WO2024136662A1

WO2024136662A1 - Method for predicting a fluid type of a reservoir fluid

Info

Publication number: WO2024136662A1
Application number: PCT/NO2023/060111
Authority: WO
Inventors: Tao Yang
Original assignee: Equinor Energy As
Priority date: 2022-12-22
Filing date: 2023-12-04
Publication date: 2024-06-27
Also published as: GB202219542D0; GB2625748A

Abstract

Method of generating a model for predicting a fluid type of a reservoir fluid at a sample location within a hydrocarbon reservoir. The method comprises: providing a training data set comprising input data and target data, the input data comprising mud-gas data and/or PVT data, and geospatial data, for each of a plurality of sample locations in a field comprising the hydrocarbon reservoir and/or in a nearby field, and the target data comprising a fluid type for each of the plurality of sample locations; and instructing a machine learning algorithm to generate a model using the training data set such that the model can be used to predict the fluid type of the reservoir fluid at the sample location based on measured mud-gas data and geospatial data for the sample location, wherein a drilling fluid recycling correction has not been applied to the mud-gas data.

Description

Method for predicting a fluid type of a reservoir fluid

Technical Field

The present invention relates to a method for predicting a fluid type of a target reservoir fluid.

Background

Drilling fluid is a fluid used to aid the drilling of boreholes into the earth. The main functions of drilling fluid include providing hydrostatic pressure to prevent formation fluids from entering into the well bore, keeping the drill bit cool and clean during drilling, carrying out drill cuttings, and suspending the drill cuttings while drilling is paused and when the drilling assembly is brought in and out of the hole.

Drilling fluids are broadly categorised into water-based drilling fluid, non-aqueous drilling fluid, often referred to as oil-based drilling fluid, and gaseous drilling fluid. Liquid drilling fluids, i.e. water-based drilling fluid or non-aqueous drilling fluid, are commonly referred to as “drilling mud”.

Mud-gas logging entails gathering data from hydrocarbon gas detectors that record the levels of gases brought up to the surface in the drilling mud during a bore drilling operation.

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

There are two types of mud-gas data that can be collected, which are sometimes referred to a “standard” mud-gas logging, and “advanced” mud-gas logging. The equipment for standard mud gas logging and advanced mud gas logging are different. For a standard mud gas system, the degasser does not usually have heating or use constant volume gas separation. There is also only one mud sampling point (“out”) and therefore it is not suitable for recycling correction. The measured gas composition is usually referred to as standard mud-gas data, which is not directly comparable to the actual Ci to Cs composition of the reservoir fluid sample.

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

When generating advanced mud-gas data, in order to make the data closely correspond to the actual reservoir fluid Ci to Cs concentrations, two correction processes are applied to the “raw” mud-gas data from the advanced mud gas logging system.

Firstly, a recycling correction is made to eliminate contamination of the gas by gases originating from previous injections of the drilling mud. This correction is applied based on a separate mud-gas measurement that was taken before the drilling mud was injected into the drilling string.

Secondly, an extraction efficiency correction step is applied to increase the concentration of intermediate components (from C2 to Cs), such that the concentration of these components, relative to the Ci concentration, more closely resemble 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 borehole.

In the past, the advanced mud gas data has been examined to estimate certain fluid properties of the reservoir fluid using broad, empirical correlations between the advanced mud-gas composition and certain fluid properties of the reservoir fluid. For example, extremely dry gas reservoirs should comprise mostly Ci and not much C2+, e.g. with C1/C2 ratios being greater than 50. Wet gas reservoirs will often have ratios between 15 and 50, and oil reservoirs will have ratios between 2 and 15. These empirical corrections are known to be highly inaccurate, particularly close to the range boundaries, and so were rarely relied upon in isolation, but rather were used to guide where to take downhole fluid samples. However, more recently, an advanced machine learning model has been developed that has made it possible to predict reservoir fluid properties much more accurately from the advanced mud-gas data.

Details of how such a machine learning model was trained to determine a gas-oil ratio of the reservoir fluid based on the advanced mud-gas data can be found in the paper 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

Advantageously, this model can be used to generate a substantially continuous log of the respective reservoir fluid property. This was not previously possible, and in the past, it was necessary to rely on downhole fluid samples taken at discrete intervals. Furthermore, the model allows reservoir fluid property predictions to be made at a very early stage of the drilling process and without needing to interrupt the drilling process, as might be required to take downhole fluid samples or the like.

