the Reservoir
voidage management (RVM) parameters whose optimisation significantly
can improved the waterflood’s performance.
A high volume injection of water from the start of the production maintains
reservoir pressure, but also causes early water breakthrough. This can cause a
poor well outflow performance and reduces the efficiency of the water/oil
displacement process. On the other hand, a reduced rate of injection of water
may improve the sweep efficiency and delay the water breakthrough but will not
support the reservoir pressure sufficiently to maintain the required production
rate from the reservoir. Reservoir voidage management improves flooding
efficiency. It will increase oil production, reduce excessive water production and
require a low injection volume of water. This understanding lead to a proposed
new workflow based on identifying the optimum reservoir pressure in terms of
well’s outflow performance and the time to water breakthrough, in order to
increase the water injection efficiency.
8.3 Inter-well connectivity measurement: Statistical techniques

8.3.1 Summary
Three common statistical techniques have been tested: the Spearman rank correlation,
multi-linear regression (MLR) and capacitive resistive model (CRM). A superposition
approach was proposed to identify the injectors connected to each producer followed by
the application the above techniques to quantify this injector/producer connectivity.
Connectivity results can then be combined with the water cut of the production wells in
a new, but simple, algorithm to update and manage water allocation factor.
169
8.3.2 Conclusions
1. Spearman rank correlation coefficient was not able to determine inter-well

connectivity.
Only a small improvement was obtained from water allocation management

using the inter-well connectivity results delivered by this technique. It was
concluded that the Spearman rank correlation coefficient is not a suitable method
to define inter-well connection in the field studied. However it can identify those
production wells whose performance is affected more by the entire injection than
by individual injection wells.
2. It is very important to recognise which injection wells are the connected to

each production well when we measure the inter-well connectivity by the
multi linear regression (MLR) or the capacitive resistive model (CRM)
techniques.
One of the important steps in determining well connectivity is to understand the

injection well production well connection. An incorrect assessment of
connections will lead to a wrong quantitative measurement of the inter-well
connectivity by these two methods. Superposition analysis incorporated with the
location of the wells and other geological parameters can help us to ensure that
the injection/production well connections are correctly assigned
3. Inter-well connectivity measurement is a good and effective tool in water

allocation management (WAM). WAM improved flood performance.
A simple new technique is proposed to define a new water allocation factor for
each injection well based on the connectivity results of MLR and CRM coupled
with the water cut of the production wells. This methodology allocates a greater
volume of water to injection wells that are well connected to a production well
with a low water cut. WAM improved the waterflood performance by producing
more oil (improved sweep efficiency) and reducing the water production.
4. WAM based on CRM connectivity results gave the greatest improvement
The injection efficiency of WAM based on CRM gave better results than MLR
because it considers extra parameters, particularly the production well’s
productivity index. By calculating the time constant associated with CRM the
170
model also respects the delay time between the injector and producer. Therefore
CRM was preferred above MLR for determining the inter-well connectivity.
8.4 Inter-well connectivity measurement: Artificial intelligence techniques

8.4.1 Summary
Two types of networks designed and optimised to forecast production rate by inputting
the injection rate. The networks were feed-forward back propagation network (FFBP)
and Co-active neuro-fuzzy inference system (CANFIS). Super position methodology
used again to identify the connected injectors to each producer. In order to determine
the inter-well connectivity sensitivity analysis carried on each producer. Connectivity
results of these techniques combined with water cut to do WAM.
8.4.2 Conclusion
1. Both Feed-forward back propagation (FFBP) network and fuzzy logic

network (CANFIS) were acceptable for determining inter-well connectivity
Injection performance was improved by applying the FFBP and CANFIS based
WAM. Therefore they can be employed to determine inter-well connectivity.
However FFBP was better than CANIS
2. Statistical techniques (MLR and CRM) were better than FFBP.
FFBP based WAM improved oil production and reduce the production of water.
However MLR and CRM were even better than FFBP at measuring the inter-
well connectivity as well as employing a simpler calculation workflow. Inter-
well connectivity can be obtained directly from MLR and CRM while a
sensitivity analysis was required to determine the connectivity with FFBP. CRM
is thus recommended for calculating the inter-well connectivity.
8.5 New development in water allocation management (WAM) and production

