Method for predicting a well log during well drilling

Abstract

The invention relates to a method for predicting well logs beyond the depth of the drill bit. Starting with at least one training curve of the log acquired in a training well deeper than the drill bit, matched with a curve of the log acquired in the well, a model is built to predict the log in the well using a supervised machine learning method. For each training curve, a composite learning curve is constructed, formed by the portion of the training curve matched with the curve from the well to the drill bit, and by the portion of the training curve matched with the other training curves beyond the drill bit. The log values ​​in the well beyond the drill bit depth are predicted by applying the model to the composite learning curves. Figure 3A to be published

Description

Title of the invention: Method for predicting a well log during well drilling. Technical field

[0001] The present invention relates in particular to the field of geosteering, which aims to guide the drilling operations of a well in an underground formation, based on the results of measurements (generally logging and seismic measurements) carried out in real time.

[0002] More specifically, the invention relates to a method for predicting the values ​​of a log in a well beyond the drill bit, simultaneously with the drilling of that well, in particular with a view to guiding the drilling of the well beyond the depth of the well drill bit, according to the prediction of the log values.

[0003] Geoguidance consists of guiding the direction of a wellbore in real time, based on an initial plan determined from a model of the subsurface formation structures built upstream and from well log measurements taken during drilling. Thus, in the drilling process, geoguidance involves adjusting the wellbore's position (angles of inclination and azimuth) on the fly to reach one or more geological targets. In the field of petroleum exploration, the objective is generally to orient the wellbore towards an area of ​​the subsurface formation containing hydrocarbons. In mature zones, geoguidance can be used to keep a wellbore within a particular section of a reservoir in order to minimize gas or water breakthroughs and maximize the well's economic production.

[0004] Geoguiding requires the prior construction of a model of the subsurface formation, also called a geological model. These models, well-known and widely used in the petroleum industry, must be representative of the boundaries of the geological layers as well as the rocks constituting these layers. This type of subsurface model is most often represented on a computer by a mesh or grid, generally three-dimensional, each cell being characterized by one or more petrophysical properties (facies, porosity, permeability, etc.). The construction of such a model, which must notably reflect the geometry of the geological layer boundaries, requires at least an interpretation of a seismic image, obtained following seismic acquisition and processing.Filling the grids with values ​​also requires analysis techniques for both seismic data (for example via stratigraphic inversion) and log data (for example gamma ray, sonic, density logs). apparent, electrical logs or well imaging logs). Thus, obtaining a geological model is a very long step and requires the intervention of several experts.

[0005] While the borehole is being drilled according to the information derived from the geological model, new geological information is collected using logging tools. Typically, a conventional setup of tools lowered into a well includes directional and inclination sensors, as well as a gamma-ray logging tool. Other conventional measurements include neutron density, prospective seismic surveys, and downhole pressure measurements. The measurements taken during drilling generally show some differences from what is expected from the model. The geological model is then updated with the new geological information, as is the well plan, to reach the corrected geological targets. Prior art

[0006] The following documents will be cited during the description: - Berg, C. R., & Newson, A. C. (2013, August). Geosteering using true stratigraphie thickness. In SPE / AAPG / SEG Unconventional Resources Technology Conférence. OnePetro. - Lallier, F., Caumon, G., Borgomano, J., Viseur, S., & Antoine, C. (2009). Dynamic time warping for stochastic stratigraphie well corrélation. Search and Discovery Article, 40473. - Naville, C., Serbutoviez, S., Throo, A., Vincké, O., & Cecconi, F. (2004). Seismic while drilling (SWD) techniques with downhole measurements, introduced by IFP and its partners in 1990-2000. Oil & Gas Science and Technology, 59(4), 371-403. - Sakoe, H., & Chiba, S. (1978). Dynamic programming algorithm optimization for spoken word récognition. IEEE transactions on acoustics, speech, and signal processing, 26(1), 43-49. - Song, F., & Toksoz, M. N. (2009). Model-guided Geosteering for Horizontal Drilling. Massachusetts Institute of Technology. Earth Resources Laboratory. -Wang, Z. (2017). Integrated geosteering workflow for optimal well trajectory. Doctoral dissertation, Memorial University of Newfoundland.

[0007] Geoguiding methods have seen significant growth in recent years.

[0008] We are familiar in particular with the document (Berg and Newson, 2013) which describes a method geoguiding allows the use of only natural radioactivity logs (known as Gamma-Ray logs). This method involves comparing a Gamma-ray logging in a horizontal well is compared to the corresponding log in a nearby vertical well (the pilot well). According to this method, the log record from the horizontal well is compressed and / or inverted until the distorted record can be matched as closely as possible with the record in the pilot well. Compression is necessary because relatively small stratigraphic intervals in the pilot well will be traversed over much greater distances in the horizontal well. Inversion is also necessary because drilling can cover a given stratigraphic interval from bottom to top as well as from top to bottom. The scale factor that enabled the matching of the two records is directly related to the inclination of the horizontal well.However, while this method requires little input data, it relies, like all known geoguiding techniques, on actual log measurements, taken on the fly, which are strictly valid only up to the drill bit.

