Shale Gas Reservoir Simulation in Eclipse

Transcription

1 PNG 512- Project Report Shale Gas Reservoir Simulation in Eclipse Submitted By: Priyank Srivastava Thought by: Dr. Turgay Ertekin Spring-2017

2 Model Description From Given Eclipse File Reservoir dimensions Nx (No. of blocks Y Ny (No. of blocks Y Nz (No. of direction) direction) layers)* *For Dual Porosity/Perm. Model it has to be even (so 1 layer will be divided into two parts 1 st layer matrix and 2 nd as fracture) Grid block Dimensions X and Y direction Dimensions are such that Gridblock size is small near well (to capture higher pressure variation) and Grid block size is increased as we go away from well. Total Length of Reservoir in X- Direction: 1148 ft. Total Length of Reservoir in Y- Direction: 1246 ft. Z direction Constant height of 100 ft. PVT Properties and Initialization Fluid Phases Present - Gas Density (lb/ft3) Oil Water Gas Rock Compressibility: Psia; For Dual porosity and Dual Perm. Rock Comp. is ZERO. Initial Conditions: Pressure ft. Well location and Constraints Well name: W1; Located in (Gridblock (20, 1)) Well constraints by BHP = 500 psia No relative Perm. functions are given so we assume single phase 100% gas saturation. We model the quarter of field to utilize advantage of symmetry of field that s why all blocks in X and Y direction containing well are multiplied by 0.5 since the gridblocks. Compressibility of rock in case-1 is taken as 1e-06 1/psia. While In case 2 & 3 compressibility of rock is taken as 0.

3 Description of Main Mechanism for Shale gas Production and Modelling Techniques Unconventional Shale Gas Reservoirs which have source rock same as reservoir rock that produces from very tight shale formation whose permeability is order of nanodarcies. Although there is still a wide amount of skepticism in industry about real mechanism of production of shale gas reservoirs and factors which effect life of these unconventional reservoir. Shale Formations are considered so tight that they can be commercially exploited only when extensive networks of natural fractures exist, many of these natural fractures are filled with calcite and other minerals so by intersection of natural fractures with hydraulic fractures, it is possible to create extensive networks of fractures thus creating large surface area for production of gas. The Most critical part of modeling process of shale gas flow in reservoir is the ability to describe flow from a very tight rock matrix to network of fractures. Following techniques can be used for modelling matrix and fracture network combinations 1.) Discrete fracture network (DFN) model- In which fractures are defined in a multimillion cells reservoir model by their discrete location in reservoir (microsiesmic can be used to pick their locations) and fracture network connectivity is defined from production characteristics/conductivity or rock mechanics analysis can be used if estimation of insitu stress is easy. However, being a very data intensive process which is seldom available in real life. Nevertheless it can provide important insight into physics of flow between fracture and matrix for simple configurations. 2.) Dual Porosity Warren and Root (1963, From MPGE, OU) first proposed the very simple model of fluid flow from naturally fractured reservoir. The schematic is given below, this model assumes each matrix block to be in connection with a fractures. The advantage of this model is detailed geometry of fracture need not be specified in this model. It is really simple to apply this simple model to Shale gas reservoir with matrix having low permeability but high gas storage capacity while fractures will have low gas storage capacity with high permeability. The flow form matrix to fracture is described by a connection factor (sigmav in case of Eclipse) and a Darcy flow characteristics while original Warren-Root model assumes no transient flow between matrix and fracture blocks so matrix blocks are treated in pseudo steady state with single value of pressure within the blocks; the mass transfer rate from the matrix to fracture depends on pressure differential between the matrix and fracture. Also fracture spacing can play an important role in this model. Kazemi extended warren and root model for transient flow between matrix and fracture while Kuchuk and sawyer extended the model

4 further to incorporate desorption effect and Knudsen flow in the pores. Recent advances in coupling of partial differential equations for the pore and fracture fluid pressures using finite difference methods and Laplace transform can be used to obtain Semi-analytical solutions for fluid flow in shale as reservoirs. 3.) Dual Permeability Model This model is similar to Dual Porosity model except that there is an interaction between matrix-matrix blocks so if pressure gradient exist between adjacent matrix blocks then gas can flow from matrix- matrix connection.