This model has been found to be very useful, but is limited in that it requires the availability of advanced mud-gas data. Whilst the collection of advanced mud-gas data is less costly than some techniques, such as downhole fluid sampling, it still adds additional costs to the drilling process.

A need exists for a technique that can be used when advanced mud-gas data is not available.

Summary

It is an object of the present invention to overcome or at least mitigate the problems identified above.

In accordance with a first aspect of the present invention there is provided a method of generating a model for predicting a fluid type of a reservoir fluid at a sample location within a hydrocarbon reservoir. The method comprises providing a training data set comprising input data and target data, the target data comprising a fluid type for each of the plurality of sample locations, and the input data comprising mud-gas data and/or PVT data, and geospatial data, for each of a plurality of sample locations in a field comprising the hydrocarbon reservoir and/or in a nearby field. The method further comprises instructing a machine learning algorithm to generate a model using the training data set such that the model can be used to predict the fluid type of the reservoir fluid at the sample location based on measured mud-gas data and geospatial data for the sample location, wherein a drilling fluid recycling correction has not been applied to the mud-gas data.

The input data of the training data set may further comprise petrophysical data.

The plurality of sample locations may be a subset of a larger number of sample locations, wherein the subset is selected based on petrophysical data obtained for the larger number of sample locations.

The petrophysical data may comprise one or more of: bulk density; neutron porosity; resistivity data; acoustic data; natural gamma ray; nuclear magnetic resonance data; and gamma ray spectroscopy data.

The training data set may comprise data measured from one or more existing wells in the field comprising the hydrocarbon reservoir and/or the nearby field.

The measured mud-gas data may be measured during drilling of a well.

The mud-gas data of the training data set may comprise measured standard mud-gas data for each of the plurality of sample locations.

A simulated extraction efficiency correction may have been applied to the mud-gas data of the training data set.

An extraction efficiency correction may not have been applied to the mud-gas data of the training data set, and the training data may comprise drilling mud compositional data. The measured mud-gas data may be collected without recycling corrections.

The plurality of sample locations may be in one or more existing wells extending into a subsurface formation, and the sample location within the hydrocarbon reservoir may be in a newer well that is different from the existing wells and that extends into a subsurface formation, wherein the plurality of sample locations may be selected based on a similarity between properties of the subsurface formation into which the newer well extends and properties of the subsurface formation into which the one or more existing wells extend, and/or a similarity between properties of fluids produced from said subsurface formations.

In accordance with a second aspect of the present invention there is provided a method of predicting a fluid type of a reservoir fluid along a length of a well through a hydrocarbon reservoir, the method comprising: predicting a fluid type of a reservoir fluid at a plurality of sample locations along a length of a well using a method according to the first aspect.

In accordance with a third aspect of the present invention there is provided a computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method of the first or second aspect.

In accordance with a fourth aspect of the present invention there is provided a computer-readable storage medium comprising the computer program of the third aspect.

In accordance with a fifth aspect of the present invention there is provided a computing device configured to execute the computer program of the third aspect.

Embodiments of the invention will now be described by way of example only and with reference to the accompanying drawings, in which:

Brief Description of the Drawings

Figure 1 shows a flow diagram illustrating a method in accordance with the invention. Figure 2 shows a schematic illustration of a mud-gas analysis tool.

Detailed Description

Existing techniques for predicting or identifying a fluid type of a reservoir fluid have used empirical correlations between properties of the reservoir fluid, e.g. the correlation between a gas-oil ratio and a C1/C2 ratio of a sample of the reservoir fluid, where the term gas-oil ratio refers to the ratio of the volume of gas that comes out of solution at surface conditions to the volume of oil.

However, a large proportion of reservoirs have fluid compositions in a C1/C2 ratio range in which there is significant overlap between gas and oil fluid samples, i.e. in a range within which it is not possible to confidently distinguish between oil and gas fluid samples based only on the C1/C2 ratio. This means that the existing empirical correlations typically cannot be relied upon for accurate fluid type prediction.