[0009] We also know of the document (Wang, 2017) which concerns an integrated geoguidance methodology for determining an optimal well trajectory, combining geostatistical algorithms, in-drill logging measurements, and reservoir simulation methods. More specifically, the objective of this integrated technique is to explore a relatively rapid update process for the exploration of oil fields without complex structures. It presents a procedure for using geostatistical realizations to guide the drilling process and optimize the well trajectory with respect to production performance. The statistical reservoir realizations are updated with in-drill logging data throughout the drilling process. However, as with all known geoguidance techniques, this method relies on actual logging measurements, which are only available down to the drill bit depth..

[0010] We are also familiar with the document (Naville et al., 2004), which concerns seismic imaging during drilling. This technique allows, among other things, imaging ahead of the drill bit using the noise of the drill bit as a seismic source. It is therefore possible, thanks to this technique, to identify changes upstream of the drill bit that could impact the initial well plane. However, this method has the drawback of requiring, at a minimum, a seismic recording system, firstly of the echoes from the geological formations ahead of the drill bit, and secondly of the vibrations of the drill bit itself, in order to eliminate the signature of this seismic source in post-processing to refine the image. This technique is therefore more complex to implement than methods based on simple well logs.

[0011] We also know of the document (Song and Toksoz, 2009) which describes a method similar to seismic surveys during drilling but which uses real sources seismic activity is used instead of drilling noise. This technique therefore has the disadvantage of requiring both seismic sources and sensors in the field, but also requires seismic processing which can be significant.

[0012] The present invention overcomes these drawbacks. Indeed, the method according to the invention aims to predict well log values ​​beyond the depth of the drill bit, values ​​which can be directly used by conventional geoguidance methods, instead of the actual measured well log values ​​down to the drill bit depth only. The predictions before the drill bit can make it possible to identify, in advance and during drilling, future changes in lithology (for example, the transition from clay cover to reservoir), which is important information that can be immediately used as input data for any geoguidance process. Summary of the invention

[0013] The present invention relates to a method for predicting, during the drilling of a well in an underground formation, the values ​​of a log beyond the depth of the drill bit of said well, in particular with a view to guiding the drilling of said well beyond the depth of said drill bit of said well according to said prediction.

[0014] Said method is characterized in that at least the following steps are carried out:

[0015] A. a learning curve for said logging is acquired as a function of the depth in at least one training well, by means of a logging tool, said depth of said training well extending beyond the depth of the drill bit of said well;

[0016] B. a curve of said logging is acquired as a function of the depth in said well, by means of said logging tool, said second curve being acquired up to said depth of the drill bit of said well;

[0017] C. for each of said learning curves of said log, the learning curve of said log acquired in said training well is matched, up to said depth of the drill bit of said well, with the learning curve of said log acquired in said well, by means of a method for searching for similarity between signals recorded according to different scales;

[0018] D. A model is constructed to predict said log in said well by means of a supervised machine learning method trained on a training set comprising said log learning curves matched to said well drill bit depth with said curve acquired in said well, as well as said curve of said logging as a function of depth in said well;

[0019] E. Optionally, if a plurality of learning curves of said log have been acquired, said learning curves are matched with each other by means of said method of searching for similarity between signals recorded according to different scales;

[0020] F. For each of said learning curves, a learning curve is constructed composed of said log, formed on the one hand by the portion of said learning curve of said log matched with said curve of said log up to the depth of said drill bit, and on the other hand by the portion of said learning curve, optionally matched with said other learning curves when a plurality of learning curves of said log have been acquired, for depths beyond said depth of said drill bit; and

[0021] G. on predicting said log values ​​in said well at least beyond said depth of said drill bit of said well by applying said model to predict said log in said well constructed in step D) to said compound learning curves.

[0022] According to one embodiment of the invention, said method of searching for similarity between signals recorded according to different scales can be the dynamic time deformation method.

[0023] According to one embodiment of the invention, said log can be a natural radioactivity log.

[0024] According to one embodiment of the invention, said supervised machine learning method can be selected from the following list: the random forest method, extremely random decision trees, and gradient amplification.

[0025] According to one embodiment of the invention, steps B) to G) can be repeated for a succession of depths of said well drill bit.

