Method for generating a model to predict at least one property of a fluid, computer-based model for predicting at least one property of a fluid, tangible computer-readable medium and method for predicting a value of a property of a fluid

Abstract

ESTIMATING RESERVOIR FLUID PROPERTIES USING SLUDGE GAS DATA. A method is disclosed for generating a machine learning model to predict a reservoir fluid property, such as gas-oil ratio or density, based on standard slurry gas and petrophysical data. This model was found to predict these reservoir fluid properties with an accuracy that is close to that which can be achieved using advanced slurry gas data. This is advantageous because standard slurry gas and petrophysical data are much more readily available than advanced slurry gas data.

Description

1 / 22 “METHOD FOR GENERATING A MODEL TO PREDICT AT LEAST ONE PROPERTY OF A FLUID, COMPUTER-BASED MODEL TO PREDICT AT LEAST ONE PROPERTY OF A FLUID, TANGIBLE COMPUTER-READABLE MEDIUM AND METHOD FOR PREDICTING A VALUE OF A PROPERTY OF A FLUID”

[0001] This disclosure relates to a logging technique for use during well drilling and, in particular, to a technique that uses mud gas data to predict reservoir fluid properties.

[0002] Drilling fluid is a fluid used to assist in drilling wells in the ground. The main functions of drilling fluid include providing hydrostatic pressure to prevent formation fluids from entering the wellbore, keeping the drill bit cool and clean during drilling, removing drilling cuttings, and suspending drilling cuttings while drilling is paused and when the drill assembly is brought into and out of the hole.

[0003] Drilling fluids are broadly categorized into water-based drilling fluid, non-aqueous drilling fluid, often referred to as oil-based drilling fluid, and gaseous drilling fluid. Liquid drilling fluids, that is, water-based drilling fluid or non-aqueous drilling fluid, are commonly referred to as drilling mud.

[0004] Mud gas logging involves collecting data from hydrocarbon gas detectors that record the levels of gases brought to the surface in the drilling mud during a borehole drilling operation.

[0005] Conventionally, mud gas logging is used to identify the location of oil and gas zones as they are penetrated, which can be identified by the presence of hydrocarbon gas in the mud system. This can be used to provide a general indication of the reservoir type, as well as to determine where to collect downhole fluid samples for more detailed fluid composition analysis. The presence of hydrocarbon gas can be detected, for example, with a total gas detector. Petition 870260016305, dated 23 / 02 / 2026, page 11 / 56 2 / 22

[0006] Once the presence of hydrocarbon gas is detected, its composition can be examined, for example, with a gas chromatograph.

[0007] The most common gas component present is usually methane (Ci). The presence of heavier hydrocarbons, such as C2 (ethane), C3 (propane), C4 (butane), and C5 (pentane), may indicate an oil or wet gas zone. Heavier molecules, up to about C7 (heptane), may also be detectable, but are usually present only in very low concentrations. Consequently, the concentrations of these hydrocarbons are often not recorded.

[0008] There are two types of mud gas data that can be collected, which are sometimes referred to as standard mud gas logging and advanced mud gas logging. The equipment for standard mud gas logging and advanced mud gas logging is different.

[0009] For a standard slurry gas system, the degasser usually does not have heating or uses constant volume gas separation. There is also only one slurry sample sampling point (outside) and therefore it is not suitable for recycling correction. The measured gas composition is usually referred to as standard slurry gas data, which is not directly comparable to the actual Ci to C5 composition of the reservoir fluid sample.

[0010] For an advanced slurry gas system, the degasser has heating and generally uses a constant volume for gas separation. There are two slurry sampling points (“out” and “in”), and therefore it is possible to perform recycling correction. The measured gas composition is generally referred to as advanced slurry gas data.

[0011] When generating advanced mud gas data, in order to make the data closely match the actual reservoir fluid concentrations C1 to C5, two correction processes are applied to the raw mud gas data from the advanced mud gas logging system.

[0012] First, a recycling correction is made to eliminate gas contamination from gases originating from previous injections of sludge. Petition 870260016305, dated 23 / 02 / 2026, p. 12 / 56 3 / 22 drilling. This correction is applied based on a separate measurement of mud gas that was taken before the drilling mud was injected into the drill string.

[0013] Secondly, an extraction efficiency correction step is applied to increase the concentration of intermediate components (from C2 to C5), so that the concentration of these components, relative to the concentration of C1, more closely resembles the relative compositions of a corresponding reservoir fluid sample. The extraction efficiency correction is applied based on the type of drilling mud used for the well.