Geochemical parameters other than the C1/C2 ratio can be used to distinguish between gas phase and oil phase fluids, and similar empirical correlations exist for many of these parameters, but these parameters have similar degrees of uncertainty. Such geochemical parameters include the Bernard parameter (Ci I C2+C3), the balance ratio (C1+C2/ C3+C4+C5), the wetness ratio (C2+C3+C4+C5/ C1+C2+C3+C4+C5), the dryness ratio (Ci I C1+C2+C3+C4+C5), and the hydrocarbon character (C4+C51 C3).

In the above-referenced existing techniques, when standard mud-gas data was to be examined, multiple geochemical parameters would be calculated and each compared to the respective empirical threshold. An estimation of whether a particular reservoir contains oil or gas would then be made based on what the majority of geochemical parameters indicated. This would typically achieve about 50% to 60% accuracy in identifying the reservoir fluid type. Whilst this provided a useful indicator, it was not sufficiently precise to be confidently relied upon.

The present invention provides a method of generating a model for predicting a fluid type of a reservoir fluid at a sample location within a hydrocarbon reservoir. The method comprises providing a training data set comprising input data and target data, the target data comprising a fluid type for each of the plurality of sample locations, and the input data comprising mud-gas data and/or PVT data, and geospatial data, for each of a plurality of sample locations in a field comprising the hydrocarbon reservoir and/or in a nearby field. The method further comprises instructing a machine learning algorithm to generate a model using the training data set such that the model can be used to predict the fluid type of the reservoir fluid at the sample location based on measured mud-gas data and geospatial data for the sample location, wherein a drilling fluid recycling correction has not been applied to the mud-gas data.

The inventors have realised that using standard mud-gas data and/or PVT data, and corresponding geospatial data, measured from one or more existing wells in one or more fields to train a machine learning algorithm enables a fluid type to be predicted more accurately. In particular, the machine learning algorithm is able to generate a model that more accurately predicts the fluid type of a reservoir fluid, where the reservoir fluid is obtained from a sample location in a reservoir that is located in the above-referenced one or more fields or a nearby field, and the model can be applied using standard mud-gas data and geospatial data measured from the sample location. The present method may therefore be particularly useful in mature fields. Nothing in the existing state of the art suggested that such a surprising improvement in accuracy in predicting a fluid type could be achieved by combining geospatial information with standard mud-gas data and/or PVT data for training the model, enabling an accurate prediction of fluid type using only measured standard mud gas data and geospatial data when the model is applied in use.

Using the present method, it is possible to achieve a higher accuracy when determining the fluid type of a target reservoir fluid, using standard mud-gas data (instead of advanced mud-gas data or PVT data). Testing by the inventors indicates that this technique can achieve an accuracy higher than 80% using standard mud-gas data and geospatial data, when applied to appropriate reservoirs.

This technique is advantageous because it can be performed using standard mud-gas data collected when drilling a new well. Not only is this significantly cheaper than other fluid analysis techniques, but it also does not require interruption of the drilling process. Furthermore, because mud-gas data is collected as a substantially continuous log along the length of the well, it is possible to create a substantially continuous reservoir fluid type log along the length of the well. This is not possible using techniques such as downhole fluid analysis, which only identify the fluid composition at a relatively small number of target locations along the length of the well.

By using mud-gas data, the reservoir fluid type can be identified in real-time as the well is drilled. Therefore, it can be used for geosteering, which is the process of adjusting the borehole position (such as inclination and azimuth angle) as the borehole is drilled in order to reach one or more geological targets.

The output of the present method can also inform decisions (including in real-time) regarding completion, e.g. how to place and complete the well, including where to perforate the well casing.

Further, whilst advanced mud-gas analysis is comparatively cheap compared to collecting a large number of downhole fluid samples and subsequently measuring PVT data, it does still represent a significant additional cost compared to collecting only standard mud-gas data. Therefore, the fact that the analysis of the present method can be performed using only standard mud-gas data is highly advantageous.

The present method uses a machine learning algorithm to produce a model for predicating a fluid type of a reservoir fluid, based on at least standard mud-gas data and geospatial data.

The geospatial data comprises positional information indicating a location at which a fluid sample is or has been collected. In particular, the geospatial information comprises longitude and latitude information indicating a position on the Earth’s surface, and depth information indicating a depth below the Earth’s surface. In an embodiment, the geospatial data comprises a definition of a 3D volume containing the location at which the fluid sample is or has been collected. In an embodiment, the depth information comprises a depth range of e.g. 20 m, 10 m, or 5 m.