[0026] The invention further relates to a method for geoguiding the drilling of a well, in which, for each depth in a succession of depths of said drill bit of said well:

[0027] a) values ​​of a log beyond said depth of the drill bit of said well are predicted by means of the method for predicting, during the drilling of a well in an underground formation, the values ​​of a log beyond the depth of said drill bit of said well as described above;

[0028] b) a drilling direction is determined for said well for depths beyond the depth of said drill bit of said well, based at least on said values ​​of said log predicted beyond the depth of the drill bit of said well.

[0029] In addition, the invention relates to a computer program product downloadable from a communication network and / or recorded on a computer-readable medium and / or executable by a processor, comprising program code instructions for the implementation of steps C) to G) of the process as described above, when said program is executed on a computer.

[0030] Other features and advantages of the process according to the invention will become apparent from the following description of non-limiting examples of embodiment, with reference to the figures attached and described below. List of figures

[0031] [Fig.1A] [Fig.1B]

[0032] Fig.1A and Fig.1B present curves of natural radioactivity as a function of depth measured respectively in a training well and in a target well.

[0033] [Fig.lC]

[0034] Fig. 1C illustrates a matching of homologous events of the curves of figures IA and IB detected by means of a method of searching for similarity between signals recorded according to different scales.

[0035] [Fig.1D] [Fig.1E]

[0036] Fig.1D and Fig.1E show the difference between the curves of figures IA and IB, respectively before and after matching of homologous events.

[0037] [Fig.2A] [Fig.2B] [Fig.2C]

[0038] Fig. 2A, Fig. 2B and Fig. 2C show natural radioactivity log curves measured respectively in a target well and in two training wells.

[0039] [Fig.3A]

[0040] Fig. 3A presents a natural radioactivity log curve predicted by the model according to the invention in the target well of Fig. 2A, as well as a natural radioactivity log curve actually measured in this target well.

[0041] [Fig.3B]

[0042] Fig. 3B shows the difference between the two curves of Fig. 3A. Description of the implementation methods

[0043] The invention relates to a method for predicting, during the drilling of a well in an underground formation, the values ​​of a log beyond the depth of the well drill bit, in particular with a view to guiding the drilling of the well according to the prediction.

[0044] According to the invention, and as will be described in more detail below, the prediction of well log values ​​is performed for a well, hereafter referred to as the "target well," at a specific drilling depth in that well, i.e., at the depth of the well drill bit. The objective is to predict the values ​​of a well log, which has been measured to the depth of the target well drill bit, beyond the depth of the target well drill bit. As will be described below, this prediction is made from a curve of the well log measured in the target well to the drill bit depth, and from at least one curve of the same well log measured in another well in the subsurface formation, hereafter referred to as the "training well," extending beyond the drill bit depth.Thus, the method according to the invention aims to predict log values ​​in a target well, beyond the drill bit depth, from logs measured in training wells extending beyond the drill bit depth. Preferably, the training well(s) traverse the same subsurface formation as the target well. Alternatively, a training well may traverse a geological analogue of the subsurface formation.

[0045] Generally, a well log results from a measurement taken using a logging tool (or probe) moving through a well. The result of a well log measurement is a curve representing the variations of a characteristic of a geological formation as a function of depth in the well. Examples of well logs include gamma ray logging, sonic logging, apparent density logging, electrical logging, and well imaging logging.

[0046] According to a preferred embodiment of the invention, the predicted log can be a natural radioactivity log, also known as a gamma-ray log or GR log. GR logs measure natural radioactivity in underground formations. They can be used to identify lithologies and to correlate depth intervals from one well to another. Sandstones and carbonates lacking clay have low concentrations of radioactive material and give low GR responses. As the clay content increases, the GR response increases due to the concentration of radioactive material in the clays. Predicting the gamma-ray response in a well ahead of the drill bit can therefore provide advance information about future changes in lithology (for example, the transition from a clay cover to a reservoir).This can reduce drilling risks and costs and improve well placement. It should be noted that the method according to the invention can be applied to any well log. GR logging is preferred here because of its strong potential for geoguidance.

[0047] The method according to the invention comprises at least the following steps:

[0048] 1) Acquisition of a logging learning curve in at least one well of learning

[0049] 2) Acquisition of a log curve in the target well

[0050] 3) Matching each learning curve with the curve of target well to the depth of the target well drill bit

[0051] 4) Construction of a model to predict the logging in the target well

[0052] 5) Matching learning curves with each other (Optional)

[0053] 6) Construction of a compound logging learning curve

[0054] 7) Prediction of logging in the target well beyond the depth of trepan

[0055] The method according to the invention is implemented for a drilling depth of the well. According to an embodiment of the invention which will be described in more detail below, step 1) can be performed only once, beforehand, and steps 2) to 7) of the method can be repeated as the target well is drilled, for example for a succession of drilling depths of the well.

[0056] According to one embodiment of the invention, in the case where several logs are measured in the target well, steps 1) to 7) of the method according to the invention can be repeated at a given drill depth for each log of the plurality of logs that one wishes to predict, in order to take into account the plurality of predicted logs in particular to guide the drilling of the well beyond the drill depth.