5 Case 1: Single porosity model with only hydraulic fracture In this case gas is produced according to Darcy s law from fractures which may be natural or artificial, sometimes artificial hydraulic fracturing treatment is necessary for connecting unconnected network of natural fractures to make commercial quantities of gas available on surface. Figure-1 shows the schematic of this model with reservoir properties as given to us. Notice the presence of high perm. Hydraulic Fracture in 1 st row in X direction. X Direction 20 Grid Blocks K = X 0.5 md Φ = X0.5 Y Direction 42 Grid Blocks K = 0.1 md Φ = 0.2 K = 0.1 X 0.5 md Φ = 0.2 X 0.5 Figure 1 : Schematic of Single porosity model with hydraulic Fracture Following results can be obtained by running given Eclipse file Initially GIIP (MMSCF) Table 1 : Recovery Factor Table for Case-1 GIP After 4 yrs. (MMSCF) Recovery factor % Reservoir Pressure Depletion Behavior Notice that rate of pressure decline is low although reservoir appears to reach PSS at time approx. after 880 Days (In PSS the Cartesian plot of Reservoir pressure Vs. time will become straight line). Low Recovery Factor of 52% can be attributed to low permeability of reservoir and no support from natural fracture reservoir.

6 Above Pictures shows the pressure depletion in reservoir 3-D model with time for year 2000 for following dates 01/01; 01/05; 01/17; 01/29; 11/04 in that order. Notice that pressure transient travels very fast in direction of fracture plane while due to low permeability of reservoir it takes 3 months of production to reach Peudosteadystate (PSS), Far East reservoir boundary is hit at time of approx. 03/15/2000. So there is a transient infinite acting flow before reservoir hits outer boundary to produce in Pseudo steady state.

7 Case 2: Dual porosity model with hydraulic fracture Another model considers presence of fractures alongside of matrix blocks on which gas is adsorbed in huge quantities, Thus shale gas reservoirs may be considered as a Dual porosity system with each matrix block in connection with fractures through fick s law of diffusion (Due to very low permeability of matrix there is negligible Darcy flow between matrix and fracture) but there is no communication between two matrix blocks. Fractures flow to wells according to Darcy s law thus gas is provided to well-connected fracture system by desorption phenomenon (This model is yet to be proven as I don t think that Shale Gas reservoir will have big Depletion of Pressure as to have huge amount of desorption occurring as given by Langmuir curve. This Model is adopted from CBM reservoirs).we don t consider Adsorption phenomenon for our Case. Figure-2 Shows Schematic of Dual Porosity model as given to us note presence of hydraulic fracture in 2 nd (fracture layer). In Eclipse Dual porosity is modelled as non-neighboring connection (NNC) between matrix and fracture, so each grid cell of matrix layer (1 st ) is connected to corresponding grid cell of natural fracture layer (2 nd ) note the presence of hyd. Frac. in X-direction in 2nd layer. SigmaV quantifies the connection factor between two cells, Note that there is no connection between matrixmatrix blocks. Figure 2 : Schematic of Dual porosity System Picture on Right depicts reservoir setting. Left Picture shows modeling as done in Eclipse simulator. Following results can be obtained by running given Eclipse file Initially GIIP (MMSCF) Table 2: Recovery Factor Table for Case-2 GIP After 4 yrs. (MMSCF) Recovery factor %

8 Since reservoir is naturally fractured reservoir boundary is hit as soon as we start production. This means due to very high conductivity of fractured system reservoir produces reaches Pseudo steady state within few days of beginning of production. Following Pictures shows the pressure depletion in reservoir 3-D model for Year 2000 on following dates respectively: 01/01; 01/05; 01/17; 01/29; 11/04. Reservoir Pressure on 11/04/2000 is psia. Although it is not shown in pictures but there is a pressure gradient between Layer1 and Layer2 non neighboring connections and flow is occurring from layer1 (matrix) to layer 2(fracture) by Darcy s law. Reservoir Pressure Depletion Behavior Note that Rapid Decline of Reservoir Pressure within first 120 days where it declines from initial reservoir pressure of 5000 psia to 600 psia. This significant decline can be due to production of all of free gas from fracture matrix system. Note that there is coupled behavior between matrix and fracture pressure i.e. due to low porosity of natural fracture system no free gas is available to it while matrix (having high porosity) will have huge amt. of gas.the

