Errors in History Matching

Z. Tavassoli, Jonathan N. Carter, and Peter R. King, Imperial College, London Summary The usual procedure in history matching is to adopt a Bayesian approach with an objective function that is assumed to have a single simple minimum at the correct model. In this paper, we use a simple cross-sectional model of a reservoir to show that this may not be the case. The model has three unknown parameters: high and low permeabilities and the throw of a fault. We generate a large number of realizations of the reservoir and choose one of them as a base case. Using the weighted sum of squares for the objective function, we find both the best production- and best parameter-matched models. The results show that a good fit for the production data does not necessarily have a good estimation for the parameters of the reservoir, and therefore it may lead to a bad forecast for the performance of the reservoir. We discuss the idea that the true model (represented here by the base case) is not necessarily the most likely to be obtained using conventional history-matching methods. Introduction Reservoir history matching is a difficult inverse problem arising in the petroleum industry. The aim of history matching is to find a model such that the difference between the performance of the model and the history of a reservoir is minimized. Traditionally, this is done by hand. But the task of varying the parameters of a reservoir description by hand until a satisfactory match is obtained is extremely onerous and time-consuming. Therefore, gradientbased optimization techniques are increasingly adopted by the oil industry for computer-aided history matching. This is because of the great time-saving benefits they can offer over conventional trial-and-error approaches. Starting with an initial reservoir description, these optimization techniques automatically vary reservoir parameters until stopping criteria are achieved and a history match of field performance is obtained. This mechanistic activity is referred to as automatic history matching. In Refs. 1 through 18, there are samples of studies that use the method of gradients for automatic history matching in order to find the best-matched model of a reservoir. Many other methods have been studied and applied to the history-matching problem. Examples include optimal control theory,1922 stochastic modeling techniques,2326 sensitivity analysis techniques, 2729 and gradual deformation method.31 In Refs. 31 through 45, some other studies on history matching are listed. The use of 3D streamline paths to assist in history matching has become increasingly common, because of their speed compared to that of the conventional methods. Samples of such studies are given in Refs. 46 through 52. These techniques are designated as assisted history matching to distinguish them from automated history-matching techniques. In most of the studies above, independent of the method used for the history matching, there is usually an assumption that there exists a simple unique solution at the correct model. They therefore neglect the inherent nonuniqueness of the solution of the underlying inverse problem. This, consequently, leads to the assumption that a good history-matched model is a good representation of the reservoir and therefore gives a good forecast. In this paper, we want to challenge these assumptions and show that the problem may have multiple solutions, and subsequently, a good history-matched model might have geological properties quite far from those of the truth and therefore could lead to a bad forecast. There have been some studies that attempt to tackle the problem of nonuniqueness of the solution, for example Refs. 1, 6, and 53 through 55. Ref. 53 presents a field story in which the risk arising from such nonuniqueness is highlighted. In Ref. 6, an approach is proposed that couples a chaotic sampling of parameter space with a local minimization technique. Through the evolution of a nonlinear dynamical system, several points are successively used as initial guesses for a local gradient-based optimizer. This provides a series of alternative matched models with different production forecasts that improve the understanding of the possible reservoir behaviors. In Ref. 1, the application of a global optimization algorithm called the tunneling method is presented for two test cases in which a series of minima was found. In Ref. 54, there is a study on the parameter estimation in the oil industry carried out by applying a modified genetic algorithm on a simple reservoir model. It demonstrated that it is difficult to obtain a history match using simple optimization methods. The simple model used in Ref. 54 was a 2D cross section of a reservoir with a sequence of alternating good- and poor-quality sands and a simple fault. The three unknown parameters were the high and low permeabilities of the sands, and the throw of the fault. In a later work by J.N. Carter,55 there is an attempt to capture the effects of modeling errors in inverse problems using Bayesian statistics. He then used a similar cross-sectional model of a reservoir to test the results. It appeared that the proposed error model yields a multimodal objective function that leads to multiple acceptable solutions. The approach used in Carter55 was to generate a large number of models, which could be searched to find the best match according to the criteria chosen. In this paper, we use the same cross-sectional model of a reservoir as in Refs. 54 and 55, and we also use an approach similar to the one in Carter55 to generate a large number of realizations of the reservoir. We then choose one of these models as a base case, and use the sum of squares as the measure of fit between the performance of the base case and each realization. We also find the distance between the two models on the parameter space. By obtaining both the best production- and the best parametermatched models, we then study the history matching and also the prediction of the both models. The results show that a model with a good history matching could have parameters quite different from those of the true case, and it may give a bad forecast. To show that the results are generic, a general study of the forecast then is carried out that confirms the robustness of the results. For the rest of the paper, firstly in the next section, the reservoir model is described and objective functions are defined. Then, a base case is chosen in the best-models section, and the best production- and parameter-matched models are obtained. Then, in the following section, the prediction of the best models is assessed, and then a generalized discussion of the forecast is given, followed by conclusions. Reservoir Model and Objective Function We use a simple 2D cross-sectional model of a layered reservoir, as shown in Fig. 1.57 The model is 2D, with no-flow boundary conditions. Water is injected on the left side, and a production well is located on the right. There is a single, normal fault at the midpoint between the two wells. The reservoir contains six alternating layers of good- and poor-quality sands with high and low permeabilities, respectively. The porosities of the good- and poor-quality sands are 0.30 and 0.15, respectively. We also assume that the three good-quality layers have identical properties, and the three poor-quality layers have a different set of identical properties. The thickness of the layers has an arithmetic progression, with the top layer having a thickness of 12.5 ft, the bottom layer a thickness of 7.5 ft, and with a total thickness of 60 ft. The width of the model
September 2004 SPE Journal