[0057] The process according to the invention includes steps implemented by computer means (a computer), in particular steps 3) to 7). The computer means may include data processing means (a processor) and data storage means (memory), as well as an input and output interface for capturing data and outputting the results of the process.

[0058] According to one embodiment of the invention, steps 2) to 7) can be carried out in real time during a drilling operation, so as to be able to implement geoguidance of well drilling.

[0059] The steps of the process according to the invention are detailed below.

[0060] 1) Acquisition of a logging learning curve in at least one well of learning

[0061] During this step, a learning curve of the well log of interest is acquired as a function of depth in at least one training well, using a logging tool. This acquisition can be carried out during the drilling of the training well or after the training well has been drilled.

[0062] According to the invention, a training well is a well whose depth extends beyond the depth of the drill bit of the target well (preferably beyond of at least 10 meters, preferably beyond at least 20 meters), in order to predict the log for depths exceeding the drill bit depth. It is clear that a learning curve acquired in a training well according to the invention includes log values ​​for depths exceeding the drill bit depth of the target well.

[0063] The method according to the invention requires at least one training well, preferably at least two, and most preferably at least five. The invention is particularly well-suited to be relevant from the oil exploration phase onward, when few wells are available. It is clear that the more training wells there are, the more reliable the well log prediction will be.

[0064] It is quite clear that the learning curves of the log of interest may have depth scales that are distinct from each other, but also from the depth scale of the log curve measured in the target well. Indeed, the geological layers of an underground formation are rarely flat and horizontal. Moreover, the thickness of the geological layers can vary laterally. Thus, the log recording corresponding to measured values ​​as a function of depth, a representative value of the same geological event (for example, a geological layer boundary) may not necessarily be at the same depth in one well as in another.

[0065] 2) Acquisition of a log curve in the target well

[0066] During this step, a curve of the log of interest is acquired as a function of the depth in the target well, using the logging tool that was used in step 1). It is clear that it is not necessarily strictly the same tool, but rather any tool suitable for measuring this same log.

[0067] According to the invention, the well log curve is acquired down to the depth of the drill bit in the target well, since the goal is precisely to predict the well log beyond the drill bit depth. As discussed above, the well log curve recorded in the target well may have a depth scale distinct from the depth scale of the training curves. In other words, a representative value of the same geological feature (for example, a geological layer boundary) may not necessarily be at the same depth in the target well as in the training well(s).

[0068] 3) Matching each learning curve with the curve of target well to the depth of the target well drill bit

[0069] During this step, each learning curve is matched with the curve measured in the target well up to the depth of the well drill bit, using a method for searching for similarity between signals recorded according to different scales.

[0070] In other words, this step aims to map the events of a curve measured in a training well to the corresponding events of the curve measured in the target well, down to the depth of the drill bit in the target well. Clearly, this results in a training curve whose events are scaled to the curve measured in the target well. In other words, the mapping is performed by considering the curve measured in the target well as the reference.

[0071] Figures IA to 1E illustrate the principle of this matching. More specifically, Figures IA and 1B respectively show a Cl curve and a C2 curve of the measurement of natural radioactivity GR as a function of depth Z, the Cl curve having been measured in a training well and the C2 curve having been measured in the target well. It can be observed that the Cl and C2 curves show homologous events, but shifted along the x-axis. Figure 1C illustrates, by means of gray lines, a matching of the homologous events of the Cl and C2 curves detected by means of a method for searching for similarity between signals recorded at different scales (in this case the dynamic time warp method described below).It can be observed that while some homologous events were recorded at very similar depths (primarily vertical gray lines), other homologous events were recorded at distant depths (slanted gray lines). Figures 1D and 1E show the difference between the Cl and C2 curves, respectively before and after matching. These figures clearly show that the difference between the Cl and C2 curves is significantly less pronounced after matching than before. More precisely, the standard deviation of the difference between the Cl and C2 curves decreases from 16% to less than 4% of the maximum amplitude of the natural radioactivity measurement (GR) after matching.

[0072] According to one embodiment of the invention, the Dynamic Time Warping (DTW) method can be used to perform the matching. This method was first described in the document (Sakoe and Chiba, 1978). For clarity, a simple acoustic analogy can be used to provide a practical understanding of DTW. Consider two singers, one rather good and the other rather mediocre. The latter sings off-key and does not respect the rhythm (sequence of time intervals between the sung notes), the key (absolute pitch of the notes), or the tempo (speed of execution) of the song. When faced with two recordings of the same song by these two singers, even an untrained ear will easily detect a certain similarity and, above all, several differences between these recordings.The DTW not only accurately quantifies the difference between these records, . But most importantly, it automatically establishes a temporal correspondence between the corresponding events observed in these two recordings. It is also quite clear that this technique, although called dynamic time warping, can be applied to any signal whose x-axis is different from time, in this case, whose x-axis corresponds to depth.