Each PVT log entry, mud gas sample, and/or petrophysical log entry therefore has corresponding geospatial information. A PVT log entry, mud gas sample, and/or petrophysical log entry may share the same geospatial information. The geospatial data is collected using methods known in the art. For example, where a downhole tool is used to collect a fluid sample for subsequent analysis (e.g. to determine PVT data), the downhole tool may include position-determining means so that positional information (including at least depth information) indicating the location at which the fluid sample is collected can be recorded in ‘real time’, when the sample is collected. Alternatively, during well drilling operations, petrophysical logs and fluid samples for standard mud gas analysis will typically be collected at regular intervals as drilling progresses (e.g. every metre for petrophysical logs, and every 3 m for mud gas samples), and the drilling time and drilling speed can be used to later determine the position at which each petrophysical log and/or fluid sample for mud gas analysis was collected.

Figure 1 illustrates a workflow 100 for training the machine learning algorithm to generate a model for predicting a fluid type of a reservoir fluid, where the model is trained using geospatial data and at least one of standard mud-gas data and PVT data, and optionally petrophysical data; and the model is for use in predicting the fluid type based on measured standard mud-gas data and geospatial data, and optionally petrophysical data.

In the embodiment illustrated in Figure 1, an initial data set 102 (shown as “input data set” in Figure 1) comprises reservoir fluid properties data from a large number of reservoir fluid samples. Reservoir samples may be obtained, for example, by downhole fluid sampling. However, other techniques could also be used, for example by taking a sample of well fluid after the well has been completed.

The reservoir fluid properties data should include at least hydrocarbon composition data, which may be either in the form of direct measurements of the concentration of each hydrocarbon component within the sample, typically covering Ci to C36+ hydrocarbons. In some embodiments, the concentration data may be in the form of relative data (e.g. as a ratio of compositions of different hydrocarbons) or may be otherwise normalised. The reservoir fluid properties data may optionally also include concentrations of one or more other constituents within the well.

The reservoir fluid properties data may include one or more derived properties of the reservoir fluid sample. Such derived properties include the target property to be determined by the machine-learning algorithm, e.g. a gas-oil ratio in this case, where the gas-oil ratio is used to determine the fluid type. A typical example of a threshold gas-oil ratio value (representing the threshold between a fluid that is determined to be gas and a fluid that is determined to be oil) is about 600 Sm3/Sm3. Other derived properties may include a density of the fluid.

The reservoir fluid properties data is 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 behaviour and properties from the reservoir samples.

In the embodiment illustrated in Figure 1, the initial data set 102 further comprises measured standard mud-gas data and geospatial data for each reservoir fluid sample. The measured standard mud-gas data comprises measured hydrocarbon composition data for gas released from the drilling fluid from the sample location.

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

Due to the low reading of C4 and C5 components, the mud-gas composition data typically comprises data for C1 to C3 only. However, C4, C5 and/or other higher components may be included.

The composition data may be stored either as a direct measurement of 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 normalised.

The measured standard mud-gas data is “raw” mud-gas data, i.e. it has not been corrected for recycling or extraction efficiency. This is important as the use of “raw” mud-gas data will allow the subsequent model to be utilised more widely, where advanced mud-gas data is not available. In an embodiment, the initial data set 102 further comprises measured petrophysical data for each PVT sample at the same reservoir depth. The petrophysical data comprises any one or more of: bulk density, neutron porosity, resistivity data, acoustic data, natural gamma ray, nuclear magnetic resonance data, as well as slowing down time and gamma ray spectroscopy data from pulsed neutron measurements, and the like.

In an embodiment, the measured petrophysical data is used to select a subset of sample locations for which corresponding data will be used as input data and target data for the machine learning algorithm to generate the model. In particular, petrophysical logs are used to identify reservoir zones containing oil and/or gas, e.g. hydrocarbon bearing sands, within a given well, and then the subset of sample locations are selected as locations within one or more wells that are within such a reservoir zone. In this way, the data used to generate the model is limited to data corresponding to locations in a well where hydrocarbons are present. It is noted that when the model is used in practice, petrophysical data measured from the newer well may be included in the input data upon which the model is run.