[0014] In the past, advanced mud gas data would have been examined to estimate certain reservoir fluid properties using broad empirical correlations between advanced mud gas composition and certain reservoir fluid properties. For example, extremely dry gas reservoirs should comprise mostly C1 and not so much C2+, for example, with each of the C1 / C2, C1 / C3, C1 / C4 and C1 / C5 ratios (for crude mud gas data) being greater than 50. Wet gas reservoirs will generally have ratios between 20 and 50, and oil reservoirs will have ratios between 2 and 20.

[0015] Recently, an advanced machine learning model was developed, making it possible to predict reservoir fluid properties with much greater accuracy from advanced mud gas data, even when those properties depend on the oil (C7+) portion of the fluid that is not measured by mud gas data.

[0016] Details of how such a machine learning model was trained to determine a reservoir fluid gas-oil ratio based on advanced mud gas data can be found in the article by Tao Yang et al. (2019), “A Machine Learning Approach to Predict Gas Oil Ratio Based on Advanced Mud Gas Data”. Society of Petroleum Engineers. doi:10.2118 / 195459-MS

[0017] Advantageously, this model can be used to generate a substantially continuous profile of the respective reservoir fluid property. This Petition 870260016305, dated 23 / 02 / 2026, p. 13 / 56 4 / 22 was not previously possible and, in the past, it was necessary to rely on downhole fluid samples. Furthermore, the model allows reservoir fluid property predictions to be made at a very early stage of the drilling process and without the need to interrupt the drilling process, as might be necessary to collect downhole fluid samples or similar.

[0018] This model has been considered very useful, but it is limited in that it requires the availability of advanced mud gas data. There is a need for a technique that can be used when advanced mud gas data are not available.

[0019] The present invention provides a method for generating a model to predict at least one property of a fluid at a sample location within a hydrocarbon reservoir, comprising: a) provide a training dataset comprising input data and target data, the input data comprising slurry gas data and petrophysical data for each of a plurality of sample locations and the target data comprising at least one fluid property for each of the plurality of sample locations; and b) generate a model using the training dataset such that the model can be used to predict at least one fluid property at the sample location based on measured mud gas data and measured petrophysical data for the sample location, where a drilling fluid recycling correction has not been applied to the mud gas data.

[0020] It is a commonly held belief within the oil and gas industry that petrophysical data provide only a qualitative indication of a reservoir fluid. The data generally predict lean gas with good certainty, but have reduced accuracy when used to distinguish condensate from rich gas and oil. No Petition 870260016305, dated 23 / 02 / 2026, page 14 / 56 5 / 22 However, it was identified that by supplementing standard mud gas data with petrophysical data, it is possible to provide an estimate of certain reservoir fluid properties with an accuracy that is close to the accuracy that can be achieved using advanced mud gas data alone.

[0021] This is particularly advantageous when it is desirable to generate fluid property profiles for large quantities of existing and new wells because standard mud gas data and petrophysical data are collected for almost all wells, including exploration and production wells. Whereas, the additional cost of collecting advanced mud gas data and, particularly, of having the two sets of mud gas analysis tools needed to perform recycling correction, means that it is often only collected when drilling a few exploration wells. The number of wells with advanced mud gas data represents only a small portion of the total number of wells with standard mud gas data and petrophysical data.

[0022] Furthermore, the above technique allows reservoir fluid property profiles to be generated for new wells at a reduced cost, since it does not require the additional costs associated with collecting advanced mud gas data. Importantly in this respect, petrophysical data can be collected as a substantially continuous profile, similar to mud gas data. This contrasts with downhole fluid sampling data, which requires interrupting the drilling process, adding significant additional costs to the drilling process.

[0023] In some embodiments, the input data may not comprise downhole fluid sampling data and the model may not require downhole fluid sampling data as an input to predict at least one fluid property at the sample location.

[0024] The method is preferably a computer-implemented method and the model generation may involve instructing a machine learning algorithm to generate the model using the training dataset so that the model can be used to predict at least one fluid property. Petition 870260016305, dated 23 / 02 / 2026, page 15 / 56 6 / 22 at the sample location based on mud gas data measured for the sample location.

[0025] At least one property is preferably a property influenced by the oil-related components of the fluid. That is, a property that is not just the product of gaseous hydrocarbons within the fluid, whose composition can be predicted based on slurry gas data.

[0026] At least one property may comprise a fluid density at the sample location. It will be appreciated that the density can be calculated under atmospheric conditions or under reservoir conditions (e.g., taking into account the oil formation volume factor).

[0027] At least one property may comprise a gas-oil ratio. That is, a ratio between the amount of gaseous hydrocarbon and the amount of liquid hydrocarbon, which is normally determined under surface conditions. The gas-oil ratio is preferably a volume ratio. The gas-oil ratio may be a single-flash gas-oil measurement. However, any suitable gas-oil measurement may be used.