[0073] The document (Sakoe and Chiba, 1978) describes the original implementation of the DTW method as follows. Let X(i) and Y(j) be two signals (here time-domain, but the technique can be applied to any signal) to be warped. The function C(i,j) representing the cost density function of the warping operation for a pair (i,j) is defined as follows:

[0074] C(i, j) = Ab si, X(i) - Y(j C(L 1, j-1), C(L 1, j), C{i, j-1))

[0075] knowing that the first elements are given by the relations:

[0076] c(14) = Abs(x(l) - Y(l))

[0077] = Ab^X^ - Y(j)) +

[0078] c(i, 1) = Abs(X(i) - Y(1)) + C(i-1,1))

[0079] The warping function for the corresponding indices i and j of signals X and Y, respectively, is defined as follows: Let N be the number of samples of signals X and Y. We assume that the pairs of points (X(1), Y(1)) and (X(N), Y(N)) are the first and last corresponding points, respectively. Then, starting from the Nth corresponding point, the (Nl)th pair of corresponding points is the one whose cost density function corresponds to , and so on recurrently. Thus the (il)th pair of corresponding points is the one that corresponds to the MiiiC^ i-1), C(i-1,M), and so on until the first pair of points (X(1), Y(1)). From the correspondence function, we determine, for example, a signal Y that corresponds to the signal X. It is quite clear that we can use any refinement of this technique.

[0080] Alternatively, one can use the method for searching for similarity between signals recorded at different scales described in the document (Lallier et al., 2009), which concerns a variation of the DTW method applied to the correlation of geological horizons between wells. This method differs from the method of (Sakoe and Chiba, 1978) not only in that the curves The considerations are in depth and not in time, but primarily due to the fact that it uses a stochastic technique of stratigraphic well correlations, the aim being to automatically generate several possible sets of well correlations taking into account interpretations made along well paths. In practice, the correlation probability of each stratigraphic marker and the interval identified along the well path is calculated using:

[0081] - information available on the interpreted well (e.g. paleobathymetry, lithology),

[0082] - of a few reference correlation lines from stratigraphic boundaries first-rate clear

[0083] - of a sedimentological scenario on the studied area,

[0084] - of sedimentological concepts deemed applicable.

[0085] Using these probabilities, several possible sets of well correlations are generated stochastically in a modified version of the DTW algorithm of (Sakoe and Chiba, 1978).

[0086] 4) Construction of a model to predict the logging in the target well

[0087] During this step, a model is built to predict the log in the target well, using a supervised machine learning method trained on a training dataset comprising:

[0088] - the corresponding learning curve(s), up to said depth of the well drill bit, with the log curve measured in the target well. These curves correspond to the input data of the supervised machine learning method during this training phase;

[0089] - the curve measured in the target well. This curve corresponds to the output at reproduced by the supervised machine learning method during this training phase.

[0090] Thus, in this step, a supervised machine learning method is trained on the curves measured in the training wells, after these curves have been matched with the curve measured in the target well, so that the model can predict the log values ​​in the target well. A model for predicting the log in the target well is thus obtained. It is clear here that this is a regression problem, since we are determining values ​​of a continuous variable (the log measurement).

[0091] According to one embodiment of the invention, any one of the following methods can be used as a supervised machine learning method: the random forest method, extremely random decision trees, amplification of Gradient Boosting, Support Vector Machines, Generalized Linear Models, Generalized Additive Models, and Feedforward Artificial Neural Networks for deep learning are all methods used for deep learning. These methods are very simple to implement, even for non-experts, primarily because they require only a few parameters: the maximum computation time (typically a few minutes, for example, 200 seconds), and the distribution between the data that will actually be used for training and the data that will be used for validation (typically 80% and 20%, respectively).

[0092] According to a preferred embodiment of the invention, the supervised machine learning method can be chosen from the following list: random forest, highly random decision trees, and gradient amplification. Indeed, it has been found that for a number of application examples in which the method according to the invention has been implemented, the three aforementioned methods have most often ranked among the best performing in terms of prediction reliability.