allocation management (PAM)
8.5.1 Summary
The study described above showed that, unsurprisingly, the production well’s water cut
can be altered by changing the water allocated to the producers which were highly
affected by injectors. Also it did not necessarily fully respect the previous performance
of the well. A search was thus made for a better parameter that describes the
171
performance of the production wells. This parameter should separate a good producer
from a bad producer more clearly. New parameters that were chosen for were:
cumulative water cut (CWC), relative oil production ratio (OPR) and oil production
index (OPI). In all cases WAM based on CRM connectivity results was used to test the
newly defined parameters.
New techniques have also been developed for production allocation management with
the above parameters (PAM). Finally production and water allocation management
(W&P AM) have been done simultaneously.
8.5.2 Conclusions
1. CWC can be used for long-term WAM while WC will be better for short
term WAM.
Both WC and CWC gave improved results. But at the beginning WC worked
better than CWC at the early time, but, as expected CWC performed above WC
after longer production intervals. It is also preferred since the allocation factor
obtained from CWC was more stable to changes in performance of the
production wells than the value obtained from WC. WC thus is the preferred
short-term parameter while CWC should be applied for longer time
management.
2. New WAM methods employing OPR and OPI provide significant
improvements compared to WC and CWC with OPI giving slightly better
results.
WAM methods employing the newly defined parameters OPI and OPR were
very effective in improving injection efficiency. The power of OPR and OPI is
that they provide a greater separation of good and bad production wells since
they can assign a negative value to bad producers. This greater differentiation
ensured that less water was allocated to the injectors connected to the producers
with highest water production with more water went to support production wells
with a higher level of oil production. More oil and less water production
resulted. OPI was more effective at this since it gives more difference between
“good” and “bad” producer.
3. A clear differentiation between good producers and bad producers is very
important for efficient WAM.
172
WAM based on OPI and OPR gives a greater differentiation more between
“good” production wells and “bad” production wells. This allows application of
more “punishment” to the injectors supporting “bad” production wells,
delivering the greatest improvement in the waterflood performance
4. Improvement from WAM is better from PAM.
WAM is more effective at improving the long-term waterflood efficiency (more
cumulative oil and less cumulative water production) than PAM. PAM delivers
the greater short-term oil production while the oil production improvement only
builds up slowly in the case of WAM. This is because there will be a
considerable lag time before increase the production rate as a result of changes
to the injection strategies (the reservoir acts at a much longer time scale than a
single well). The reduction in water production was bigger than the increase in
oil production for WAM when compared to PAM.
5. The combination of WAM and PAM (W&P AM) provided best performance.
W&P AM improved significantly the performance of the analysed reservoir
models. It was much more effective at increasing the cumulative oil production
(the amount of improvement was two times more than WAM and PAM) rather
than decreasing water production. This is because PAM is better in improving oil
production rather than decreasing water production.
8.6 Water allocation optimisation

8.6.1 Summary
All the above studies performed WAM based on the inter-well connectivity results and
monitoring the resulting production well performance. An alternative approach is to use
water allocation factors based on a future forecast of the production rates following the
changes in the injector’s injection rates. Such techniques use the injection rates as an
input to estimate future liquid production rate in the production wells (MLR & CRM)
coupled with an optimiser engine to optimise the injection rate. Reviewing MLR and
CRM shows that both of them, along with their constraints and the objective function
are linear. Therefore linear programming (LP) based optimiser can then be used to find
the optimum allocation factor for the injectors. Two objective functions were
considered; maximising daily rate and maximising the cumulative oil production.
173
Finally, the results from the LP-CRM study compared with those from the combination
of Genetic algorithm (GA) optimiser with reservoir simulator.
8.6.2 Conclusions
1. The second approach of WAM (LP-CRM) was better than the previous
methodology based on connectivity measurement (OPI-CRM).
LP-CRM guided WAM improved the overall water injection performance even
better than the previous methodology based on injecting more to the injectors
connected to good producers (OPI-CRM). Allocation management based on
future performance estimation is thus better than allocation based on past
performance.
2. LP-CRM was very sensitive to the defined objective function.
Objective function has a significant effect on the determination of the optimum
solutions. Maximisation of daily oil rate is thus a better objective function for
short term optimisation while maximising the cumulative oil production was
better for long-term allocation management.
3. LP-CRM is recommended as an effective tool for WAM.
A marginal but not significant improvement was obtained from a GA-simulation
combination compared to an LP-CRM based allocation management. The
computation time of the GA-simulation was 24 hours, compared to less than 15
minutes for the LP-CRM. The LP-CRM thus is recommended as an effective and
simple methodology for water allocation management especially when a reliable
reservoir model is not available, or for a very large reservoir models containing
many of wells when the required computational time is not practical. LP-CRM
prediction will also be very useful even when powerful computers with
sophisticated reservoir models are available since it can be used to define the
first guess for the GA-reservoir simulator to reduce the computing time.
8.7 Water flooding management: new case study:

8.7.1 Summary
The proposed techniques for PAM, WAM and RVM were then applied to a more
complex real field model to evaluate the added value of waterflood management.
174
8.7.2 Conclusions
1. RVM is a “good driver” and WAM is a “good navigator”.
PAM aims to increase the liquid production from the wells with good oil production
and less water production. But if we increase the production from these wells we
also increase the water production too. And because of this we can not get
significant improvement from PAM as it will also increase the water production.
WAM and RVM will affect the displacement efficiency of the water injection. This
will cause more oil be pushed toward the producers without changing their
production rate. So, more oil will be produced water production will reduce. WAM
and RVM will both are improving sweep efficiency but they will act differently.
RVM will improve waterflood movement helping to leave less oil behind on the
other hand; WAM put the water front in the right direction. And because of this
when we apply WAM and RVM together we will get very significant improvement
in injection efficiency
8.8 Recommendations for future works:
1. New techniques have been developed for RVM, WAM and PAM based on
production and injection rate data. These techniques can be used in principle to
design a closed loop water injection management based on intelligent
completion. So a study of the feasibility of designing such a closed loop water
allocation management and the added value of it can be a good area for further
research.
2. The proposed techniques were based on exercising on reservoir models. So it is
strongly recommended to validate the proposed techniques in real forecasting
cases where real field injection and production history are available.
3. In this study we tried to calculate the connection between injection and