The initial data set 102 comprises target data and input data for each of the plurality of sample locations. Where petrophysical data has been used to ‘screen’ the plurality of sample locations, the target data and input data are from a subset of sample locations that is smaller than the more numerous, total plurality of sample locations in the initial data set. The plurality of sample locations are in one or more existing wells in a field. The target data corresponds to the desired output of the model. In some embodiments, the input data corresponds directly to the data that will be input into the eventual model. However, this is not strictly necessary. The model may be trained using types of data that are different from the type(s) of data that will be input into the model when it is applied in practice. It is envisaged that the data that will be input into the eventual model will be measured from one or more locations in a newer well in the same field in which the one or more existing wells are located, or a nearby field. That is, historical data from existing wells is used to train the model that will then be applied to predict the fluid type in a newer well, based on standard mud gas data and geospatial data measured in the newer well, e.g. during drilling. In an embodiment, the data from existing wells that is used to train the model is selected based on a similarity between properties of the subsurface formation into which the newer well extends and properties of the subsurface formation into which the one or more existing wells extend, and/or a similarity between properties of fluids produced from said subsurface formations. The similarity is determined based on one or more of PVT logs, petrophysical logs, acoustic logs, and seismic data.

The target data in this example is a fluid type, i.e. oil or gas. As discussed above, this data is stored as part of the reservoir properties data within the initial data set. Alternatively, other measurements of gas-oil ratio may be used, or a gas-oil ratio may be derived from the reservoir composition data, i.e. based on the concentrations of the various hydrocarbons.

In the embodiment illustrated in Figure 1, the input data is, or comprises, standard mud-gas data, i.e. data indicative of the composition of gases released from the drilling fluid from the sample location, and geospatial data. PVT data and/or petrophysical data may also be included in the input data. Tests by the inventors have shown that the addition of each additional data set (i.e. PVT data and/or petrophysical data) results in a corresponding improvement in model accuracy.

In an alternative embodiment, the input data does not comprise standard mud gas data. In this embodiment the input data is, or comprises, PVT data.

As mentioned above, the measured mud-gas data comprises “raw” mud-gas data, i.e. it has not been corrected for recycling or extraction efficiency.

Whilst it is not possible to apply a recycling correction after collection of the data, nor is it possible to account for the lack of heating (if heating was not used), it may be possible to apply a simulated extraction efficiency correction, also known as a “pseudo” extraction efficiency correction in the art. The pseudo extraction efficiency correction factors for oil-based drilling fluid can be estimated either from EOS simulation or may be approximated by experiment. Results show that the temperature dependent extraction efficiency correction far outweighs recycling corrections. Consequently, the mud-gas data used for the initial data set 102 preferably comprises standard mud-gas data where an pseudo extraction efficiency correction has been applied when oil-based mud is used.

Next, a model generation is performed, in which a model is generated and validated based on the initial data set 102.

The initial data set 102 is first divided into a training data set 104, and a test data set 106. The initial data set 102 is preferably curated such that at least the test data set 106 contains data that spans the various classes of the initial data set 102 as a whole (e.g. dry gas reservoirs, wet gas reservoirs, oil reservoirs).

Typically, at least 50% of the initial data set 102 should be used for training, and at least 10% of the initial data set 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 divisions may be used instead. In an embodiment, the split of the initial data set 102 is based on a temporal aspect, to prevent a case in which data corresponding to newer wells is used as part of the training set. In particular, the data is split so that the test data set 106 consists of (or mainly comprises) data from newer wells, and the training data set 104 consists of (or mainly comprises) data from the older existing wells. A similar split is performed so that the validation data set 108 consists of (or mainly comprises) data from newer wells, and the train data 104 consists of (or mainly comprises) data from the older existing wells. This is to make the conditions under which the model is trained and tested similar to, or correspond to, the conditions under which the model will eventually be used (i.e. the model will be applied to a newer well, having been generated using data from older, existing wells).

Generally the larger the training data set, the more accurate the model will be. However, if too small a test data set is used (or indeed if no test data set is used) then it is not possible to confidently verify the accuracy of the model, e.g. making it difficult to detect an over-fitted model (only accurate for the specific training data).

To generate a model, a machine learning algorithm is provided with the training data set 104, and a set of training parameters to control the machine learning algorithm. In one example, XGBoost classifier, and Random Forest classifier were found to be best performing models. However, it will be appreciated that any suitable algorithm may be used, such as standardized logistic regressor or light GBM classifier. Those operating within this field will be familiar with the procedures for selecting and utilising a machine learning algorithm. Therefore, this will not be discussed in detail.