[0093] According to a highly preferred embodiment of the invention, the gradient amplification method can be used. Indeed, it has been found that, for the application examples on which the method according to the invention has been implemented, this method is statistically the one that most often results in the best predictions. Gradient amplification uses two concepts: ensemble machine learning and machine learning boosting. In ensemble machine learning, training models are fitted to the data individually or combined into an ensemble. An ensemble is a combination of simple individual models that together create a new, more powerful model. Machine learning boosting, on the other hand, is a specific method for creating ensembles as follows: First, an initial model is fitted to the data (for example, using linear regression or a decision tree). Then, a subsequent model is built to improve the prediction of cases where the first model performs poorly.The combination of these two models is better than either of the two previous models taken separately. This process, called reinforcement, is repeated several times, with each successive model attempting to correct the shortcomings in the predictions of the previous combined models. Finally, gradient amplification is a specific type of reinforcement in machine learning. The general principle... This approach is based on the fact that the best possible next model, when combined with previous models, minimizes the overall prediction error. The term 'gradient amplification' comes from the fact that the target results for each case are defined according to the gradient of the error relative to the prediction. More precisely, each new model takes a step in the direction that minimizes the prediction error, within the space of possible predictions for each training case.

[0094] 5) Matching learning curves with each other (Optional)

[0095] This step is optional and will only be implemented if a plurality has been acquired learning curves of the log. During this optional step, the learning curves are matched with each other, using a method for finding similarity between signals recorded according to different scales.

[0096] In other words, during this step, if several training wells are available and a log curve has been recorded in each of these training wells, the curves recorded for each of these wells are matched. Indeed, each log curve has its own scale, due to the fact that geological layers are not flat, horizontal, and laterally isometric.

[0097] For the implementation of this step, the Dynamic Time Deformation technique described above can be used, or the technique described in the aforementioned document (Lallier, 2009) can be used to match the learning curves recorded for each of the learning wells.

[0098] Advantageously, the matching is performed by choosing as a reference the training well whose distance from the target well is the smallest, preferably whose distance from the target well is the smallest at the depth of the target well's drill bit. This allows the training curves to be matched with each other with, in principle (depending on the stratigraphy), the least difference in scale compared to the curve acquired in the target well.

[0099] 6) Construction of a compound logging learning curve

[0100] During this step, for each of the learning curves, a composite logging learning curve is constructed, formed on the one hand by the portion of the learning curve matched with the curve up to the depth of the drill bit, and on the other hand by a portion of the learning curve, optionally matched with the other learning curves when a plurality of logging learning curves have been acquired, for depths beyond the depth of the drill bit.

[0101] In other words, for each learning well, a composite curve is constructed, the part of which above the depth of the drill bit corresponds to the measured curve in this learning well and matching with the curve measured in the target well, and whose part below the depth of the drill bit corresponds to: - if only one logging learning curve has been acquired: to the portion of the curve measured in this learning well for depths beyond the drill bit depth, - if several log learning curves have been acquired: to the part of the curve measured in this learning well and matched with the other learning curves.

[0102] At the end of this step, we obtain for each learning well a composite learning curve, consistent with the target well above the drill bit, and consistent with any other wells below the drill bit.

[0103] 7) Prediction of logging in the target well beyond the depth of trepan

[0104] During this step, the log values ​​in the target well are predicted, beyond the depth of the well drill bit, by applying the log prediction model constructed in step 4) to the compound learning curves constructed in step 6).

[0105] In other words, the regression model constructed in step 4) is applied to the compound learning curves constructed in step 6), which include log values ​​beyond the drill bit depth of the target well, so as to predict log values ​​in the target well beyond the drill bit depth of the target well.

[0106] This gives us a curve of the predicted log of interest in the target well comprising values ​​at least for depths exceeding the depth of the drill bit of the target well.

[0107] According to one embodiment of the invention, the log of interest in the target well can be predicted not only beyond the depth of the target well drill bit, but also for depths below the depth of the target well drill bit. This can be used to verify the quality of the model built in step 4) in the portion below the depth of the target well drill bit, by comparing the predicted log values ​​with those of the measured log.

[0108] According to one embodiment of the invention, the log of interest in the target well can be predicted only beyond the depth of the target well drill bit. This reduces the execution time of the method according to the invention, and thus provides a predicted log curve more quickly, which is advantageous for the purpose of real-time geoguiding of the target well drilling.

[0109] Thus, the method according to the invention makes it possible to predict the values ​​of a log in the target well at least beyond the depth of the well drill bit.

[0110] The method can advantageously be implemented simultaneously with the drilling of this well, in order to guide the drilling of the well beyond the depth of the well drill bit, based on the prediction of the log values. Indeed, prior art geoguidance methods use actual log values, measured only to the depth of the drill bit, whereas the method according to the invention determines predicted log values ​​beyond the depth of the drill bit, which can therefore be directly used by geoguidance methods.

[0111] According to one embodiment of the invention, step 1) can be performed only once, and steps 2) to 7) as the target well is drilled, for example, for a succession of depths. This allows the logging prediction to be refined progressively for greater depths in the target well, as the logging measurements taken in the target well progress in depth with the advance of drilling.

