PTE 586 Project

29
Net Present Value Prediction Using Artificial Intelligence Group 4 Tongxin Zhang Ahmed Aljeshi Lawrence Bustos Olufemi Hussain Mesut Yildiz (TL)

Transcript of PTE 586 Project

Page 1: PTE 586 Project

Net Present Value Prediction Using Artificial Intelligence

Group 4 Tongxin Zhang Ahmed Aljeshi Lawrence Bustos Olufemi Hussain Mesut Yildiz (TL)

Page 2: PTE 586 Project

Outline

○  Introduction

○  Neural Network

○  Data

○  Net Present Value

○  Softwares

○  Results

○  Recommendations

Page 3: PTE 586 Project

Introduction

○  Objective: Create an alternative method to predict

Net Present Value of a well using Artificial Intelligence

○  Net present value (NPV) shows economic

feasibility

Page 4: PTE 586 Project

Neural Network- SPE Papers

○  Solution: •  Predicting Oil and Gas Spot Prices using Chaos

Time Series Analysis and Fuzzy Neural Network Model

•  Forecasting, Sensitivity and Economic Analysis of

Hydrocarbon Production from Shale Plays Using Artificial Intelligence & Data Mining

Page 5: PTE 586 Project

Neural Network- Supervised

Page 6: PTE 586 Project

Neural Network- Attributes

1 Hidden Layer 6, 12, 18 Neurons Vs 100 Neurons

Page 7: PTE 586 Project

Finding Data

○  No real example of both production and

economics.

○  Used Pleito oilfield data

○  Traditional NPV Calculation

Page 8: PTE 586 Project

Pleito Field- Location

○  Operator: Vintage Production California LLC

Page 9: PTE 586 Project

Pleito Well- 48-25 (DOGGR)

Production: 1977-2013 (Active) Cumulative Gas: 612,540 Average Oil Rate: 3,508 Barrels GOR: 400.6 SCF/Barrel Cumulative Oil: 1,550,365 Ave Water Cut: 40%

Page 10: PTE 586 Project

Data Used- Actual Excel

○  Real production data from a well in Pleito

oilfield in California

○  Calculate NPV - assumptions

• CAPEX, OPEX

• Tax, royalty, interest rates

• Oil and gas prices are actual

Page 11: PTE 586 Project

NPV Calculation - Equations

○  Gross revenue = produced volumes x unit price

○  Net revenue = (Gross Revenue) – (Royalty) ○  Gross Income= (Net Revenue) – (OPEX) ○  Taxable Income= (Gross Income) –

(Depreciated/Amortized CAPEX) ○  Net Cash= (Gross Income) – (Tax) ○  Net Present Value =

Page 12: PTE 586 Project

Softwares - Matlab

6 Neurons 12 Neurons 18 Neurons

Neural Network Toolbox

Page 13: PTE 586 Project

Softwares - Matlab

○  Training Function:

Levenberg-Marquardt backpropagation

○  Activation Function:

Hyperbolic tangent sigmoid function

Graph and Symbol

Page 14: PTE 586 Project

Matlab Results

Validation Results Comparison(2008-2010)

Page 15: PTE 586 Project

Matlab Results

Validation Results Comparison(2008-2010)

Page 16: PTE 586 Project

Matlab Results

Prediction Results Comparison(2011-2013)

Page 17: PTE 586 Project

Matlab Results

Prediction Results Comparison(2011-2013)

Page 18: PTE 586 Project

Matlab Results

Training Validation and Prediction Results Comparison (1977-2013)

Page 19: PTE 586 Project

Matlab Results

Regression Comparison

6 Neurons 12 Neurons 18 Neurons

R=0.9738 R=0.9811 R=0.9911

Page 20: PTE 586 Project

NEURAL NETWORK SOFTWARE

Page 21: PTE 586 Project

PRESENT VALUE DATA

Actual  Avg.  Present  Value   So3ware  Avg.  Present  Value  

TRAINING  

1977   24385.80478   24385.80478  1978   16303.96248   16303.96248  1979   21513.6047   21513.6047  1980   31047.59555   31047.59555  1981   31632.88675   31632.88675  1982   52263.0045   52263.0045  1983   40537.34312   40537.34312  .   .   .  .   .   .  .   .   .  .   .   .  