Model validation 108, e.g. train-test validation, may then be performed. During the model validation 108, the model is tested to determine how well it predicts new data that was not used in estimating the model, in order to flag problems such as features prone to over fitting or selection bias. Model validation is conducted using a generalized linear model i.e. standardized logistic regression, for first phase feature search. Further, model tuning is conducted using a classifier model ie. Random Forest classifier or XGBoost classifier

Train-test validation involves partitioning the training data set 104 into train and test data sets according to a temporal aspect as described above. , performing the model fitting using the train data set , and validating the analysis on the other subset of the training data set 104, the validation data set. To reduce variability, a generalized linear model is used first to identify features that may be prone to overfitting. Further the model is tuned using a classifier model. The validation results give an estimate of the model’s predictive performance (e.g. area under the curve, AUG) or accuracy. Further the generalized linear model is used return a probability of the predicted fluid type category given by the classifier model.

In this example a test-validation set is used in which the model validation is performed by testing the trained model using the validation set that was not used for training. The best model is then selected as the model having the best predictive performance, e.g. the highest AUG or accuracy.

A first testing step 110 is then performed, in which the model is tested using the training data set 104 as a whole.

A second testing step 112 is then performed, in which the model is tested using the test data set 106. As discussed previously, this is a curated set of data that is broadly representative of the data as a whole, takes in to account the drilling temporal aspect and was not used during the generation of the model.

The model has been found to predict a fluid type of the reservoir fluid based on standard mud-gas data and geospatial data with an accuracy or AUC of 80% that is close to that achieved using a model based on advanced mud-gas data.

Understanding the quality of the measured mud-gas data is important before performing a fluid type prediction because the mud-gas data quality will significantly impact prediction accuracy. The following characteristics of the mud-gas data values have been identified as indicating low quality or unreliable data:

• Large fluctuations of a component within a small depth range.

• First observations after missing measurements.

• Ci to C3 content below a given threshold.

To quantify the quality of the mud-gas data, the inventors derived a quality control metric (QC metric) which ranged from 0 to 1. High quality mud-gas data would have QC metric value close to 1. If one or more of the above factors are found, then the QC metric would be reduced. Low-quality mud-gas data was indicated by QC metric close to 0. A single numeric quality measure between 0 and 1 can be plotted side-by-side with a predicted fluid property log (as will be discussed below) to visualize the confidence level associated with each prediction, based on mud-gas data quality.

Samples having a higher QC metric correspond closely, whilst samples having a lower QC metric have poor correspondence. Thus, these factors provide a useful indication of the accuracy of a prediction of the reservoir fluid type (oil or gas).

Mud-gas data and geospatial data are both generated continuously during the drilling process. Therefore, by applying the machine learning model to the mud-gas data and geospatial data, it is possible to provide, at an early stage of the well installation procedure, a continuous log for the well bore of the predicted fluid type. This is something that has not been possible previously until much later in the process.

The mud-gas data may comprise a plurality of mud-gas data points, which may correspond to target reservoir fluids at a plurality of locations along a length of the well. Identifying the fluid type may then comprise identifying a fluid type of the target reservoir fluid at each location along the length of the well. Thus, the method can be used to generate a fluid type log along a length of the well.

The method described above is preferably performed by a computer program operating on a computer.

The mud-gas data used in the present method is obtained using any suitable mud-gas analysis tool, e.g the mud-gas analysis tool 20 shown schematically in Figure 2.

The tool 20 is coupled to a flow line 10 containing drilling mud returned from a borehole of a well. The drilling mud may be water-based mud or oil-based mud.

The tool 20 comprises a sampling probe 22 disposed with respect to the flow line 10 so as to collect a sample 24 of the drilling mud from the flow line 10. The drilling mud sample 24 is preferably a continuous sample, i.e. such that a portion of the flow of drilling mud within the flow line 10 is diverted through the mud-gas analysis tool 20.

The drilling mud sample 24 is supplied to a gas-separation chamber 26 where at least a portion of the gas carried by the drilling mud is released. The sample of drilling mud 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, for standard mud-gas data, the mud sample 24 is not heated with the temperature typically ranging from 10°C to 60°C. However, in some implementations, heating is used to heat the drilling mud to around 80°C to 90°C.