production wells at the well scale (whole reservoir section). But intelligent
completion will provide us with monitoring and control at a layer scale; hence
evaluation of new inter-well connectivity at each individual can be incorporated
in the overall waterflood strategy at each individual layer.
175
4. The role of an active aquifer was not considered in this study. Allocation
management in presence of active aquifer in the reservoir is another area for
study.
5. This research assumed that no new production or injection wells were drilled
during the life of the field. A systematic Study of the effect of new drilling on
well connectivity can be another useful subject for water allocation
management.
6. The effect of different injection schemes on determination of well connectivity
can be another subject for further research such as injection with constant
bottom-hole pressure or constant surface injection rate.
7. There are other techniques that have the potential to be employed to determine
inter-well connectivity. One of these possible methods could be advanced
interpolation techniques such as kriging.
176

Improved Oil Production and Waterflood Performance by Water Allocation Management

Uploaded by

Document Informationclick to expand document informationProd

Document Informationclick to expand document information

Copyright:

Available Formats

Improved Oil Production and Waterflood Performance by Water Allocation Management

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Improved Oil Production and Waterflood Performance by Water Allocation Management

Uploaded by

Copyright:

Available Formats

Improved Oil Production and

Waterflood Performance by Water

A thesis submitted for the degree of Doctor of Philosophy

Institute of Petroleum Engineering

I dedicate this thesis to my wife and my parents for their continuous

Foremost, I would like to express my gratitude to my supervisor Prof. David. R.

LISTS OF TABLES AND FIGURES...................................................................... vii

LIST OF ACRONYMS ............................................................................................ xx

NOMENCLATURE ................................................................................................. xxi

Chapter 1 – Introduction ................................................................................................... 1

1.1 Water production ................................................................................................ 1

1.2 Waterflooding management ............................................................................... 3

1.3 Intelligent Well Completion (IWC) and waterflooding management ................ 4

1.4 Research objectives ............................................................................................ 6

Chapter 2 - Monitoring The Performance Of Waterflooding ........................................... 9

2.1 Monitoring and surveillance of water injection projects .................................... 9

2.1.3 The X-Plot ................................................................................................. 12

2.1.4 The Hall Plot ............................................................................................. 15

2.2 Case study......................................................................................................... 17

2.2.1 Model description ..................................................................................... 18

2.2.1.1 Base case scenario ............................................................................. 20

2.2.2 Plot Analysis ............................................................................................. 20

2.2.2.1 Cut-Cum plot ..................................................................................... 22

2.2.2.2 WOR plot ........................................................................................... 26

2.2.2.3 X-plot ................................................................................................. 30

2.2.2.4 Hall plot ............................................................................................. 32

2.3 Reservoir voidage management ....................................................................... 35

2.3.1 Model description ..................................................................................... 35

2.3.2 Methodology for reservoir voidage management ..................................... 36

2.4 Results and discussion ...................................................................................... 44

2.5 References ........................................................................................................ 46

Chapter 3 – Producer-Injector Inter-well Connectivity Measurement; Statistical

3.1 Producer-injector inter-well connectivity ......................................................... 48

3.2 Inter-connectivity measurement ....................................................................... 49

3.2.1 Spearman rank correlation coefficient ...................................................... 49

3.2.2 Multi-linear regression (MLR).................................................................. 50

3.2.3 Capacitive resistive model (CRM) ............................................................ 52

3.3 Case study......................................................................................................... 58

3.3.1 Spear Rank correlation coefficient ............................................................ 60

3.3.2 Multi-linear regression .............................................................................. 62

3.4 Allocation management .................................................................................... 73

3.5 Results and discussion ...................................................................................... 76

3.6 References ........................................................................................................ 79

Chapter 4 – Producer-Injector Inter-well Connectivity Measurement; Artificial

4.1 Artificial neural networks (ANNs) .................................................................... 81

4.1.1 Structure of ANNs .................................................................................... 82

4.1.2 Classification of ANNs ............................................................................. 83

4.1.3 Application of ANN .................................................................................. 84

4.1.3.1 Pattern classification .......................................................................... 84

4.1.3.2 Clustering ........................................................................................... 84

4.1.3.3 Function approximation (modelling) ................................................. 85

4.1.3.4 Forecasting ......................................................................................... 86

4.1.4 Application of ANN in petroleum engineering......................................... 86

4.2 ANN and inter-well connectivity measurement ............................................... 87

4.2.1 Feed-forward back propagation (FFBP) ................................................... 88

4.2.2 Fuzzy logic network (Co-adoptive neuro-fuzzy inference system

4.3 Case Study ........................................................................................................ 88