[0028] At least one property may comprise a saturation pressure of the fluid at the sample location. That is, the pressure at which a secondary phase will appear as the pressure is depleted.

[0029] At least one property comprises a fluid formation volume factor at the sample location. That is, the ratio of the fluid volume under reservoir (in situ) conditions to the fluid volume under surface conditions.

[0030] At least one property may comprise a concentration of a hydrocarbon within the fluid at the sample location. The hydrocarbon may be a hydrocarbon not included in the mud gas data. For example, the hydrocarbon may be a C7+ hydrocarbon. That is, the hydrocarbon may be a C7 hydrocarbon or it may be a hydrocarbon heavier than C7, for example, a C8 hydrocarbon or heavier. The hydrocarbon may be a hydrocarbon that is substantially an oil under reservoir conditions. Petition 870260016305, dated 23 / 02 / 2026, page 16 / 56 7 / 22 The hydrocarbon concentration can be an absolute concentration (e.g., a molar concentration), or it can be a relative concentration (e.g., a ratio compared to Ci), or it can be a normalized concentration.

[0031] The reservoir can be a gas reservoir, a multiphase reservoir, or an oil reservoir.

[0032] At least one property for each sample location can be determined from reservoir fluid property data associated with the sample location. Reservoir fluid property data may comprise composition data measured for a fluid at the sample location. Reservoir fluid property data may contain the composition of C1 to C7+ hydrocarbons at the sample location and preferably C1 to C20+ hydrocarbons and, more preferably, C1 to C36+ hydrocarbons at the sample location. As used herein, the notation Cx+ should be understood to mean Cx or heavier hydrocarbons.

[0033] The mud gas data from the training dataset may comprise mud gas data measured for the sample location, i.e., standard mud gas data measured for the sample location.

[0034] Measured mud gas data may be indicative of a composition of gases released from the drilling fluid used during drilling through the sample location (i.e., passing through a drill bit performing the drilling). Measured mud gas data may be indicative of a concentration of at least C1 to C4 gaseous hydrocarbons and preferably at least C1 to C5 gaseous hydrocarbons that were released from the drilling mud.

[0035] As discussed above, mud gas data preferably do not include advanced mud gas data, i.e., where the mud gas data have not been corrected to match the gaseous hydrocarbon composition of the fluid at the sample location.

[0036] A drilling fluid recycling correction refers to correcting mud gas data to remove errors due to gases released from Petition 870260016305, dated 23 / 02 / 2026, page 17 / 56 8 / 22 previous drilling operations, such as due to drilling fluid recycling. Typically, this would require a measurement of reference mud gas data collected before injecting drilling mud into the drill string.

[0037] Optionally, an extraction efficiency correction was not applied to the slurry gas data.

[0038] An extraction efficiency correction refers to correcting mud gas data (Ci to C5) to closely match reservoir fluid composition due to different hydrocarbon components having different abilities to vaporize from drilling mud.

[0039] When an extraction efficiency correction has not been applied, the training data may additionally include drilling mud composition data. This may allow the machine learning model to correct the data within the model generated by the machine learning algorithm.

[0040] Alternatively, an extraction efficiency correction may have been applied to the standard mud gas data. Often, the type of drilling mud used for a well is known, and in many cases, extraction efficiency corrections can be estimated by Equation of State (EOS) simulation or approximated by testing. Consequently, even when using standard mud gas data, it may be possible to retrospectively apply an extraction efficiency correction to the standard mud gas data.

[0041] Optionally, mud gas data were collected without the use of heating. Although standard mud gas data may use heating; heating has often not been used when collecting mud gas data. Consequently, where it is desirable to use the model to examine existing wells, a model trained using mud gas data collected without the use of heating is particularly useful.

[0042] Petrophysical data may include any one or more of: apparent density, neutron porosity, resistivity data, data Petition 870260016305, dated 23 / 02 / 2026, page 18 / 56 9 / 22 acoustic, natural gamma rays, nuclear magnetic resonance data, as well as time-delay and gamma-ray spectroscopy data from pulsed neutron measurements and the like. Optionally, petrophysical data may comprise two or more of these data types.

[0043] Model generation may comprise: training a machine learning algorithm with a first subset of the training dataset; and testing the machine learning algorithm with a second disjoint subset of the training dataset. The first subset preferably comprises at least 50% of the samples from the training dataset. The second subset preferably comprises at least 10% of the samples from the training dataset.

[0044] Viewed from a second aspect, the present invention provides a computer-based model for predicting at least one property of a fluid at a sample location within a hydrocarbon reservoir based on measured mud gas data and measured petrophysical data for that sample location, the computer-based model having been generated by the above method.

[0045] Viewed from the third aspect, the present invention provides a tangible, computer-readable means of storing the computer-based model.