Copyright 2004 Society of Petroleum Engineers Original SPE manuscript received for review 6 August 2003. Revised manuscript received 21 April 2004. Paper (SPE 86883) peer approved 28 May 2004.

352

Fig. 1Reservoir model with fault throw of h. Blue (black) and red (grey) represent poor- and good-quality sands with low and high permeabilities, kp and kg, respectively.

is 1,000 ft. The viscosity used for water was 0.31 cp; for oil, between 0.51 and 0.74 cp within the pressures of 4,041.7 and 9,014.7 psi. Power-law forms are used for the water and oil relative permeabilities as functions of the water and oil saturations, respectively. To produce a set of measurements, we therefore need three properties: the size of the fault throw h, the permeability for the good-quality sand kg, and the permeability of the poor-quality sand kp. The three parameters are selected independently from uniform distributions with ranges: fault throw, h [0,60], good-quality permeability, kg [100,200], and poor-quality permeability, kp [0,50]. The simulation model is 10012 gridblocks, with each geological layer divided into two simulation layers of equal thickness; each gridblock is 10 ft wide. The simulation is carried out with wells operating at fixed bottomhole pressures, and both are completed on all layers. The oil/water production/injection rates are then periodically recorded 43 times: first monthly for 36 months (3 years), and then annually for 7 years. As a goodness-of-fit measure, we use an objective function to be a weighted sum of squares of the difference between the production data of the base case and each realization from the start of production until the end of the history matching:
m

= 0.03Qwb h0,kp0,kg0,ti , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (2) = 0.03Qob h0,kp0,kg0,ti . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (3)

We normalize m by dividing it by the number of the measurements, n. We also define a measure of distance between the base case and each realization on the parameter space as p,
p=

Pi Pbi 2 1 , . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . (4) 2 3 i=1 2 pi

where Pi and Pbi correspond to the three parameters of each realization and the base case, respectively. These parameters are h, kp, and kg for each realization and h0, kp0, and kg0 for the base case. p, therefore, is a 1D measure of distance in a 3D parameter space. pi the standard deviation of the base-case parameter, which here represents a weighting factor proportional to the measuring parameter, pi 0.03Pbi. We normalize p by dividing it by the number of parameters, three. Best Models We choose a point in the parameter space to be the base case; h0 10.4 ft, kg0 131.6 md, and kp0 1.3 md. The production rates obtained are shown in Fig. 2. We then randomly sample another 159,645 realizations of the reservoir from the parameter space and run the simulator to obtain the production data for each sample model. Obtaining objective function m gives a measure of fit between the production data of the base case and each realization at the history-matching period, which we assume to be 3 years, n 36 months. Also calculating p from Eq. 4 gives a measure of distance between the base case and each realization on the parameter space. Best Production-Matched Model. Here, we first look for one single model which has the best match to the production data and call it the best history-matched model. This is a representation of common exercise that is usually done in the industry. The model for which m is minimum is the one with the best production match. In our case, the parameters for such a model are h 33.1 ft, kg 135.9 md, and kp 2.62 md, which gives m 0.118. In
353

h,kp,kg|h0,kp0,kg0
n

1 n i=1

Qw h,kp,kg,ti Qwb h0,kp0,kg0,ti 2

2 w

ti
2

Qo h,kp,kg,ti Qob h0,kp0,kg0,ti 2

2 o

ti

. . . . . . . . . . . . . . . . . (1)

