You are on page 1of 11
SPE/ADC 37644 Optimization of Multiple Bit Runs Soaety a Petooun Emnoers R.V. Barragan, Petrobras, and O.L.A, Santos, Petrobras, and E.E. Maidia, Csiro Petroleum Cony 197 SPEND Og Coreen 2/3 then bil = bi + hj ‘Thus, the choice of point bi+1 is equally probable in any direction from bi within a distance of hj. If the parameter drilling time generated during this random process exceeds the ‘maximum drilling time, this maximum drilling time should be used. Thus, the maximum drilling time is a constraint to the drilling time values generated randomly. (©) Compare f(bi) with f(bi+1). If FIFI) < Abi), repeat the process from (b) using point bit and zero the trails counter. Otherwise, keep the point bi and increment the tails counter by 1. If the trails counter achieves Imax, reduce stepsize hj. (d) Repeat the process from (b), unless the stepsizes forall variables are within prefixed values wien the process finishes. This technique has shown to be very consistent and efficient. To improve its efficiency, a modification on it has been introduced! When f(bi) > bit, store the successful direction and increase the probability to 0.5 in that cirection It is important to realize that even thought the heuristic 581 approach does not warrant the best solution, it finds a good solution even for a small number of trails as can be seen in Figures | and 2. Figure 1 shows the reduction in metric cost when it is allowed to modify weight on bit and rotary speed as formation changes (Method 3 of drilling optimization) against the optimization with constant weight on bit and rotary speed (Method 2). Figure 2 shows the reduction in total cost when the optimization is done for an entire well phase (Method 4) against the optimization for each single bit run (Method 2). In those figures, for each number of trails for stepsize reduction, the heuristic method has been used five times to yield the maximum, the minimum and the average values for ‘the metric cost fora single bit (Fig. 1) and the well phase cost, ig, 2) simulations. It can be seen from the figures that when the number of trails for stepsize reduction increases, the metric, cost and well phase cost tend to converge and the range between the maximum and minimum values reduces. Results Computer programs have been written for each drilling ‘optimization method. They were run for a well phase starting at a depth of 1000 m with well phase lengths varying from 100 to 1000 m. The simulations have been done for both ‘modified drilling models (Bourgoyne & Young and Warren). Two drilling situations have been analyzed: single formation (one lithology) and multiple formations (several lithologies). The results are shown in Figures 3 to 9 and the other input data are shown in the Appendix. ‘Single Formation Figures 3 and 4 for Bourgoyne & Young model and Figure 5 for Warren model show that the simulation results for Methods 1, 2 € 3 are exactly the same since there is no difference between these methods when just one formation is considered. Methods 4 and $ displayed very close results ‘Actually, the difference between these two methods is due to statistical variation inherent to the applied methodology. The optimization by well phase (Methods 4 and 5) has shown an economy of 5 % over the optimization for single bit runs (Methods 1, 2 and 3), Multiple Formations Figures 6 and 7 for Bourgoyne & Young model show that Grilling optimization by Method 3 yielded the greatest drilling length and the lowest metric cost when compared with Methods 1 and 2. For the frst bit run, the optimization by Method 3 brought an increase in drilling length of more than 50% over Method 1 and more than 15% over Method 2. This resulted in savings around 15% over Method 1 and around 10% over Method 2. ‘The optimization by well phase (Method 4 and 5), allowed