2004   6959.993386   6959.993386  2005   8277.689996   8277.689996  2006   7952.507929   7952.507929  2007   8226.340719   8226.340719  

VALIDATION  2008   9301.399161   9301.399161  2009   2609.61142   2609.61142  2010   4472.392672   4472.392672  

FORECASTING  2011   4910.086479      

2012   1905.528339      

2013   2270.595203      

Page 22: PTE 586 Project

VALIDATION PRESENT VALUE

Page 23: PTE 586 Project

Actual  NPV   Predicted  NPV   Error  

2011  

Jan-­‐11   6254.210998   3968.980439   0.365  Feb-­‐11   4210.616433   3047.597601   0.276  Mar-­‐11   6678.912587   4009.373653   0.400  Apr-­‐11   5752.610198   3124.110503   0.457  May-­‐11   5967.42932   2654.518389   0.555  Jun-­‐11   5754.404235   2144.55317   0.627  Jul-­‐11   5938.082057   1918.349572   0.677  Aug-­‐11   5341.303176   2459.746053   0.539  Sep-­‐11   5010.045412   3011.531335   0.399  Oct-­‐11   2072.744265   4279.799643   1.065  Nov-­‐11   2739.046112   5254.787882   0.918  Dec-­‐11   3201.632958   5734.639865   0.791  

2012  

Jan-­‐12   1816.91691   5591.116417   2.077  Feb-­‐12   166.1086102   5763.480279   33.697  Mar-­‐12   1447.199972   4964.070376   2.430  Apr-­‐12   2269.753447   4567.958398   1.013  May-­‐12   1950.825323   3718.48113   0.906  Jun-­‐12   1947.415668   2728.736722   0.401  Jul-­‐12   2188.639601   3365.730407   0.538  Aug-­‐12   2051.109089   2983.073937   0.454  Sep-­‐12   2325.984921   3162.876587   0.360  Oct-­‐12   2480.888188   2131.712137   0.141  Nov-­‐12   2121.758724   1956.521861   0.078  Dec-­‐12   2099.739618   2215.269226   0.055  

2013  

Jan-­‐13   2122.665132   2033.839355   0.042  Feb-­‐13   2215.965402   2021.75256   0.088  Mar-­‐13   2301.997505   1947.024223   0.154  Apr-­‐13   2204.113861   1707.351853   0.225  May-­‐13   2151.571775   2324.322663   0.080  Jun-­‐13   1887.566247   2121.71863   0.124  Jul-­‐13   2601.538906   2706.626824   0.040  Aug-­‐13   2786.514269   3880.555683   0.393  Sep-­‐13   2790.561057   3845.351377   0.378  Oct-­‐13   2031.388907   3912.483847   0.926  Nov-­‐13   1758.230914   3270.305518   0.860  Dec-­‐13   2395.02846   2730.511095   0.140  

2011-­‐2013  NPV:   109,035     117,259     0.075  Total  NPV:   6,440,001     6,448,225    

Page 24: PTE 586 Project

PRESENT VALUE

Page 25: PTE 586 Project

PREDICTION PRESENT VALUE

R Square = 0.5694 R = 0.7546

Page 26: PTE 586 Project

Results

Software Matlab Neuroxl Predictor

Activation Function Hyperbolic Tangent Sigmoid Function

Hyperbolic Function

Number of Neurons 6 12 18 100 neurons

Predicted NPV Error 0.039 0.034 0.012 0.075

Prediction R square 0.9483 0.9626 0.9823 0.5694

Prediction R 0.9738 0.9811 0.9911 0.7546

NPV(2011-2013) 100,007 114,483 102,181 117,259

NPV 6,394,828 6,403,693 5,562,674 6,448,225

Page 27: PTE 586 Project

Recommendations

○  Use Reservoir Data as input attributes

○  Apply decline curve analysis DCA

○  Compare with Multiple Wells

○  Comparison to Real NPV Data

○  Use in house ANN for predicting

Page 28: PTE 586 Project

Conclusion

Utilizing a Neural Network to predict the Net Present

Value for a well is a viable option compared to

traditional Methods. Overall results proved that an ANN

can predict within an 8% error.

○  Time Efficient

○  May use simulation Data

○  Dependent on Technological Resources

○  May provide a forecast for a business investment

Page 29: PTE 586 Project

Thank you