Here, index m stands for the match of the history. Qw and Qwb the water production rates for each realization and the base case, respectively, at each measurement time ti. Similarly, Qo and Qob the oil production rates for the same models. The subscript b corresponds to the base case, and n the number of the measurements during the history-matching period. o and o the standard deviations of the water and oil production rates, respectively, of the base case. In this paper, we consider the standard deviations as weighting factors for normalization purposes. We assume the weighting factor to be 3% of the measuring value (either water or oil production), which gives
September 2004 SPE Journal

Fig. 2Oil/water production rates vs. time for the base case.

Fig. 3, we plot the oil/water production rates for both the base case and the best production-matched model for 4 years from the start of the production. The vertical dotted line corresponds to the end of the history matching (3 years). We see that there is a very good fit between the production profiles of the two models during the history-matching period. For these two models, we also calculate p for a measure of fit between their parameters, which gives a very large value, p 1,066. This value indicates there is a bad match between the parameters of the two models. For the best model, kg is quite well estimated, h is almost three times the throw

of the base case, and kp is almost twice as large as in the base case. These suggest that a model with a good fit for the production data of a reservoir does not necessarily have a good estimation for the parameters of the reservoir. Best Parameter-Matched Model. Alternatively, we may look for the model with the best parameter match. The realization with the best parameter fit to the base case is the one with the minimum value of p. We obtain such a model with parameters of h 11.15 ft, kg 134.8 md, and kp 1.35 md, which gives p 1.23. For this

Fig. 3Oil/water production rates vs. time for the base case (solid line) and the best production-matched model (dashed line) for 4 years from the start of production. The vertical dotted line shows the end of the history matching (3 years). 354 September 2004 SPE Journal

model, however, m 68.18, which is large, indicating that there is not a good fit between the production data of this model and that of the base case. These suggest that a model with parameters close to the true model (represented here by base case) does not necessarily give a good fit for the production data of the reservoir. To establish this, we plot graphs of oil/water production rates vs. time in Fig. 4 for both the base case and the best parameter-matched model for 4 years from the beginning of the production. The first 3 years, which are indicated by the vertical dotted line on the graph, correspond to the history-matching period. We see that there is not a close match between the production data of the two models during this period. There is also an obvious difference in the time of water breakthrough between the two models, such that the assumed model predicts an earlier water breakthrough time. Prediction of the Best Models. Here, we are interested to see how the predictions of both the best production- and parametermatched models compare with those of the base case. In Figs. 5 through 7, we plot 2D saturation maps of oil/water for all the three models at the end of the history matching (after 3 years). We see the differences in oil/water saturation in the layers of the reservoir between the base case and the best production-matched model. This is in spite of the fact that the two models have a very similar production profile at the history-matching period, as is seen in Fig. 3. The distribution of oil/water in the reservoir, however, seems to be more similar for the base case and the best parameter-matched model, although they do not show a good fit for the production profile during the history matching, as can be seen in Fig. 4. To study the forecast of the best models, we consider two quantities: the production data at the prediction period, and the variation of the water saturation at the production well at the end of the forecast. The reservoir model we use here is very simple and small and has a short lifetime. In the model at the end of the history matching (after 3 years) almost 2/3 of the oil in the reservoir has been recovered. The average lifetime of such a reservoir therefore is approximately 4 to 5 years. Therefore, to detect the discrepancies between the models, we set the end of the forecast to be 4 years, and therefore consider one year after the end of the history matching. In Figs. 3 and 4, there are production data for the best production- and parameter-matched models, respectively, over the