9 feeding of fracture system by matrix blocks is dictated by Darcy s law and sigmav (connection factor between matrix and fracture). The picture below shows the Final Pressure condition of Matrix and Fracture Layer. Note the difference of 1 Psia between 1 st layer (matrix) and 2 nd layer (fracture).the Reservoir pressure (FPR) plotted in previous pictures Gas In place/pressure Behavior for 1 st layer (matrix) and 2 nd layer (Fracture) This Graph clearly shows most of gas is present in 1 st layer (matrix) while 2 nd layer (fracture) contains comparably very less amount of gas and reason for this is high porosity of matrix (0.2) compare to that of fracture (0.0073). Thus the most important factor which will determines how much gas we will be producing is coupling between porosity and matrix (SigmaV). This factor depends on Fracture spacing and geometry of matrix blocks. This Graph shows that while pressure of matrix and fracture blocks follows the similar trend and magnitude but average pressure in matrix blocks at late times is slightly lower than ( 1 psi.) pressure in fracture blocks. This can be attributed to constraint of 500 psia that we set for well BHP and since wells are only connected with fracture blocks pressure in fracture layer is 500 psia while fracture in matrix layer is 499 psia.

10 Case 3: Dual permeability model with hydraulic fracture Another Model to Describe Shale Gas reservoirs is Dual Permeability Model where matrix blocks can communicate to each other. The flow between Matrix to matrix will be controlled by diffusion (depending on gas concentration gradient) and Darcy s law (depending on matrix permeability) While wells are only Connected to Fractures. Figure-3 Shows schematic of Dual permeability model as modelled in eclipse Note That Dual Porosity model can be considered as a special case of Dual Permeability Model in which matrix to matrix connection is zero. LAYER 1_MATRIX BLOCKS X Direction 20 Grid Blocks Matrix-Matrix Connection due to Dual Perm. K = 0.1 X 0.5 md Φ = 0.2 X 0.5 Y Direction 42 Grid Blocks K = 0.1 md Φ = 0.2 LAYER 2_FRACTURE BLOCKS X Direction 20 Grid Blocks Non Neighbouring Connection Between two layers SigmaV = K = 0.1 X 0.5 md Φ = 0.2 X 0.5 Hyd. Frac. K = md Φ = Y Direction 42 Grid Blocks Wells Connected only to Frac. Blocks Natural Frac. K = 100 md Φ = Figure 3 : Schematic of Dual Perm.-Dual Poro. Model with Hyd. Frac Following results can be obtained by running given Eclipse file Initially GIIP (MMSCF) Table 3 : Recovery Factor Table for Case-3 GIP After 4 yrs. (MMSCF) Recovery factor % In conclusion, No Matter which model we choose Conductivity and Effectiveness of fractures both artificial and natural will play a significant role in production prediction of shale gas wells. The Work done in this report tries to study appropriateness of above given mechanism of shale gas production and considers role of fracture conductivity on shale gas production profile. Also, I tried to show effect of including Adsorption model in the given dataset.

11 Reservoir Pressure Depletion for year 2000 with Dates: 01/05; 01/17; 01/29; 11/04.Notice the Coupled behavior of fracture and matrix blocks also there is a matrix-matrix connection between blocks for this model so flow will also occur within first (matrix) layer but due to very low permeability of matrix as compared to fractures it is not easy to observe this flow. This flow may be observed if we increase permeability of matrix or perforate the well in both matrix & fracture system. Another important implication of this model is since wells are only connected with fractured system some amount of gas will remain stored in matrix due to matrix-matrix coupling thus expected ultimate recovery must be lower for case of Dual Perm. Model. Fareast block pressure declines to psia on 11/04/2000 Note that Rapid Decline of Reservoir Pressure within first 120 days where it declines from initial reservoir pressure of 5000 psia to 600 psia. This significant decline can be attributed to high conductivity of fracture system.

12 Gas In place for 1 st layer (matrix) and 2 nd layer (Fracture) This Graph Shows GIP for matrix and fracture layer as expected most of gas is present in matrix (high porosity) while fracture contains very less amount of gas (low porosity).the factor that controls flow from matrix to fracture is coupling factor sigmav. This graph shows behavior of average pressure of matrix and fracture blocks. Well is connected to only fracture block and have BHP constraint of 500 psia. Therefore fracture block pressure gets stabilized once it reaches 500 psia.