[0046] Viewed from a fourth aspect, the present invention provides a method for predicting a value of a fluid property at a sample location within a hydrocarbon reservoir, the method comprising: receiving measured mud gas data and measured petrophysical data for the sample location; and predicting the value of the fluid property at the sample location by providing the measured mud gas data and measured petrophysical data to the computer-based model.

[0047] The method may also include determining a quality for the measured mud gas data and / or the measured petrophysical data.

[0048] The method may also involve generating a confidence indication associated with the predicted value of the fluid property. The confidence indication may be a numerical indication, but other indications may be used, such as Petition 870260016305, dated 23 / 02 / 2026, p. 19 / 56 10 / 22 color indications (e.g., red / yellow / green) or word indications (e.g., good / bad).

[0049] The confidence indication may be based on the quality of the measured mud gas data and / or measured petrophysical data.

[0050] For a single data point, the confidence indication may be reduced by one or more concentrations of Ci, C4 or C5 that are below a respective predetermined limit.

[0051] When standard mud gas data are taken from a series of locations at different depths, the confidence indication may be reduced by fluctuations in a component concentration of the mud gas data greater than a threshold amplitude within a predetermined depth range.

[0052] When standard mud gas data and / or petrophysical data are taken from a series of locations at different depths, the indication of confidence may be reduced by the lack of a predetermined number of previous measurements or over a predetermined depth range.

[0053] Viewed from a fifth aspect, the present invention provides a method for predicting a fluid property value of a fluid along a well length through a hydrocarbon reservoir, the method comprising: predicting a fluid property value of a fluid at a plurality of sample locations along a well length using the above method.

[0054] The method may comprise: displaying, using an electronic display screen, a graph representing the predicted value of the fluid property against a location of the respective sample location for each of the plurality of sample locations along the length of the well.

[0055] The method may also include: indicating, using the electronic display screen, a confidence indication associated with one or more of the predicted value. For example, the confidence indication may be illustrated numerically, verbally, chromatically, or iconographically. Petition 870260016305, dated 23 / 02 / 2026, page 20 / 56 11 / 22

[0056] The entire method described above, that is, the methods of the first, fourth, and fifth aspects, can be implemented in any suitable and desired manner and on any suitable and desired platform. In a preferred embodiment, the methods are each a computer-implemented method, for example, the steps of the method are performed by processing circuits.

[0057] The methods according to the present invention can be implemented at least partially using software, for example, computer programs. Thus, it will be seen that, when viewed from additional aspects, the present invention provides computer software specifically adapted to perform the methods described in this document when installed on a data processor, a computer program element comprising portions of computer software code to execute the methods described in this document when the program element is executed on a data processor, and a computer program comprising adapted code to execute all the steps of a method or methods described in this document when the program is executed on a data processing system.

[0058] The present invention also extends to a computer software carrier comprising such software arranged to perform the steps of the methods of the present invention. Such a computer software carrier may be a physical storage medium, such as a ROM chip, CD-ROM, DVD, RAM, flash memory or disk, or it may be a signal, such as an electronic signal through wires, an optical signal or a radio signal, such as a satellite or the like.

[0059] It will also be appreciated that not all steps of the methods of the present invention need to be performed by computer software and, thus, from a further broad embodiment, the present invention provides computer software and such software installed on a computer software carrier to perform at least one of the steps of the methods set forth in this document.

[0060] The present invention can, consequently, be suitably Petition 870260016305, dated 23 / 02 / 2026, page 21 / 56 12 / 22 incorporated as a computer program product for use with a computer system. Such implementation may comprise a series of computer-readable instructions, which may be fixed on a tangible, non-transient medium, such as a computer-readable medium, for example, floppy disk, CD-ROM, DVD, ROM, RAM, flash memory or hard disk. It may also comprise a series of computer-readable instructions transmissible to a computer system, via a modem or other interface device, through a tangible medium, including, but not limited to, optical or analog communication lines, or intangibly using wireless techniques, including, but not limited to, microwave, infrared or other transmission techniques. The series of computer-readable instructions incorporates all or part of the functionality described previously in this document.

[0061] Those skilled in the art will appreciate that such computer-readable instructions can be written in a range of programming languages ​​for use with many computer architectures or operating systems. Furthermore, such instructions can be stored using any memory technology, present or future, including but not limited to semiconductor, magnetic or optical, or transmitted using any communication technology, present or future, including but not limited to optical, infrared or microwave. It is contemplated that such a computer program product can be distributed as a removable medium with accompanying printed or electronic documentation, for example, shrink-wrapped software, pre-loaded with a computer system, for example, in a system ROM or hard disk, or distributed from a server or electronic bulletin board via a network, for example, the Internet or the World Wide Web.