whole history-matching and prediction periods (total of 4 years). Figs. 8 through 10 show graphs of water saturation along the production well at the end of the forecast. In Fig. 3, we see that although the production rates of the base case and the best model match each other very well until the end of the history matching, they obviously diverge during the prediction period. If we then look at the depth-saturation plots of the two models in Figs. 8 and 9 at the end of 4 years, we see major discrepancies between the two profiles of water saturation at the production well. These suggest that a model with a good history matching to the true case does not necessarily have a good forecast, and therefore might not be a good representative of the truth. In order to evaluate the forecast of the best parameter-matched model compared to the base case, we look at the production data of the two models at both the history-matching and prediction periods in Fig. 4. They seem to suggest that although the two production profiles do not have a good match during the history matching, they converge toward each other at the prediction period. They also seem to have quite similar profiles of depth vs. saturation at the end of 4 years, as one can see in Figs. 8 and 10. These suggest that a model with parameters close to those of the base case might lead to a good forecast for the true case. Generalization of the Forecast In the previous section, we considered the forecast of the two best models and assumed that the end of the prediction is 4 years. Here, we will see how generalized these results are, and to do so we look at all the realizations of the reservoir model. We therefore introduce a new quantity called f, which shows a measure of fit between the production data of the base case and each realization from the end of the history matching until the end of the forecast. Index f stands for the forecast. f has the same form as m in Eq. 1, with different limits for the summation: a lower limit related to the end of the history matching, and an upper limit related to the end of the prediction. Also, the divisor n for f corresponds to the number of the measurements from the end of the history matching until the end of the forecast. We then obtain f for all pairs of the base case and each realization for all 159,645 realizations with randomly chosen parameters. In order to study the possible correlation between the goodness of the forecast and that of the history matching between the base case and each model, we plot f vs. m.

Fig. 4Oil/water production rates vs. time for the base case (solid line) and the best parameter-matched model (dashed line) for 4 years from the start of production. The vertical dotted line shows the end of the history matching (3 years). September 2004 SPE Journal 355

Fig. 5Saturation map for the base case at the end of the history matching.

Fig. 6Saturation map of the best production-matched model at the end of history matching.

Fig. 7Saturation map for the best parameter-matched model at the end of history matching.

Fig. 8Depth vs. saturation at the production well for the true model 1 year after the end of history matching. 356 September 2004 SPE Journal

Fig. 9Depth vs. saturation along the production well for the best production-matched model 1 year after the end of history matching.

Also, to see any possible relation between the goodness of the forecast for every model with the closeness of the parameters of that model to the base case, we plot f vs. p. These graphs are plotted in Figs. 11 and 12. There seems to be a broad correlation between the forecast and both the history of production and the parameters of the models. This is such that very large values of m , which indicate very bad history-matched models, correspond to a very bad forecast with high values of f. The same follows for the relationship between f and p. However, we are interested in studying the forecasts of the realizations with productions or parameters relatively close to those of the true case. We therefore consider models in Figs. 11 and 12 with small values of m and p. Figs. 13 and 14 represent these graphs for when m<5 and p<10, respectively. The points at the origin correspond to the base case with m 0. The best production- and parameterp f matched models are indicated by arrows with, respectively, lowest m and p in Figs. 13 and 14. The corresponding f for these models are 32.7 and 7.87, respectively, for the former and the latter models. These show a better prediction for the best parametermatched model, and a worse forecast for the best history-matched model, which are similar to the results we obtained in the section on best-model prediction. However, the plots in Figs. 13 and 14 in general do not seem to indicate any systematic correlation between the goodness of the forecast f and that of the history match, m, or parameter match, p. Well history-matched models (with small m) or models with parameters close to those of the base case (with small p) both might lead to bad forecasts with relatively large values of f. We also see that a good forecast with small f might correspond to small or large values of m or p. This seems to suggest that models with bad history matching or those with parameters far away from the true case still might result in relatively good prediction of the production in a reservoir. The other point in Fig. 13 is the presence of two distinct clusters: one with models of good prediction (low values of f) and the other with models of bad prediction (high values of f). Here, we