13 Comparison of Production Profile for given Base Cases It does not makes sense to compare cumulative Production and Rate for three cases as GIIP for case 1 is different from case 2 & 3.However, since well constraints are same it seems that well present in fractured reservoir will give much Higher Production rate initially. It is not advisable to compare FGPT graph as reservoir potentials (GIIP) is different for different cases. Figure 4: Total Cumulative Production for Three Cases Reasons for same behavior of DUALPOROSITY and DUALPERM model Since there is no Adsorption Model and Φ & K for both cases are same there is No Noticeable difference between the two. However, if we zoom in there is slight difference between Two Cases Which shows Lower gas production in DUALPERM. Case this can be attributed to Gas which is stored in matrix due to matrix-matrix connection and is not contributed to fracture and since only fracture can produce to wells thus this gas is never released from reservoir. Figure 5 : Zoomed View of Total Cumulative Production for Case-2 & 3 Due to presence of natural fracture network (also giving high gas rates in Case 2 & 3. There is rapid decline in Reservoir Pressure as compare to case 1 where we only have hyd. fractures present. Figure 6 : Reservoir Pressure Decline for three Cases

14 Vol. of Gas Adsorbed (Mscf/ft3) Case 4: Dual porosity ADSORBTION model with hydraulic fracture Adsorption Model assumes that gas flow from matrix to fracture occurs using Fick s law of diffusion instead of normal Darcy s flow and as such since permeability of matrix is low, so dominant force for matrix to fracture flow will be Diffusion. There are many parameters which can Control rate of desorption in a shale gas play which includes shape of matrix blocks to matrix drainage time, Fracture Spacing. Following parameters and Langmuir adsorption curve is used for generation of adsorption model. Also in this model since gas is stored in matrix surface (porosity of matrix have to be increase to 0.98) Matrix Diffusion Coeff. : 0.2 ft2/day Langmuir Adsorbtion Curve Pressure, Psia Figure 7 : Barnett 21 Langmuir isotherm Curve taken from Zoback paper According to Salman Akram Mengal Master s Thesis from TAMU OGIP estimation methods when applied on field data from selected wells showed that inclusion of adsorbed gas resulted in approximately 30% increase in OGIP estimates and 17% decrease in recovery factor (RF) estimates. In Eclipse GIIP for Adsorption Dual Ф is calculated using Langmuir parameters for Barnett 21 shale s as given in Zoback paper Langmuir parameters are P l = Psia ; V l = scf/ton. Reservoir Pressure used in given data deck is 5000 psia. So according to given Langmuir curve maximum amt. of gas that can be adsorbed for given reservoir pressure will be mscf/ft3 (Density of shale is assumed to be 180 lb/ft3)

15 Comparison with Case 2 Case Case 2 4 GIIP (MMscf) GIP after 4 yrs. (MMSCF) Recovery % As shown the GIIP and Recovery both are reduced for case when adsorption is taken into account. The reason for decrease of GIIP can be attributed to amount of adsorbed gas and values of Initial Langmuir parameters chosen while in Case-2 complete flow was Darcy s in Case-4 we will have flow from desorption of gas from matrix to fracture. Following curves gives Reservoir Pressure decline for both cases note low cumulative production for Case4, Also this case gives rapid decline of reservoir pressure as compare to case-2 hence it could be deduced that due to adsorption more gas is depleted earlier from the system then with case of no adsorption. Green Curve: Case 2 (No Adsorption) Red Curve: Case 4 (With Adsorption)

16 Gas In place for 1 st layer (matrix) and 2 nd layer (Fracture) Comparison of All the Cases Presence of Natural Fractures Increases Recovery Factor of Reservoir From 52% (Case-1) to 89% (Case- 2&3). However, there is no noticeable difference between Dual porosity and Dual Perm. Model due to low perm. Of matrix system and we have not modelled any gas Adsorption in the system. However, when Adsorption is modelled Recovery factor is reduced. Comparison Criterion Single Ф (Only Hyd. Frac.) Dual Ф Dual Perm. Adsorption Dual Ф GIIP (MMSCF) Recovery % FPR = 500 Not Reached psi in how many days Flow regime Transient Flow PSS PSS PSS after 1 Month of production Drainage Area Low High High High

17 Case 5, 6, 7: Reservoir model With Barnett Shale Properties Figure 8: Key Reservoir Properties of Barnett Shale (Nelson and Hill, 2000) Reservoir parameters Values Depth (ft.) 7500 Net thickness, ft. 100 Porosity Res. Perm. md 0.09 Fracture wing length (ft.) 126 Fracture permeability --unchanged-- Reservoir Pressure (psia) 3500 Figure 9: lower Barnett shale properties taken from Case 5: Single Φ Hyd. Fracture Model for Barnett Shale Case 5: Barnett Single Porosity Initially GIIP (MMSCF) GIP After 4 yrs. (MMSCF) Recovery factor %