[0062] Certain preferred embodiments of the present disclosure will now be described in more detail by way of example only and with reference to the accompanying drawings in which:

[0063] Figure 1 is a schematic illustration of a mud gas analysis tool; and Petition 870260016305, dated 23 / 02 / 2026, page 22 / 56 13 / 22

[0064] Figure 2 illustrates a workflow for a machine learning algorithm to generate a first model to predict a gas-oil ratio using a training dataset;

[0065] An exemplary standard mud gas analysis tool 20 is shown schematically in Figure 1.

[0066] Tool 20 is coupled to a flow line 10 containing drilling mud returned from a wellbore hole. As discussed above, the drilling mud can be water-based mud or oil-based mud.

[0067] The tool 20 comprises a sampling probe 22 disposed in relation to the flowline 10 so as to collect a sample 24 of the drilling mud from the flowline 10. The drilling mud sample 24 is preferably a continuous sample, i.e., such that a portion of the drilling mud flow within the flowline 10 is diverted through the mud gas analysis tool 20.

[0068] The drilling mud sample 24 is fed to a gas separation chamber 26 where at least a portion of the gas carried by the drilling mud is released. The drilling mud sample may optionally be heated by a heater 28 upstream of the gas separation chamber 26. Heating the drilling mud sample 24 helps to release the gas from the drilling mud sample 24. Typically, the mud sample 24 is not heated and the temperature is typically ranging from 10°C to 60°C. However, in some implementations, heating is used at 80°C to 90°C.

[0069] The released gas 30 is directed from the separation chamber 26 to a gas analysis unit (not shown), while the degassed sludge 32 is returned to the flow line 10 or to another location for reuse.

[0070] The gas analyzer may comprise a total gas detector, which can provide a basic quantitative indication as to the amount of gas being extracted from the drilling mud by the tool 20. Total gas detection typically incorporates a catalytic filament detector, also called a hotwire detector, or a hydrogen flame ionization detector. Petition 870260016305, dated 23 / 02 / 2026, page 23 / 56 14 / 22

[0071] A catalytic filament detector operates on the principle of catalytic combustion of hydrocarbons in the presence of a platinum wire heated to a gas concentration below the lower explosive limit. The increase in heat due to combustion causes a corresponding increase in the resistance of the platinum wire filament. This increase in resistance can be measured using a Wheatstone bridge or equivalent detection circuit.

[0072] A hydrogen flame ionization detector works on the principle of ionizing a hydrocarbon molecule in the presence of a very hot hydrogen flame. These ions are subjected to a strong electric field, resulting in a measurable current flow.

[0073] The gas analysis device may additionally or alternatively comprise an apparatus for detailed analysis of the hydrocarbon mixture. This analysis is generally performed by a gas chromatograph. However, various other detection devices may also be used, including a mass spectrometer, an infrared analyzer, or a thermal conductivity analyzer.

[0074] A gas chromatograph is a rapid sampling and batch processing instrument that provides a proportional analysis of a series of hydrocarbons. Gas chromatographs can be configured to separate almost any set of gases, but typically oilfield chromatographs are designed to separate the paraffin series of hydrocarbons from methane (Ci) to pentane (Cs) at room temperature, using air as a carrier. The chromatograph will report (in units or molar percent) the amount of each component of the gas detected.

[0075] A carrier gas stream 34, commonly comprising air, may be supplied to the separation chamber 26 and mixed with the released gas 30 to form a gas mixture 36 which is supplied to the gas analysis unit. The carrier gas stream 34 provides a continuous flow of carrier gas in order to provide a substantially continuous flow rate of the gas mixture 36 from the separation chamber 26 to the gas analysis unit. Furthermore, in the case Petition 870260016305, dated 23 / 02 / 2026, page 24 / 56 15 / 22 of a gas analyzer comprising a combustor, the use of air as a carrier gas can provide the oxygen necessary for combustion.

[0076] In some arrangements, tool 20 can be configured to detect and / or remove H2S from the gas to avoid adverse effects that could influence hydrocarbon detection.

[0077] In some embodiments, non-combustible gases, such as helium, carbon dioxide and nitrogen, can be detected by the gas analyzer in conjunction with hydrocarbon profiling.

[0078] The following technique seeks to use a machine learning algorithm to produce a model that accurately estimates certain reservoir fluid-related properties, in particular the gas-oil ratio and reservoir fluid density, based on standard mud gas data and other petrophysical data.

[0079] Figure 2 illustrates a workflow 100 for training the machine learning algorithm in order to generate a model for predicting a reservoir's gas-oil ratio based on measured standard mud gas data.

[0080] In the following example, an input dataset of 102 is used as a training dataset and comprises data relating to a plurality of reservoir samples.