The released gas 30 is directed from the separation chamber 26 to a gas analysis unit (not shown), while the degassed mud 32 is returned to the flow line 10 or to another location for re-use.

The gas analyser may comprise a total gas detector, which may provide a basic quantitative indication as to how much gas is being extracted from the drilling mud by the tool 20. Total gas detection typically incorporates either a catalytic filament detector, also called a hotwire detector, or a hydrogen flame ionization detector. The gas analysis device may additionally or alternatively comprise an apparatus for detailed analysis of the hydrocarbon mixture. This analysis is usually performed by a gas chromatograph. However, several other detecting devices may also be utilised including a mass spectrometer, an infrared analyser or a thermal conductivity analyser.

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

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 that 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 separation chamber 26 to the gas analysis unit. Additionally, in the case of a gas analyser comprising a combustor, the use of air as the carrier gas may provide the necessary oxygen for combustion.

Commonly, mud-gas data provides a concentration for each of the Ci , C2, C3, iC4, nC4, iCs, and nCs hydrocarbon gases.

In some arrangements, the tool 20 may be configured to detect and/or remove H2S from the gas to prevent adverse effects that could influence hydrocarbon detection.

In some embodiments, non-combustibles gases, such as helium, carbon dioxide and nitrogen, can be detected by the gas analyser in conjunction with the logging of hydrocarbons.

It will be appreciated by the person of skill in the art that various modifications may be made to the above described embodiments without departing from the scope of the present invention.

Claims

1. A method of generating a model for predicting a fluid type of a reservoir fluid at a sample location within a hydrocarbon reservoir, comprising: providing a training data set comprising input data and target data, the input data comprising mud-gas data and/or PVT data, and geospatial data for each of a plurality of sample locations in a field comprising the hydrocarbon reservoir and/or in a nearby field, and the target data comprising a fluid type for each of the plurality of sample locations; and instructing a machine learning algorithm to generate a model using the training data set such that the model can be used to predict the fluid type of the reservoir fluid at the sample location based on measured mud-gas data and geospatial data for the sample location, wherein a drilling fluid recycling correction has not been applied to the mud-gas data.

2. The method of claim 1, wherein the input data of the training data set further comprises petrophysical data.

3. The method of claim 1 or claim 2, wherein the plurality of sample locations are a subset of a larger number of sample locations, wherein the subset is selected based on petrophysical data obtained for the larger number of sample locations.

4. The method of claim 2 or 3, wherein the petrophysical data comprise one or more of: bulk density; neutron porosity; resistivity data; acoustic data; natural gamma ray; nuclear magnetic resonance data; and gamma ray spectroscopy data.

5. The method of any one of the preceding claims, wherein the training data set comprises data measured from one or more existing wells in the field comprising the hydrocarbon reservoir and/or the nearby field.

6. The method of any one of the preceding claims, wherein the measured mudgas data is measured during drilling of a well.

7. The method of any one of the preceding claims, wherein the mud-gas data of the training data set comprises measured standard mud-gas data for each of the plurality of sample locations.

8. The method of claim 7, wherein a simulated extraction efficiency correction has been applied to the mud-gas data of the training data set.

9. The method of claim 7, wherein an extraction efficiency correction has not been applied to the mud-gas data of the training data set, and wherein the training data comprise drilling mud compositional data.

10. The method of any one of claims 7 to 9, wherein the measured mud-gas data is collected without recycling corrections.

11. The method of any one of the preceding claims, wherein the plurality of sample locations are in one or more existing wells extending into a subsurface formation, and the sample location within the hydrocarbon reservoir is in a newer well that is different from the existing wells and that extends into a subsurface formation, wherein the plurality of sample locations are selected based on a similarity between properties of the subsurface formation into which the newer well extends and properties of the subsurface formation into which the one or more existing wells extend, and/or a similarity between properties of fluids produced from said subsurface formations.

12. A method of predicting a fluid type of a reservoir fluid along a length of a well through a hydrocarbon reservoir, the method comprising: predicting a fluid type of a reservoir fluid at a plurality of sample locations along a length of a well using a method according to any one of the preceding claims.

13. A computer program comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method of any one of the preceding claims.

14. A computer-readable storage medium comprising the computer program of claim 13.

15. A computing device configured to execute the computer program of claim 13.