18 Above Pictures shows the pressure depletion in reservoir 3-D model with time for year 2000 for following dates 01/01; 01/05; 01/17; 01/29; 11/04 in that order. Notice that pressure transient travels very fast in direction of fracture plane while due to low permeability of reservoir it takes 3 months of production to reach Pseudo steady state (PSS), Far East reservoir boundary is hit at time of approx. 03/15/2000. So there is a transient infinite acting flow before reservoir hits outer boundary to produce in Pseudo steady state. Role of Natural Fracture in Shale Gas Production According To kent bowker of bluescape energy : It generally takes about 2 3 yrs. of experience in the Barnett Shale play of north Texas for the geologist or engineer to realize that open natural fractures are insignificant to the productivity of the Barnett, that is not to say that natural fractures are not abundant in Barnett cores; they are, but they are nearly always healed with cement, commonly calcite Our natural bias as conventionally trained engineers and geologists is to identify a fracturepermeability network that will transmit gas from the matrix of the rock to the wellbore. The usual logic of those new to the Barnett is that induced fractures created during completion operations must enhance the existing open natural-fracture network The following thought experiment helps to

19 illustrate the point (R. Suarez-Rivera, 2005, personal communication). If one drills a hole in a thick block of glass that is under anisotropic lateral strain and then pressurizes the hole with water, the glass will crack along a single plane. However, if we first shatter the glass and then glue the pieces back together perfectly (so that there are no remaining voids) and repeat the procedure, the block of glass will fracture along many planes. The presence of open natural fractures would actually inhibit the growth of induced fractures. When a crack forms in a steel beam or piece of sheet metal, the propagation of the fracture can be halted by drilling a hole at the tip of the fracture. By eliminating the stress point at the tip of the fracture, the stresses causing the propagation of the fracture are dispersed, and the fracture no longer grows. I believe that the combination of diffusion, a very high gas concentration within the Barnett, and the rock s ability to fracture is what makes the play successful. The Barnett is not a fractured-shale play; it is a shale that-can-be-fractured play (D. Miller, 2004, personal communication). 1.) Does Orientation of Fracture matters Fracture length in Case ft. Let s Change Only Orientation of Fracture as given below to compare to it With Case-1 X Direction 20 Grid Blocks K = X 0.5 md Φ = X0.5 Location of Well Y Direction 42 Grid Blocks K = 0.1 md Φ = 0.2 K = 0.1 X 0.5 md Φ = 0.2 X 0.5 Following are the Grid blocks which we can change Grid Cell X Location Grid Cell Y Location X cell length Y cell length Equivalent Diagonal Length (ft) Distance of Block from well (ft)

20 Since, we want to keep Fracture Length same so Changing Permeability/Porosity of blocks till highlighted row in base model will give fracture length of 1109 ft. (slightly lower than base model of 1148 ft.) Red Curve: Case 1 Green Curve: Diagonal Fracture Analysis - Since we are using Cartesian Grid System. So we are forcing fluid to travel in Particular direction (X, Y and Z) So, Hydraulic Fracturing will give maximum recovery when done in direction of grid block. This is a Numerical Artifact due to methodology use in solving this problem numerically rather than being a real situation in Reservoir.

21 Conclusions What Assumptions we have taken in above models (Relevance to Practical Issues) 1.) No Adsorption model is made in data deck so contribution from matrix is only due to gases present in its pores which may not be in huge amount. In real life shale gas wells are excepted to show phenomenon of gas production due to Desorption of gas from matrix, Hence Rate decline will not be so drastic as seen in the production profile. Case-4 Models Adsorption in Dual Porosity Model and compare its performance with original model of Dual Porosity (Case2) with no Adsorption. 2.) Role of natural fracture network this is shown by Comparing Case-1 with Case-2 as Case-2 Contains Natural Fractures also along with Hyd. Frac. While case 1 contains only Hyd. Frac. Presence of Natural Fractures Increases Recovery Factor of Reservoir From 52% (Case-1) to 89% (Case-2&3). 3.) Model appropriateness While Shale Reservoirs are very low permeability they are expected to contain natural fractures which may be closed due to high overburden stress, Thus in order to reactivate these natural fracture we use hydraulic fracturing techniques. Thus Dual Porosity Model with Hydraulic Fracture (Case-2) seems to be the most appropriate model for these kinds of reservoir. We can increase the complexity of problem by including Adsorption and Dual permeability behavior of matrix. Eclipse gives us the ability to model Dual Porosity/Permeability system in many different ways.while Different Shapes of matrix can be chosen we can also discretized single matrix blocks into many different sub-grids thus enhancing our ability to incorporate more heterogeneity in our models. 4.) Comparing Real Barnett shale model to Given Case Case 5,6,7 compares cases 1,2,3 respectively