[0081] The input dataset comprises reservoir fluid property data from a large number of reservoir fluid samples. Reservoir samples can be obtained, for example, by downhole fluid sampling. However, other techniques can also be used, for example, by taking a well fluid sample after the well has been completed.

[0082] Reservoir fluid property data should include at least hydrocarbon composition data, which may be in the form of direct measurements of the concentration of each hydrocarbon component within the sample, typically covering C1 to C36+ hydrocarbons. In some embodiments, concentration data may be in the form of relative data (e.g., as a ratio of compositions of different hydrocarbons) or may be Petition 870260016305, dated 23 / 02 / 2026, page 25 / 56 16 / 22 normalized otherwise. Reservoir fluid property data may optionally also include concentrations of one or more other constituents within the well.

[0083] Reservoir fluid property data may include one or more properties derived from the reservoir fluid sample. Such derived properties may include the target property to be determined by the machine learning algorithm, for example, a gas-to-oil ratio in this case. Other derived properties may include a fluid density.

[0084] Reservoir fluid property data are sometimes referred to as PVT data, as it is commonly obtained in a pressure-volume-temperature (PVT) laboratory, where researchers will employ various instruments to determine reservoir fluid behavior and reservoir sample properties.

[0085] Input dataset 102 further comprises standard mud gas data measured for each PVT sample at the same reservoir depth. The measured standard mud gas data comprises hydrocarbon composition data measured for gas released from the drilling fluid at the sample location.

[0086] It will be appreciated that there is a time lag between the drill bit passing through the sample location and when the mud reaches the surface and is analyzed. However, workers in this field will be familiar with the procedures for calculating the time lag to determine the depth to which the mud gas sample corresponds. Therefore, this will not be discussed in detail.

[0087] Composition data for mud gas preferably comprise data for at least C1 to C4 hydrocarbons and preferably at least C1 to C5 hydrocarbons (as is the case in the present example). In some cases, concentrations for up to C7 or more hydrocarbons may be included.

[0088] Composition data can be stored as a measurement Petition 870260016305, dated 23 / 02 / 2026, page 26 / 56 17 / 22 directly from the concentration (e.g., measured in ppm or similar units), or alternatively as a relative concentration (e.g., as a proportion of another hydrocarbon, usually Ci). In some embodiments, the composition data may be normalized.

[0089] The measured standard sludge gas data are raw sludge gas data, i.e., they have not been corrected for recycling or extraction efficiency. This is important because using raw sludge gas data will allow the subsequent model to be used more broadly where advanced sludge gas data are not available.

[0090] Input dataset 102 further comprises petrophysical data measured for each PVT sample at the same reservoir depth. The petrophysical data may comprise any one or more of: apparent density, neutron porosity, resistivity data, acoustic data, natural gamma rays, nuclear magnetic resonance data, as well as time delay and gamma ray spectroscopy data from pulsed neutron measurements and the like.

[0091] The input dataset 102 comprises target data and input data for each sample that passed screening. The target data corresponds to the desired output of the model. The input data corresponds to the data that will be entered into the eventual model.

[0092] The target data in this example is a gas-oil ratio, and in this example, it is the single flash gas-oil measurement of the sample. As discussed above, this data is stored as part of the reservoir properties data within the initial dataset. Alternatively, other gas-oil ratio measurements could be used, or a gas-oil ratio could be derived from reservoir composition data, i.e., based on the concentrations of the various hydrocarbons.

[0093] The input data are standard mud gas data, i.e., data indicative of the composition of gases released from the drilling fluid from the sample location and at least one type of petrophysical data, for example, Petition 870260016305, dated 23 / 02 / 2026, page 27 / 56 18 / 22 apparent density and neutron porosity.

[0094] As mentioned above, the measured sludge gas data comprise raw sludge gas data, i.e., they have not been corrected for recycling or extraction efficiency.

[0095] Although it is not possible to apply a recycling correction after data collection, nor is it possible to account for the lack of heating (if heating was not used), it may be possible to apply a retrospective extraction efficiency correction. This is because the drilling mud composition for a specific well is generally known, and the extraction efficiency correction factors for that specific drilling fluid can be estimated from EOS simulation or approximated experimentally. The results show that the temperature-dependent extraction efficiency correction far outperforms the recycling corrections.

[0096] Consequently, the mud gas data used for input dataset 102 preferably comprise standard mud gas data where an extraction efficiency correction has been applied.

[0097] Next, a model generation is performed, in which a model is generated and validated based on the input dataset 102.

[0098] The input dataset 102 is first divided into a training dataset 104 and a test dataset 106. The input dataset 102 is preferably curated so that at least the test dataset 106 contains data that span the various classes of the input dataset 102 as a whole (e.g., dry gas reservoirs, wet gas reservoirs, oil reservoirs).

