SHGAs Forecasting.pdf

8/17/2019 SHGAs Forecasting.pdf

1/39

Forecasting Production in Shale Gas Reservoirs- ABetter Assessment of ReservesOccidental Petroleum Corporation ■ GCS Reservoir Study Group■ Anadarko Petroleum Convention Center ■ 10th May ■ Krunal Joshi, Reservoir Engineer

Forecasting Production in Shale Gas Reservoirs- A Better Assessment of Reserves


2/39

Disclaimer

This presentation in no way represents or bears upon theReserves process of Oxy or any of its subsidiaries



3/39

Outline

• Problem

• Deterministic forecasting models

• Fixes to the Duong method

• Comparison of deterministic forecasting models forIndividual wells

• Comparison of deterministic forecasting models for

grouped well sets

• Conclusions



4/39

We Have a Problem

• Forecasting methods we use in conventional reservoirsmay not work well in

– Tight oil, gas

– Oil, gas shales

– Unconventional resources generally

• There have been various methods proposed



5/39

Criteria for Ideal Decline Model in Ultra-TightReservoirs

• Forecasts are reasonable and realistic for the well life

• Forecasts reasonable even with


6/39

A Superior Model Has Higher Accuracy andPrecision For a Large Number of Wells



7/39

Long-Term Horizontal Shale Gas WellSimulation: Linear Flow Plot

10

100

1000

10000

1 10 100 1000 10000

R a t e ,

M S C F / D

a y

Time, Months

Fracture Linear Flow

Fracture Interference Flow

Outer Matrix Linear

Flow

Structural Boundary Flow



8/39

Forecasting Models



9/39

Arps (Minimum Decline)

• Best-fit “b” until predetermined minimum decline ratereached; then impose exponential decline (SPE 16237)

• Problems

– Apparent “best” b decreases continually with time

– Appropriate minimum decline rate based on observed long-termbehavior in appropriate analogy – unavailable in new resource

plays

)/1()1(

1b

i

it bD

qq



10/39

SEPD/SEDM Model

• „Validated‟ for wells with both transient and stabilized flow

in Barnett Shale

• Forecasts unreliable for


11/39

Duong Model



12/39

Determination of a & m (Duong)

a=0.731 , m=1.067

a

slope = m



13/39

Determination of q1 & q ∞(o r q in f )

qinf or q∞ is the x-intercept on the above plot

q=q1t (a, m) + q∞



14/39

Duong Forecast



15/39

Field Data Set

Johnson

Tarrant

Wise Denton

• 250 Well Dataset

Barnett Shale (200wells)

• Denton

• Tarrant

• Wise

• Johnson

Fayetteville Shale (50 wells)

• Van Buuren

Drilling Info

Horizontal Wells

Monthly Rate Data

• 1st production starts 1/1/2004

• Range of total production: 30 to 85 months



16/39

Simulated Data Set

• Composite Model

– Analytical Simulator (Fekete WellTest) – SRV permeability different from Outer Matrix permeability

• Barnett (25 simulations) – 133874(Chong et al. 2010), 146876(Cipolla et al. 2011), 144357(Strickland et

al. 2011), 96917(Frantz et al. 2005), 125530(Cipolla et al. 2010)and147603(Ehlig-Economides and Economides 2011)

• Marcellus (25 simulations) – 133874(Chong et al. 2010), 125530(Cipolla et al. 2010), 144436 (Thompson

et al. 2011) and 147603(Ehlig-Economides and Economides 2011).

• Properties Varied:

– Fracture stages, fracture length and fracture conductivity.

– Stimulated Reservoir Volume (SRV) permeability

– In accordance with the ranges in the above papersForecasting Production in Shale Gas Reservoirs- A Better Assessment of Reserves


17/39

Barnett Shale Simulation (Base Case)

Hz Multifrac-Comp ModelSchematic

xfy

=200ft

xfy

=200ft

xfy

=200ft

xfy

=200ft

xfy

=200ft

Xw =-1600.0 ft

Xw =-800.0 ft

Xw =2000.0 ft

Xw =0.0 ft

Xw =800.0 ft

Xw =1600.0 ft

Y w

= 1 3 2 0

. 0 f t

Xe =4000.0 ft

Ye

=2640.0ft

100 nd

500 nd



18/39

Marcellus Shale Simulation (Base Case)

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

xfy

=150ft

Xw =-1913.5 ft

Xw =-1565.6 ft

Xw =-1217.7 ft

Xw =-869.8 ft

Xw =-521.9 ft

Xw =-174.0 ft

Xw =2087.5 ft

Xw =174.0 ft

Xw =521.9 ft

Xw =869.8 ft

Xw =1217.7 ft

Xw =1565.6 ft

Xw =1913.5 ft

Xe =4175.0 ft

Ye

=1120.0ft

920nd

200 nd



19/39

Fixes to the Duong Model

• Use of qinf – Not suggested for short term data.