consider the two distinct sets of models with m<1, which have relatively good history match: those with good prediction, f<20, and those with poor forecast, f>20. For each set, we fit a Gaussian distribution to the likelihood of m over the parameter space using the least-squares method. As the prior probability is uniform, the likelihoods represent the posterior probabilities for the two modes. We then find the values for the mean and standard deviation of each parameter for both Gaussian distributions. The results are as follows. For the cluster with m<1 and f<20, the measured means are 1 5.20, 2 127.0, and 3 2.74, and the standard deviations are 1 0.30, 2 0.37, and 3 0.52. For the models with 39.4, m<1 and f 20, the fitted Gaussian distribution gives 1 136.7, and 3 2.53, and the standard deviations are 2 3.04, 2 1.88, and 3 0.004. In these values, the indices 1, 1 2, and 3 correspond to the parameters h, kg, and kp, respectively. Conclusions We studied a simple 2D cross-sectional model of a reservoir with alternating layers of good- and poor-quality sands with high and low permeabilities, and a simple fault in the middle. The model then had three unknown parameters: the high and low permeabilities, kg and kp, respectively, and the throw of the fault, h. We ran a simulator for several realizations of the reservoir and obtained the production data. Then, we chose a model as a base case and assumed the end of the history matching to be shortly after water breakthrough in the base case (3 years). As a goodness-of-time measure we used a weighted sum of squares of the difference between the production of each realization and the base case until the end of the history matching, m. We also defined a measure of distance on the parameter space between the base case and each realization, p. We then found both the best production-matched and the best parameter-matched models. The results showed that the best production-matched model does not necessarily have a good fit for the parameters of the reservoir. On the other hand, a

Fig. 10Depth vs. saturation at the production well for the best parameter-matched model 1 year after the end of history matching. September 2004 SPE Journal 357

Fig. 11

vs.

for all the realizations of the reservoir.

model with its parameters close to those of the base case might not have a good match to the production data. We then looked at the forecast of the reservoir for both best models one year after the end of the history matching. The quantities we considered were the production data and the watersaturation profile at the production well. The production data of the best history-matched model and of the base case seem to diverge during the forecast period. The two models show very different saturation profiles at the end of the forecast. The production data of the best parameter-matched model, however, seem to converge toward those of the base case in the prediction period, which gives

a saturation profile very similar to that of the base case at the end of the forecast. This suggests that a good history-matched model with productions close to those of the true case does not necessarily have a good forecast, and therefore might not be a good representative of the truth. However, it seems that the model with parameters close to those of the base case might lead to a better forecast for the true case. To show the robust nature of the results above, we then studied the forecast on a more general basis. We therefore introduced a goodness-of-fit measure, f, between the production data of the base case and each realization from the end of the history matching until the end of the forecast. By plot-

Fig. 12 358

vs.

for all the realizations of the reservoir. September 2004 SPE Journal

Fig. 13

vs.

for when

< 5.

ting f vs. m and also p, we did not observe any general systematic correlation between the goodness of the forecast with either production match or parameter match. Well-history-matched models or models with parameters close to those of the base case could both lead to relatively bad forecasts. Alternatively, models with bad history matching or those with parameters far away from the true case might still result in a relatively good forecast for the reservoir. To show that the results obtained in this paper are more generic, we have carried out a further study in Ref. 56, which investigates the errors both in history matching and in prediction by systematically varying each parameter in the reservoir model.

We would also like to mention that the problem we looked at in this paper is well-posed. In a typical history-matching exercise, the number of data values available are typically less than the number of unknown parameters in the reservoir, so it is an ill-posed problem and leads to multiple solutions. However, in our case, there are only three possible variables with 108 measurement data with no modeling errors, but still we run into errors in history matching. Therefore, the results obtained here are not caused by any illposedness of the problem. Consequently, using more data might change the so-called best history-matched model, but the principles of the occurrence of errors will remain unchanged.

Fig. 14 September 2004 SPE Journal

vs.

