RU2773670C1 - Method (variants), system and machine-readable medium for determining the proportion of reservoir fluid in a fluid mixture - Google Patents

Abstract

FIELD: oil and gas industry.

SUBSTANCE: group of inventions relates to the oil and gas industry and can be used for separate accounting of products during joint operation of several reservoirs. To implement a method for determining the proportion of reservoir fluid in a mixture of fluids, at least one sample of individual reservoir fluid is obtained from at least two different layers. Calibration mixtures are prepared by mixing individual reservoir fluids of different layers in different ratios. The hydrocarbon composition of the obtained calibration mixtures is determined. Correlations are established between the content of individual reservoir fluids in calibration mixtures and the obtained data on the hydrocarbon composition of calibration mixtures using the multidimensional regression method. A sample of the studied fluid is obtained and its hydrocarbon composition is determined. The proportions of each individual reservoir fluid are determined based on the data obtained on the hydrocarbon composition of the sample of the fluid under study and on the established relationship between the content of individual fluids in calibration mixtures and the hydrocarbon composition of calibration mixtures. According to the second option, at least three samples of individual reservoir fluid are obtained. A system and a machine-readable carrier for determining the proportion of reservoir fluid in a fluid mixture is proposed.

EFFECT: increase in the accuracy of determining the content of the fraction of each reservoir fluid in a mixture of two or more fluids, as well as an increase in the reliability of the distinguishability of hydrocarbon fluids in a mixture that belong to different formations/deposits.

51 cl, 11 dwg, 6 tbl

Description

The invention relates to the oil and gas industry, in particular, to a method for determining the content of hydrocarbon fluids in mixtures and can be used for separate accounting of products in the joint operation of several reservoirs.

Information about the content of various hydrocarbon fluids in complex mixtures, including oil and condensate, is used both to assess changes in the properties of reservoir fluids during production in order to control the depletion of reserves of hydrocarbon deposits, and during the study of various processes occurring in formations when exposed to them, for example, to increase oil recovery. In the case of simultaneous production of fluids from several reservoirs in the section of one well, it becomes an important issue to take into account the contribution of each of the fluids to the resulting production.

The determination of the proportion of fluids from different reservoirs in a mixture is based on the difference in their hydrocarbon composition. The determined composition of hydrocarbons in formation fluid samples allows identification of a separate fluid taken from a specific oil and gas bearing formation, since oils from different reservoirs (or individual sections of a complex reservoir), as a rule, have measurable systematic differences in composition. To assess the differences in the oil fluid related to different reservoirs, in particular, the so-called “fingerprint” of the reservoir fluid is used (in this context, the “fingerprint” is a set of parameters that uniquely characterize the hydrocarbon fluid taken from a particular oil and gas reservoir, i.e. individualizing parameters of the composition and properties of this fluid).

Thus, an oil sample consisting of various fluids, formed, for example, as a result of the mutual contribution of oil-bearing formations during production, can be considered as a mixture of unique "fingerprints", each of which characterizes a separate reservoir. Identification of the share of each such “fingerprint” in the mixture allows one to estimate the fluid content in the composition of the mixture (analyzed sample). The accuracy of determining the fluid fraction is determined by the reliability of sampling, sample preparation, high accuracy (reliability) of data on the hydrocarbon composition of reservoir fluids and their mixture, as well as the accuracy of the data processing methods used to determine the contribution of an individual fluid to the mixture.

There are a number of methods that improve the accuracy of determining the hydrocarbon composition of the reservoir fluid and, accordingly, the visibility of these fluids in a mixture (patent RU2341792, publ. April 29, 2020, IPC: E21B 49/00, E21B 49/08, G01N 33/26, G06F 17/18). However, an accurate determination of the proportion of the contribution of various fluids to one production stream is required, which can facilitate timely decisions and interventions at all stages of development and management of an oil field.

There is a known method for analyzing the distribution of production (patent US 9074465, publ. 07/07/2015, IPC: E21B 43/14, E21B 47/00), which includes obtaining the results of spectral analysis (at least 2 of the indicated methods of analysis: electromagnetic absorption spectroscopy, spectroscopy Raman scattering, X-ray fluorescence, electromagnetic scattering spectroscopy, NMR spectroscopy or time-resolved terahertz spectroscopy) for each individual reservoir fluid, obtaining the spectral characteristics of the mixed fluid, assuming a linear relationship between the individual spectral characteristics and the spectral characteristics of the mixed fluid, determining the relationship the amount of each of the plurality of fluids in the produced fluid mixture as a result of the analysis of individual spectra for each of the fluids and the assumed linear relationship.