[0112] According to one embodiment of the invention, steps 2) to 7) of the process can be repeated for a succession of well drill bit depths, for example every 10m, preferably every 5m, preferably every 1m. Alternatively, steps 2) to 7) of the process can be repeated for a succession of time steps during the drilling of the target well, for example every 60 minutes, preferably every 30 minutes, preferably every 10 minutes, these durations being predefined according to the rate of advance of the well drilling (typically between 1.5m / hour and 15m / hour).

[0113] It should be noted that for the application examples on which the method according to the invention was implemented, the execution time of steps 3) to 7) was on the order of 3 minutes on an Intel(R) Xeon(R) CPU E5-1620 v3 @ 3.50GHz processor (16.0 GB RAM), which is significantly less than the conventional advance time of the drilling tool over a few meters: the logging prediction can therefore be done almost in real time.

[0114] The invention further relates to a method for geoguiding the drilling of a well, in which, for each depth in a succession of depths of the drill bit of said well:

[0115] a) values ​​of a log are predicted beyond the depth of the drill bit of said well by means of the method for predicting, during the drilling of a well in an underground formation, the values ​​of a log beyond the depth of the drill bit of the well as described above;

[0116] b) the direction of drilling the well is determined for depths beyond the depth of the well drill bit, based at least on the predicted log values ​​beyond the depth of the well drill bit.

[0117] The person skilled in the art has perfect knowledge of the means to carry out step b) above since, in the field of geoguidance, he determines the direction of drilling of the well from curves of logs actually measured, whereas in the case of this case, it is sufficient to consider the log curves predicted according to the invention in addition.

[0118] According to one embodiment of the invention, in step b), the technique described in the aforementioned document (Berg and Newson, 2013) can be applied, which relates to a geoguiding method allowing the use of only Gamma-ray logs, but using the log curves predicted according to the invention beyond the depth of the well drill bit and not only the log curves actually measured.

[0119] According to another embodiment of the invention, the technique described in the aforementioned document (Wang, 2017), which relates to an integrated geoguidance methodology combining geostatistical algorithms, in-drill logging techniques, and reservoir simulation methods, can alternatively be applied. According to this embodiment, the method described in the document (Wang, 2017) can be applied using the logging curves predicted according to the invention beyond the depth of the well drill bit, and not only the logging curves actually measured.

[0120] According to one embodiment of the invention, in step b) at least the angles of inclination and azimuth of the well trajectory are determined for depths extending between 10 and 20 m beyond the depth of the drill bit, for each depth of the well drill bit in the succession of well drill bit depths. Examples

[0121] The characteristics and advantages of the method according to the invention will become clearer upon reading the application example below.

[0122] More specifically, the process according to the invention has been applied to an underground formation located in the Western Australian offshore, more specifically in the North Carnarvon Basin, considered to be the main hydrocarbon basin in Australia.

[0123] Figure 2A shows a natural radioactivity log curve (NRL) measured in a target well, hereafter denoted T (curve labeled T in Figure 2A). The target well passes through a 6-meter oil column in the Barrow Group sandstones. This well also passes through water-bearing sands, with a thin oil column, marls, and shales.

[0124] Two training wells, denoted Li (i=l to 2), are available, which are respectively about 3.5 km and 8 km away from the target well T. Figures 2B and 2C show the natural radioactivity log curves GR measured respectively in well L1 and in well L2 (curves denoted L1 and L2 in Figures 2B and 2C).

[0125] Natural radioactivity values ​​are available for depths extending to at least 3200 m: 3277 m for target well T, 3295 m for well L1, and 3200 m for well L2. To illustrate the reliability of the method according to the invention, we will attempt to predict the GR log values ​​for depths extending beyond 3100 m, and we will verify the quality of the prediction using the GR log curve actually measured in target well T for depths ranging from 3100 to 3200 m. In other words, for the purpose of validating the method according to the invention, we apply the method according to the invention assuming that the depth of the drill bit in the target well is 3100 m.

[0126] In accordance with step 3) described above, the GR curves of wells L1 and L2 are matched with the GR curve of target well T up to a depth of 3100m, using the method described in the aforementioned document (Sakoe and Chiba, 1978).

[0127] Furthermore, the method according to the invention is implemented using the gradient boosting method as a supervised machine learning method. This machine learning method is trained on the GR curves of wells L1 and L2, matched with the GR curve of the target well T down to a depth of 3100 m, as well as on the GR curve of the target well T. This yields a model for predicting a well log (here, an GR log) in the target well T, in accordance with step 4) described above.

[0128] Then, according to step 5) described above, the GR curves of wells L1 and L2 are matched with each other, taking as a reference well L1 which is the training well closest to the target well T.

[0129] In accordance with step 6) described above, composite learning curves are generated for the training wells L1 and L2, formed by the portion of the GR curves of these wells up to the depth of 3100m after their matching with the target well T, and by the portion of the GR curves of the wells L1 and L2 beyond the depth of 3100m after their matching with each other.