for when

< 10. 359

In summary, all the results seem to suggest that in using the conventional history-matching methods, one cannot practically guarantee to recover the true model, which represents the real geological structure of the reservoir. Nomenclature h fault throw, ft permeability of good-quality sand, md kg kp permeability of poor-quality sand, md h0 fault throw of the base case, ft kg0 permeability of good-quality sand for the base case, md kp0 permeability of poor-quality sand for the base case, md Pbi parameters of the base case Pi parameters of the reservoir Qo oil production rate, B/D Qob oil production rate of the base case, B/D Qw water production rate, B/D Qwb water production rate of the base case, B/D ti time at each measurement, days measure of t at the forecast period f objective function at the history-matching period m measure of distance on the parameter space p standard deviation of the oil production rate, B/D o standard deviation of the parameters pi standard deviation of the water production rate, B/D w References
1. Gomez, S., Gosselin, O., and Barker, J.W.: Gradient-Based History Matching With a Global Optimization Method, SPEJ (June 2001) 200. 2. Bissell, R.C. et al.: Combining Geostatistical Modeling With Gradient Information for History Matching: The Pilot-Point Method, paper SPE 38730 presented at the 1997 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 58 October. 3. Rahon, D., Blanc, G., and Guerillot, D.: Gradients Method Constrained by Geological Bodies for History Matching, paper SPE 36568 presented at the 1996 SPE Annual Technical Conference and Exhibition, Denver, 69 October. 4. Bissell, R.C., Sharma, Y., and Killough, J.E.: History Matching Using the Method of Gradients: Two Case Studies, paper SPE 28590 presented at the 1994 SPE Annual Technical Conference and Exhibition, New Orleans, 2528 September. 5. Bissell, R., Killough, J.E., and Sharma, Y.: Reservoir History Matching Using the Method of Gradients on a Workstation, paper SPE 24265 presented at the 1992 SPE European Petroleum Computer Conference, Stavanger, 2527 May. 6. Mantica, S., Cominelli, A., and Mantica, G.: Combining Global and Local Optimization Techniques for Automatic History Matching Production and Seismic Data, SPEJ (June 2002) 123. 7. Zhang, F., Reynolds, A.C., and Oliver, D.S.: Model Errors Inherent in Conditioning a Stochastic Channel to Pressure Data, paper SPE 62987 presented at the 2000 SPE Annual Technical Conference and Exhibition, Dallas, 14 October. 8. Ory, J. et al.: A Semi-Automatic History-Matching Technique Applied to Aquifer Gas Storages, paper SPE 38862 presented at the 1997 SPE Annual Technical Conference and Exhibition, San Antonio, Texas, 58 October. 9. Savioli, G.B. and Bidner, M.S.: Comparison of Optimization Techniques for Automatic History Matching, Journal of Petroleum Science and Engineering (1994) 25. 10. Sultan, A.J., Ouenes, A., and Weiss, W.W.: Automatic History Matching for an Integrated Reservoir Description and Improving Oil Recovery, paper SPE 27712 presented at the 1994 SPE Permian Basin Oil and Gas Recovery Conference, Midland, Texas, 1618 March. 11. Ouenes, A. et al.: New Algorithm for Automatic History Matching: Application of Simulated Annealing Method (SAM) to Reservoir Inverse Modeling, paper SPE 26297 available from SPE, Richardson, Texas (1993). 360