[0099] Typically, at least 50% of the input dataset 102 should be used for training and at least 10% of the input dataset 102 should be used for testing. Common ratios include 50:50, 70:30, 75:25, 80:20, 90:10. However, it will be appreciated that other splits may be used instead.

[0100] Generally, the larger the training dataset, the more Petition 870260016305, dated 23 / 02 / 2026, page 28 / 56 19 / 22 The model will be accurate. However, if a test dataset is too small (or even if no test dataset is used), then it is not possible to confidently verify the model's accuracy, for example, making it difficult to detect an overfitted model (only accurate for the specific training data).

[0101] To generate a model, a machine learning algorithm is provided with the training dataset 104 and a set of training parameters to control the machine learning algorithm.

[0102] In one example, Gaussian Process Regression and Random Forest were considered to be the best performing models. However, it will be appreciated that any suitable algorithm can be used, such as Universal Kriging, KMean, or Elastic Net algorithms. Those operating within this field will be familiar with the procedures for selecting and using a machine learning algorithm. Therefore, this will not be discussed in detail.

[0103] Model validation 108, for example, cross-validation, can then be performed. During model validation 108, the model is tested to determine how well it predicts new data that were not used in the model estimation, in order to flag problems such as overfitting or selection bias. Model validation 108 is an optional step.

[0104] Cross-validation involves partitioning the training dataset 104 into complementary subsets, performing model fitting using one subset of the training dataset 104, and validating the analysis on the other subset of the training dataset 104. To reduce variability, most methods use multiple rounds of cross-validation, performed using different partitions, and the validation results are combined (e.g., by averaging) across the rounds to give an estimate of the model's predictive performance (e.g., a mean prediction error, MAPE).

[0105] In this example, K-fold cross-validation and, in particular, 4-fold cross-validation is used. In K-fold cross-validation, the training data 104 Petition 870260016305, dated 23 / 02 / 2026, p. 29 / 56 20 / 22 are separated into disjoint subsets K (in this case, four), known as folds. Then, cross-validation is performed by training the model on all data except one fold, and validating the trained model using the fold that was not used for training. The best model is then selected as the model with the best predictive performance, for example, the lowest MAPE.

[0106] A first test stage 110 is then carried out, in which the model is tested using the training dataset 104 as a whole.

[0107] A second test stage 112 is then performed, in which the model is tested using the test dataset 106. As discussed earlier, this is a curated dataset that is broadly representative of the data as a whole and was not used during model generation.

[0108] It was found that the model predicts a reservoir fluid gas-oil ratio based on standard Ci to C5 mud gas data and petrophysical data with MAPE that is close to that achieved using a model based on advanced Ci to C5 mud gas data.

[0109] Understanding the quality of measured slurry gas data is important before making a fluid property prediction (e.g., gas-oil ratio) because the quality of the slurry gas data will significantly affect the accuracy of the prediction. The following characteristics of slurry gas data values ​​have been identified as indicating poor quality or unreliable data: • Large fluctuations of a component within a small depth range. • Initial observations after missing measurements. • C1 content below a certain limit. • C4 or C5 content below a certain limit.

[0110] To quantify the quality of slurry gas data, the inventors derived a quality control metric (QC metric) that ranged from 0 to 1. High-quality slurry gas data would have a QC metric value close to 1. If one or more of the above factors are found, the QC metric will be reduced. Petition 870260016305, dated 23 / 02 / 2026, p. 30 / 56 21 / 22 Low-quality mud gas data were indicated by a CQ metric close to 0. A single numerical quality measure between 0 and 1 can be plotted side-by-side with a predicted fluid property profile (as will be discussed below) to visualize the confidence level associated with each prediction, based on the quality of the mud gas data.

[0111] Samples with a higher CQ metric match closely, while samples with a lower CQ metric have little match. Thus, these factors provide a useful indication of the accuracy of a gas-oil ratio forecast.

[0112] Mud gas data and petrophysical data are generated continuously during the drilling process. Therefore, by applying the machine learning model to mud gas data and petrophysical data, it is possible to provide, at an early stage of the well installation procedure, a continuous profile for the wellbore of the predicted reservoir property, for example, gas-oil ratio or fluid density. This is something that was not possible previously until much later in the process.

[0113] Although the above examples have been described in the context of a gas-to-oil ratio as the target reservoir fluid property, the same technique can also be employed to create a model to estimate other reservoir fluid properties of the reservoir fluid at a sample location, based on measured mud gas data. Exemplary reservoir fluid properties include a reservoir fluid density, a stock tank oil density or a live reservoir density, a reservoir fluid saturation pressure, and a reservoir fluid formation volume factor.