– Debatable for long-term data

– Simulated data can solve the conundrum of whether qinf is

necessary or not.

• Modified Duong

– Accounts for fracture interference

– Dswitch of 5%, i.e. when decline rate reaches 5%, forecastswitches to Arps



20/39

Using qinf Does Not Work For Short Term FieldProduction Data



21/39

Using q∞ For Simulated Production Data Does

Not Work Well



22/39

Modified Duong (Dswitch @5%) Works BetterThan the Original Duong



23/39

Individual Well Field Production Data



24/39

Comparison of The Modified Duong, SEDM andArps For a Field Data Set

Discrepancy (error %) in remaining reserves for a field datasetHistory Matched Duong_Dswitch@5% SEDM Arps (Dmin 5%)

6 Mean -15.98 40.91 10.97

Std.Dev 29.24 39.06 33.16

% Wells


25/39

How Well Do Different Models Forecast WithShort Term Data ?

Comparison of various empirical models for API# 42-121-32245,

matching 12 months of historical data.



26/39

How Well Do Different Empirical ModelsForecast With Long Term Data ?

Comparison of various empirical models for API# 42-497-35453,

matching 36 months of historical data.



27/39

Individual Well Simulated Data



28/39

Discrepancy (% error) in Remaining ProductionFor the 3 Empirical Methods on Simulated Wells


Discrepancy (Error %) in remaining production


29/39

How Well Do Different Models Forecast WithShort Term Data ?

A Barnett Shale simulation matching 12 months of history



30/39

How Well Do Different Empirical ModelsForecast With Long Term Data ?

A Barnett Shale simulation matching 36 months of history



31/39

Field Grouped Data Sets



32/39

0.1

1

10

0 50 100 150 200 250 300 350 400

R a t e ,

B S C F / D a y

Time, Months

Historical

Modified Duong (Dswitch@5%)

SEPD

Arps (Dmin@5%)

EOH EOP

How Well Do Different Models Forecast ForShort Term Grouped Data ?

Johnson County (130 wells)- 18 months matched



33/39

How Well Do Different Models Forecast ForLong Term Grouped Data ?

Denton County (81 wells) – 36 months matched

0.1

1

10

0 50 100 150 200 250 300 350 400

R a t e , B S C F / M o n t h

Time, Months

Historical

SEPD

Modified Duong (Dswitch@5%)

Arps (Dmin@5%)

EOH EOP



34/39

What About Oil Wells?

• Same as Gas Wells

– Dswitch/Dmin values vary for different plays

– Interference

• Account for solution gas

• Operational issues need to be accounted for – Pump Issues, Paraffin Issues

– Higher reserves potential if issues fixed



35/39

Conclusions

• Previously mentioned modifications to the Duong makes the Duong

model even more robust and accountable for fracture interference

• The Modified Duong (Dswitch) method provides more accurate results

than the SEDM and Modified Arps (Dmin) Model when more than 12

months of historical production data is available, although some error

is still associated with those forecasts

• None of the models studied produces accurate forecasts with 6

months or less of historical production data

• For grouped well sets the SEPD and Modified Duong (Dswitch) work

exceptionally well providing reasonable forecasts



36/39

Acknowledgements

Crisman Institute at Texas A&M for their funding

Paper # • Paper Title • Presenter Name

http://www.pe.tamu.edu/


37/39

Quiz

• With greater than 12 months of historical production datawhich of these decline models provided the lowest error

in remaining production for an individual well?

a. SEPD/SEDMb. Modified Duong (Dswitch @ 5%)

c. Arps (Dmin @ 5%)

d. Duong



38/39

Forecasting Production in Shale Gas Reservoirs- ABetter Assessment of ReservesOccidental Petroleum Corporation ■ GCS Reservoir Study Group■ Anadarko Petroleum Convention Center ■ 10th May ■ Krunal Joshi, Reservoir Engineer



39/39

Date post:	06-Jul-2018
Category:	Documents
Upload:	ismail-mohammed
View:	214 times
Download:	0 times

SHGAs Forecasting.pdf

Documents