12. Smith, R.A.W. and Tan, T.B.: Reservoir Characterization of a Fractured Reservoir Using Automatic History Matching, paper SPE 25251 presented at the 1993 SPE Symposium on Reservoir Simulation, New Orleans, 28 February3 March. 13. Carlson, F.M. and Stein, M.H.: Automatic Waterflood History Matching Using Dimensionless Performance Curves, paper SPE 24897 presented at the 1992 SPE Annual Technical Conference and Exhibition, Washington, DC, 47 October. 14. Yang, P.-H. and Watson, A.T.: Automatic History Matching With Variable-Metric Methods, SPERE (August 1988) 995. 15. Dye, L.W., Horne, R.N., and Aziz, K.: New Method for Automated History Matching of Reservoir Simulators, paper SPE 15137 presented at the 1986 SPE California Regional Meeting, Oakland, California, 24 April. 16. Moore, L.T. and Lamba, J.: Automatic History Matching In Production Modeling, paper presented at the 1985 Annual Conference and Convention of the American Inst. of Industrial Engineers, Boston, Massachusetts,1417 May. 17. Chen, W.H. et al.: A New Algorithm for Automatic History Matching, SPEJ (December 1974) 593; Trans., AIME, 257. 18. Thomas, L.K., Hellums, L.J., and Reheis, G.M.: Nonlinear Automatic History-Matching Technique for Reservoir Simulation Models, SPEJ (December 1972) 508; Trans., AIME, 253. 19. Zhang, J., Dupuy, A., and Bissell, R.: Use of an Optimal-Control Technique for History Matching, paper presented at the 1998 ISIP Intl. Symposium on Inverse Problems in Engineering Mechanics, Nagano City, Japan, 2427 March. 20. Wasserman, M.L., Emanuel, A.S., and Seinfeld, J.H.: Practical Applications of Optimal-Control Theory to History-Matching Multiphase Simulator Models, SPEJ (August 1975) 347; Trans., AIME, 259. 21. Chavent, G., Dupuy, M., and Lemonnier, P.: History Matching by use of Optimal Theory, SPEJ (February 1975) 74; Trans., AIME, 259. 22. Wasserman, M.L. and Emanuel, A.S.: History matching threedimensional models using optimal control theory, JCPT (1976) 15, No. 4, 70. 23. Jimenez, Z., Azpiritxaga, I., and Lozada, T.: Stochastic Modeling Technique as Applied to History Matching of Eocene-Misoa Reservoir, Lake Maracaibo, Venezuela, paper SPE 38020 presented at the 1997 SPE Reservoir Simulation Symposium, Dallas, 811 June. 24. Calatayud, P.M. et al.: Combined Stochastic and Deterministic Reservoir Characterization Leads to Faster History Matching of a Horizontal Well in a Complex Fluvio-Lacustrine Environment, paper SPE 28585 presented at the 1994 SPE European Petroleum Conference, London, 2527 October. 25. Tyler, K.J., Svanes, T., and Omdal, S.: Faster History Matching and Uncertainty in Predicted Production Profiles With Stochastic Modeling, paper SPE 26420 presented at the 1993 SPE Annual Technical Conference and Exhibition, Houston, 36 October. 26. Giudicelli, C.B. et al.: Anguille Marine, a Deep-Sea Fan Reservoir Offshore Gabon: From Stochastic Modeling Toward History Matching, paper SPE 24894 presented at the 1992 SPE Annual Technical Conference and Exhibition, Washington, DC, 47 October. 27. Watson, A.T.: Sensitivity Analysis of Two-Phase Reservoir History Matching, SPERE (August 1989) 319. 28. Dogru, A.H. and Seinfeld, J.H.: Comparison of Sensitivity Coefficient Calculation Methods in Automatic History Matching, SPEJ (December 1981) 551. 29. Hirasaki, G.J.: Sensitivity Coefficients for History Matching Oil Displacement Processes, SPEJ (March 1973) 39. 30. Le Gallo, Y. and Le Ravalec-Dupin, M.: History Matching Geostatistical Reservoir Models With Gradual Deformation Method, paper SPE 62922 presented at the 2000 SPE Annual Technical Conference and Exhibition, Dallas, 14 October. 31. Denney, D. et al.: Geological Parameterization of a Reservoir Model for History Matching, JPT (April 2002) 76. 32. Wu, Z. and Datta-Gupta, A.: Rapid History Matching Using a Generalized Travel-Time Inversion Method, SPEJ (June 2002) 113. 33. Guerillot, D. and Pianelo, L.: Simultaneous Matching of Production Data and Seismic Data for Reducing Uncertainty in Production Forecasts, paper SPE 65131 presented at the 2000 SPE European Petroleum Conference, Paris, 2425 October. September 2004 SPE Journal