[0114] Furthermore, a similar technique can be used to train a model to estimate the reservoir fluid composition and corresponding C7+ fraction properties. This is advantageous because this information can be used for an equation of state (EOS) model calculation. The EOS model for a particular fluid is an expression that describes the relationship between pressure, Petition 870260016305, dated 23 / 02 / 2026, page 31 / 56 22 / 22 temperature and volume of the fluid and can be used to predict the phase behavior of the fluid in order to derive additional properties of the same.

[0115] It is generally considered necessary to know at least the following fluid properties in order to determine the equations of state:

[0116] The absolute composition of each of the C1 to C6 hydrocarbons and the absolute composition of the combined C7+ hydrocarbons;

[0117] The average hydrocarbon density of C7+ hydrocarbons; and

[0118] The average molecular weight of C7+ hydrocarbons.

[0119] When determining equations of state for a fluid, C7+ hydrocarbons are usually grouped together because these hydrocarbons generally remain in the liquid / oil phase. A standard C7+ characterization method can divide C7+ into multiple pseudo-components for EOS calculation.

[0120] Although models of individual fluid properties (such as density and GOR) were developed in the first examples, it will be appreciated that a physical model could be generated that would calculate all fluid properties. The EOS model approach in the second example demonstrates a good solution for predicting all reservoir fluid properties.

[0121] Although preferred embodiments have been described above, it will be appreciated that these have been provided by way of example only, and the scope of the invention should be limited only by the following claims. Petition 870260016305, dated 23 / 02 / 2026, p. 32 / 56

Claims

1 / 3 CLAIMS 1. A method for generating a model to predict at least one property of a fluid at a sample location within a hydrocarbon reservoir, characterized in that it comprises: providing a training dataset comprising input data and target data, the input data comprising mud gas data and petrophysical data for each of a plurality of sample locations, and the target data comprising at least one fluid property for each of the plurality of sample locations; and generating a model using the training dataset such that the model can be used to predict at least one fluid property at the sample location based on measured mud gas data and measured petrophysical data for the sample location, wherein a drilling fluid recycling correction has not been applied to the mud gas data.

2. A method according to claim 1, characterized in that the model generation comprises instructing a machine learning algorithm to generate the model using the training dataset.

3. A method according to any one of claims 1 or 2, characterized in that at least one property comprises a property influenced by the oil-related components of the fluid.

4. A method, according to any one of claims 1 to 3, characterized in that at least one property comprises one or more of: a fluid density at the sample location; a gas-oil ratio of the fluid at the sample location; a saturation pressure of the fluid at the sample location; a fluid formation volume factor at the sample location; a concentration of C₆⁺ hydrocarbons within the fluid at the sample location. Petition 870220123828, dated 12 / 30 / 2022, pp. 33 / 38 2 / 3 5. A method, according to any one of claims 1 to 4, characterized in that the mud gas data of the training dataset comprise standard mud gas data measured for the sample location.

6. Method according to claim 5, characterized in that an extraction efficiency correction was applied to the sludge gas data of the training dataset.

7. A method according to claim 5, characterized in that an extraction efficiency correction was not applied to the mud gas data of the training dataset, and wherein the training data comprises drilling mud composition data.

8. Method, according to any one of claims 5 to 7, characterized in that the measured mud gas data were collected without the use of heating.

9. A method, according to any one of claims 1 to 8, characterized in that the petrophysical data comprise one or more of: apparent density; neutron porosity; resistivity data; acoustic data; natural gamma rays; nuclear magnetic resonance data; and gamma-ray spectroscopy data.

10. A computer-based model for predicting at least one property of a fluid at a sample location within a hydrocarbon reservoir based on measured mud gas data and measured petrophysical data for that sample location, the computer-based model characterized in that it was generated by a method as defined in any one of claims 1 to 9.

11. Tangible computer-readable medium characterized in that Petition 870220123828, dated 12 / 30 / 2022, page 34 / 38 3 / 3 stores a computer-based model, as defined in claim 10.

12. A method for predicting the value of a fluid property at a sample location within a hydrocarbon reservoir, the method characterized in that it comprises: receiving measured mud gas data and measured petrophysical data for the sample location; and predicting the value of the fluid property at the sample location by providing the measured mud gas data and the measured petrophysical data to a computer-based model, as defined in claim 10.

13. A method for predicting a fluid property value of a fluid along a well length through a hydrocarbon reservoir, characterized in that it comprises: predicting a fluid property value of a fluid at a plurality of sample locations along a well length using a method as defined in claim 12 for each sample location.

14. Method according to claim 13, characterized in that it further comprises: displaying, using an electronic display screen, a graph representing the predicted values ​​of the fluid property against a location of the respective sample site for each of the plurality of sample sites along the length of the well. Petition 870220123828, dated 12 / 30 / 2022, pp. 35 / 38

Images