[0130] The model for predicting a log in the target well constructed according to step 4) is then applied to the compound learning curves constructed as described above in step 6). [Fig.3A] shows the GRP curve which is the GR curve predicted by the model, as well as the GRM curve which is the GR curve actually measured, and [Fig.3B] shows the difference between these two curves.

[0131] It can be observed that the predicted GRP curve is perfectly consistent with the actually measured GRM curve from 3100 m to 3200 m, i.e., up to a depth of 100 m beyond the depth to which the model was trained (the drill bit depth in the method according to the invention). Indeed, in this depth range, the error, quantified by the standard deviation of the difference between the GR curve predicted by the model according to the invention and the actually measured GR curve, is less than 2% of the maximum amplitude of the actually measured GR curve. Beyond 3200 m, the errors increase significantly and exceed 10% on average of the maximum amplitude of the actual GR measurement.

[0132] Thus, the application of the method according to the invention made it possible to reliably predict a well log (here, a GR log) beyond the depth of the well drill bit, for approximately one hundred meters. This is very advantageous because, for typical drilling speeds, between 1.5 m / hour and 15 m / hour, this corresponds to more than 6 hours of drilling to reach this limit. However, the calculation time of the method according to the invention is significantly shorter, typically on the order of a few minutes. Furthermore, it is possible to rerun the method according to the invention as the drilling tool advances and to update the model as soon as new well log measurements are available. Thus, the well log prediction can be updated almost in real time.Alternatively, it may be sufficient to update the prediction model of a log and predict a new log using this updated prediction model preferably at least every 10 m, or very preferably every 5 m of drill bit advance in the target well.

[0133] Thus, the present invention makes it possible to predict log responses at depths not yet reached by drilling, typically at distances of several tens of meters in front of the drill bit during drilling, in particular to guide drilling to such distances from these predictions, and not solely from actual measurements, as according to the prior art.

Claims

20 Demands

1. A method for predicting, during the drilling of a well in an underground formation, the values ​​of a well log beyond the depth of the drill bit of said well, in particular for the purpose of guiding the drilling of said well beyond the depth of said drill bit. drilling of said well according to said prediction, characterized in that at least the following steps are carried out: A. a learning curve of said logging is acquired as a function of depth in at least one training well, by means of a logging tool, said depth of said training well extending beyond the depth of the drill bit of said well; B. a curve of said logging is acquired as a function of the depth in said well, by means of said logging tool, said second curve being acquired up to said depth of the drill bit of said well; C. for each of the said learning curves of the said log, the said learning curve of the said log acquired in the said training well is matched, up to the said depth of the drill bit of the said well, with the said curve of the said log acquired in the said well, by means of a method of searching for similarity between signals recorded according to different scales; D. a model is constructed to predict said log in said well by means of a supervised machine learning method which is trained on a training set comprising said learning curves of said log matched up to said drilling depth of said well with said curve acquired in said well, and said curve of said log as a function of depth in said well; E. Optionally, if a plurality of learning curves for said log have been acquired, said learning curves are matched with each other by means of said method of searching for similarity between signals recorded according to different scales; F. For each of the said learning curves, a learning curve is constructed composed of the said log, formed on the one hand by the portion

2.

3.

4.

5. of said learning curve of said logging, matched with said learning curve of said logging up to the depth of said drill bit, and on the other hand by the portion of said learning curve, optionally matched with said other learning curves when a plurality of learning curves of said logging have been acquired, for depths beyond said depth of said drill bit; and G. on predicting said log values ​​in said well at least beyond said depth of said drill bit of said well by applying said model to predict said log in said well constructed in step D) to said compound learning curves. A method according to claim 1, wherein said method of searching for similarity between signals recorded according to different scales is the dynamic time deformation method. A method according to any one of the preceding claims, wherein said log is a natural radioactivity log. A method according to any one of the preceding claims, wherein said supervised machine learning method is selected from the following list: random forest, highly random decision trees, and gradient amplification. A method according to any one of the preceding claims, wherein steps B) to G) are repeated for a succession of depths of said well drill bit.

6. A method for geoguiding the drilling of a well, wherein, for each depth in a succession of depths of said drill bit of said well: a. values ​​of a log beyond said depth of the drill bit of said well are predicted by means of the method for predicting, during the drilling of a well in an underground formation, the values ​​of a log beyond the depth of said drill bit of said well according to any one of the preceding claims; b. a drilling direction of said well is determined for depths beyond the depth of said drill bit of said well, based at least on said values ​​of said log predicted beyond the depth of the drill bit of said well.

7. Product computer program downloadable from a communication network and / or stored on a computer-readable medium and / or executable by a processor, comprising program code instructions for carrying out steps C) to G) of the method according to any one of claims 1 to 5, when said program is executed on a computer.

Images