an average savings of 15% over Method I and a highest SPE 37544 OPTIMIZATION OF MULTIPLE BIT RUNS ‘ saving of 35% for the first bit run (Figure 7). Figures 8 and 9, for Warren model, show similar behaviors. [Note that the results for Methods 4 and 5 are very close again, Since the number of formations is small in this example, there is no big savings when itis allowed to modify the drilling parameters when the formation changes (Methods 3 or § when compared with Methods 2 or 4). During the ‘optimization, bits are normally changed when they reach hard formations as shown in Figures 10 and 11 Conclusions |. Optimization by well phase is more economical than optimization by single bit runs. These savings tend to reduce with the increase of well phase length due to the increase of the number of bits needed to drill the well phase. 2. The heuristic approach is efficient, robust and versatile. It accepts easily constraint values to the drilling parameters making it very suitable to optimization of directional drilling. This technique can also be modified to include a procedure 10 choose the best bit to be used. 3. The optimization methods showed in this paper can be applied during the project and drilling phases of the well. The data acquired during the well drilling can be used to modify the best drilling parameters for the next bits to be run, Acknowledgments ‘The authors thank Petrobras and Csiro Petroleum for allowing. the preparation and presentation of this paper. References | Bourgoyne A.T et Ali, "Applied Drilling Engineering”, SPE Text Book Series, Vol I. 2. Bourgoyne A.T., Young F.S., "A Multiple Regression Approach ‘to Optimal Drilling and Abnormal Pressure Detection”, Society of Petoleum Engineers Journal, Aug. 1974 TM, “Penetration Rate Performance of Roller Cone Bits", Society of Petroleum Engineers N° 13259, 4. Winters W.t., Warren TM., Onyia E.C, “Roiler Bit Model with Rock Ductility and Cone Offset’, Society of Petroleum Engineers N° 16696, 5. Black AD., Tibbitts G.A, Sandtrom JL, DiBona B.G,, “Effects of Size on Three Cone Bit Performance in Laboratory Drilled Shale”, Society of Petroleum Engineers N° 11231 6. Press WH, Veterling W.T., Plannery BP, Teukolsky S.A., “Numerical Recipes in Paseal”, Cambridge. 7. AbBetairi EA, Moussa MM, ALOtibi SS, "Multiple Regression Approach to Optimize Drilling Operations in the Arabian Gulf Area, Society of Petroleum Engineers N° 13694 8. Wee W., Kalogerakis N, “Modelling of Drilling Rate for Canadian Offshore Well Data", Journal of Canadian Petroleum Technology, Nov.-Dec, 1989, 3. 9, Hareland G., Hoberack L-L., “Use of Drilling Parameters to Predict in-situ Stress Bounds”, Society of Petroleum Engineers MADC N° 25727, 10, Bentesen R.G., Wilson D.C, “Optimization Techniques for 582 Weight-on-bit and Rotary Speed, Part I - Point and Interval Optimization, Part I - Mult-interval Optimization", Joural of (Canadian Petroleum, Oct-Dec. 1976 and Jan-Mar. 1977. Appendix Constants forthe Bourgoyne & Young's mode: 11 Global 100 Weight Exponent 10 Weight Constant 00 Rotary Speed Exponent 10 Hydraulics Exponent 0.0 Differential Pressure Exponent 0.0 Bit Wear Exponent 0.08 Constants for the Warren's model: a 25 > 05 ¢ 25.0 Constant forthe Rig: Maximum Depth Maximum Flow Rate Maximum Pump Pressure Tripping Time: 0,003. depth + 1 epprop 10000 m 300 gpm 5000 psi (hours) Connection Time: 0,006 .(tepth pat ~depthniet) USS 500,00/hour (hours) Rig Operating Cost Bit Constants: Type Diameter Nozzles mM 8s” 1o/10/10 USS 1000,00 Cost Fluid Constants: Rheological Model Type Density Formation Data: Single Formation Bingham Clay/Water 9.0 Ibvgal Finat Dept (ny TROT] 30007 Multiple Formations iar Depth mn, in000 TOTO T0500 11500 2200 225007 z000 Form: Press (ibaa) 7 DP a 020. ‘20. 0207 0} 020. 