34. Spivey, J.P. and Frantz, J.H.J.: History Matching Production Data Using Analytical Solutions for Linearly Varying Bottomhole Pressure, paper SPE 29167 presented at the 1994 SPE Rocky Mountain Regional Meeting/Low-Permeability Reservoir Symposium, Denver, 2022 March. 35. Makhlouf, E.M. et al.: General History Matching Algorithm for Three-Phase, 3D Petroleum Reservoirs, SPE Advanced Technology Series (1993) 1, No. 2, 83. 36. Parish, R.G. et al.: Effective History Matching: The Application of Advanced Software Techniques to the History-Matching Process, paper SPE 25250 presented at the 1993 SPE Symposium on Reservoir Simulation, New Orleans, 28 February3 March. 37. Watson, A.T. et al.: An Analytical Model for History Matching Naturally Fractured Reservoir Production Data, SPERE (August 1990) 384. 38. Watson, A.T., Lane, H.S., and Gatens, J.M. III.: History Matching With Cumulative Production Data, JPT (January 1990) 96. 39. Datta-Gupta, A. et al.: A Type-Curve Approach to Analyzing TwoWell Tracer Tests, SPEFE (March 1995) 40; Trans., AIME, 299. 40. Ghoniem, S.A. et al.: Simplified method for petroleum reservoir history matching, Applied Mathematical Modelling (1984) 8, No. 4, 282. 41. Joyner, H.D. and Lovingfoss, W.J.: Use of a Computer Model in Matching History and Predicting Performance of Low-Permeability Gas Wells, JPT (December 1971) 1415. 42. Sayyouh, M.H.: Goal Programming: A New Tool for Optimization in Petroleum Reservoir History Matching, Applied Mathematical Modelling (1981) 5, No. 4, 223. 43. Watson, A.T. et al.: History Matching in Two-Phase Petroleum Reservoirs, SPEJ (December 1980) 521. 44. van den Bosch, B. and Seinfeld, J.H.: History Matching in Two-Phase Petroleum Reservoirs: Incompressible Flow, SPEJ (September 1977) 398. 45. Shah, S., Gavalas, G.R., and Seinfeld, J.H.: Error Analysis in History Matching: The Optimum Level of Parameterization, SPEJ (June 1978) 219. 46. Illiassov, P.A. and Datta-Gupta, A.: Field-Scale Characterization of Permeability and Saturation Distribution Using Partitioning Tracer Tests: The Ranger Field, Texas, SPEJ (December 2002) 409. 47. Milliken, W.J., Emanuel, A.S., and Chakravarty, A.: Applications of 3D Streamline Simulation To Assist History Matching, SPEREE (December 2001) 502. 48. Caers, J. et al.: A Geostatistical Approach to Streamline-Based History Matching, SPEJ (December 2000) 250. 49. Baker, R.: Streamline Technology: Reservoir History Matching and ForecastingIts Success, Limitations, and Future, JCPT (2001) 40, No. 4, 23. 50. Wang, Y. and Kovscek, A.R.: Streamline Approach for History Matching Production Data, SPEJ (December 2000) 353.

51. He, Z., Yoon, S., and Datta-Gupta, A.: Streamline-Based Production Data Integration With Gravity and Changing Field Conditions, SPEJ (December 2002) 423. 52. Vasco, D.W., Yoon, S., and Datta-Gupta, A.: Integrating Dynamic Data Into High-Resolution Reservoir Models Using Streamline-Based Analytic Sensitivity Coefficients, SPEJ (December 1999) 389. 53. Yamada, T.: Nonuniqueness of History Matching, paper SPE 59434 presented at the 2000 SPE Asia Pacific Conference on Integrated Modeling for Asset Management, Yokohama, Japan, 2526 April. 54. Bush, M.D. and Carter, J.N.: Application of a modified genetic algorithm to parameter estimation in the petroleum industry, paper presented at the 1996 Conference on Artificial Neural Networks in Engineering, St. Louis, Missouri, 113 November. 55. Carter, J.N.: Using Bayesian Statistics To Capture the Effects of Modelling Errors in Inverse Problems, Mathematical Geology (2004) 36, No. 2, 187. 56. Tavassoli, Z., Carter, J.N., and King, P.R.: An Analysis of HistoryMatching Errors, submitted for publication.

SI Metric Conversion cp 1.0* ft 3.048* psi 6.894757

*Conversion factor is exact.

Factors E03 Pas E01 m E+00 kPa

Zohreh Tavassoli is a postdoctoral research associate at the Dept. of Earth Science and Engineering of Imperial College, London. e-mail: z.tavassoli@imperial.ac.uk. Her research interests include history matching, uncertainties, and multiphase flow in fractured porous media. She holds BSc and MSc degrees in physics from Shahid-Beheshti U. and a PhD degree in statistical physics from Brunel U. Jonathan Carter is a senior lecturer in reservoir engineering at Imperial College, London. e-mail: j.n.carter@imperial.ac.uk. His research interests include history matching, inverse problems, Bayesian statistics, genetic algorithms, and agent-based models of complex adaptive systems. Carter holds a BSc degree in mathematics from Southampton U. and a PhD degree in physics from Warwick U. Peter King is a professor of petroleum at Imperial College, London. e-mail: peter.king@imperial.ac.uk. Previously, he spent 17 years with BP plc. His research interests include uncertainties in history matching, reservoir characterization and modeling, and fast estimation of risk using percolation theory. He is a Royal Academy of Engineering Visiting Professor in the Dept. of Engineering at Cambridge U., and a Visiting Scholar in the Dept. of Physics at Boston U. King holds a PhD degree in theoretical statistical physics from Cambridge U.

September 2004 SPE Journal

361