030, 20000 330000 SPE 37644 (OPTIMIZATION OF MULTIPLE BIT RUNS Model s Ri | Rx | Moore eas | 0765 | 0,765 Maurer 749_| 0751 _| 0,748 Bingham | 7.25 | 0760 | 0,758 Cunningran | 433 | o8s7_| 0985 Exe | 430 | 850 | 0,856 cate | 3,08 | 0800 | 0.01 “Table I - Results for Rate of Penetration Analysis Method 1 Method 2 Method 3 Method 4 Method 5 1 Bit; 1 Formation / 2 27 ar 27 27 Bit 1 Bit; 2 Formation / ar ar 243, 27 243 Bit 1 Bit; 3 Formation / a a 2.187 27 2.187 Bit 2Bit; 4 Formation / 5 54 54 729 729 Bit 2Bit; 2 Formation / 54 54 486 729 59.049 Bit 2Bit; 3 Formation / 54 54 4374 729 4.782.969 Bit 3 Bit; 1 Formation / 81 at at 19.683 19.683 Bit 3 Bit; 2 Formation / a at ne 19.683 14,348,907 Bit ‘Bit; 3 Formation / at at 6.561 19.683, 1,048x10"° Bit 4 Bit; 1 Formation / 108 108 108 531.441 531.441 Bit 4 Bit; 2 Formation / 108 108 972 531.441 3.48 7K10 Bit 4.Bit; 3 Formation / 108 108 8748 531.441 2,288x10" Bit 4 Bit; 4 Formation / 108 108 78.732 531.441 2.954x10"" Bit 5 Bit; 3 Formation | 135 135 10.935 14.348.907 5,003x10'° Bit "Table 2 Number of Searches for Hooke & Jeeves Methods 583 SPE 37646 OPTIMIZATION OF MULTIPLE BIT RUNS USS/m #000 Method (Optimum Cost With Constant Weight on Bit and 29804" 2920 | ‘Metric Cost for a Single Bit 28.80 Method 3 DBR] ° “ 0 120 160 200 Number of Trails for Stepsize Reduction Figure 1 - Metric Cost as a Function of the Number of Trials ‘Well Phase Cost ‘Method 4 160 200 0 120 [Number of Trails for Stepsize Reduction Figure 2 - Well Phase Cost as a Function of the Number of Trials 584 SPE 37644 OPTIMIZATION OF MULTIPLE BIT RUNS Single Formation 45.00 Bourgoyne & Young's Model 40.00 g 35.00 4 30.00 + 7 wo| 0.00 200.00, 400,00 600.00 800.00 1000.00 ‘Well Phase Length (M1) Figure 3 - Metric Cost as Function of Well Phase Length (Single Formation = B&Y) Single Formation Bourgoyne & Young's Model 100.00, 96.00 = i | i . 0 | : / | {| — eta 1203 | 1 Method 4 or 5. e000} —, rr oo meucoS apa emco etc tem000 ‘Well Phase Length (M) Figure 4 -Cost Reduction as Function of Well Phase Length (Single Formation- B&Y) 585 SPE 37664 OPTIMIZATION OF MULTIPLE BIT RUNS Single Formation Warren's Model 0.00 200.00 400,00 600.00 800.00 1000.00 ‘Well Phase Length (M) Figure 5 - Metric Cost as Function of Welt Phase Length (Single Formation - Warren) lultiple Formations ao Bourgoyne & Young's Model j 4000 E aw 2000 0.00 200.00 400.00, 600.00 800.00 1000.00 ‘Well Phase Length (M) Figure 6 - Metric Cost as Function of Well Phase Length (Multiple Formations - BRY) 586 SPE 37644 Cost (%) ‘OPTIMIZATION OF MULTIPLE BIT RUNS Multiple Formations ‘1000 Bourgoyne & Young's Model 10000 e000 — etl 79.00 J r Meelis 0.00 200.00 400.00 600.00 200.00 1000.00 ‘Well Phase Length (1) Figure 7 - Cost Reduction as Function of Well Phase Length (Multiple Formations - B&Y) Multiple Formations Warren's Model ‘Metric Cost (USS/M) 8 8 70.00 alan aepeal 0.00 200.00 400,00 600.00 80000 +1000.00 ‘Well Phase Length (M) Figure 8 - Metric Cost as Function of Well Phase Length (Multiple Formations - Warren) 587 SPE 37644 OPTIMIZATION OF MULTIPLE BIT RUNS Multiple Formations Warren's Model 10000 i Ce fi | 9.00 NF eM \ of y s iff = Vf: é 1 8000 1 { 70.00 0.00 200.00 400.00 600.00 200.00 1000.00 ‘Well Phase Length (M) Figure 9 - Cost Reduction as Function of Well Phase Length (Multiple Formations - Warren) Bit Run in Soft Formation 160.00 12000 Second Formation (Hard) First Formation (Sof) Interval Drilled (M) 0.00 40.00 80.00 120.00 160.00 20000 Formation Thickness (M) Figure 10 - Bit Run Starting in a Soft Formation SPE 37506 OPTIMIZATION OF MULTIPLE BIT RUNS Bit Run in Hard Formation 300.00 20.00 + ‘Third Formation (Hard) 150.00 Interval Drilled (M) Second Formation (Soft) First Formation (Hard) 5009 = eet s000 6000 7000 80.00 9000 10000 11000 12000 13000 Formation Thickness (M) Figure 1 - Bit Run Starting in a Hard Formation, 589

You might also like