Common features of the known and claimed methods is the analysis of the composition of individual fluids and mixtures of fluids.

Common features of the known and claimed system for determining the proportion of reservoir fluid in a mixture of fluids is the use of methods based on linear correlation of data that make up the initial fingerprint of fluid samples (in particular, the method of principal components).

However, given the complex composition of hydrocarbon (reservoir) fluids and the difficulty of identifying individual components in each fluid, the known method does not provide high accuracy in determining the contribution of each fluid to the mixture. However, reservoir fluids do not always contain components that differ significantly between oil from one reservoir and oil from another reservoir, or they may not be very different from each other when analyzing a mixture of fluids, for example, when analyzing composition - chromatographic or spectral characteristics (peaks) can overlap on top of each other and it is not possible to establish the exact ratio of such components using the usual linear correlation.

A method is known (Kaufman, R. L. (1990). Gas chromatography as a development and production tool for fingerprinting oils from individual reservoirs: applications in the Gulf of Mexico. In GCSSEPM Foundation Ninth Annual Research Conference Proceedings, October 1, 1990 (pp. 263- 282)), according to which, using chromatographic data, a detailed analysis of the integrity of the formation is performed, for example, determining whether the fracture is tight or not, whether there are hydraulic obstructions in the reservoir. The known method is based on the determination of various fluids in oil samples using simple calibration dependencies based on the ratio of the heights of the chromatographic peaks of individual marker compounds (fingerprints of individual reservoir fluids).

Common features of the known and claimed methods are the determination of the proportion and presence of reservoir fluid in the mixture (to analyze the formation of a mixture of reservoir fluids as a result of crossflows between reservoirs) using data on the hydrocarbon composition of individual reservoir fluids, as well as identifying differences between the hydrocarbon compositions of these reservoir fluids.

However, the known method is based on the assumption of a linear combination of chromatographic peaks of individual reservoir fluids in the resulting fluid mixture chromatogram, which is not always observed when analyzing real samples. In addition, the known method allows the analysis of mixtures consisting of no more than three individual fluids. This is due to the fact that when mixing two fluids, the linear calibration dependence is presented as a curve, when mixing three fluids, the surface, however, with a larger number of fluids in the mixture, there is no simple graphical representation of the model from which a solution could be found. Chromatograms of individual reservoir fluids may have minor differences, which leads to an increase in the error in determining the contribution of each fluid to the mixture when using simple calibration dependencies.

The closest analogue (prototype) is the method of operational separate accounting for the production of a two-layer production facility (patent RU2625822, publ. from a two-layer production facility, statistical processing of the data obtained, while preparing samples, mixing samples of single-layer objects in specified proportions, conducting a study of the dynamic viscosity of the resulting model mixtures, building relationships between the content of the model mixtures of the share of oil of each of the layers and the dynamic viscosity of the model mixtures, after which the share of each of the layers in the composition of the mixture is determined.

Common features of the known and claimed methods are sampling of individual fluids (single-layer object), sample preparation, obtaining model mixtures with different ratios of individual fluids, determining the proportion of each fluid when comparing data obtained for a mixture of fluids and for model mixtures with different ratios of individual fluids.

However, inappropriate selection of samples of individual fluids that are used to prepare model mixtures can lead to a significant decrease in the accuracy of determining the proportion of each fluid in a fluid mixture. In addition, the values of the dynamic viscosity of the mixtures do not provide a sufficiently high accuracy in determining the composition of the fluid mixture and, accordingly, the contribution of the individual fluid to the mixture.

The technical result of the invention is to increase the accuracy of determining the content of the proportion of each formation fluid in a mixture of two or more fluids, as well as to increase the reliability of the distinguishability of hydrocarbon fluids in a mixture that belong to different formations/fields.

The technical result is achieved for a method for determining the proportion of reservoir fluid in a mixture of fluids, which includes obtaining at least one sample of an individual reservoir fluid from at least two reservoirs, preparing calibration mixtures by mixing individual reservoir fluids from different reservoirs in various ratios, determining the hydrocarbon composition of the obtained calibration mixtures, establishing the relationship between the content of individual formation fluids in calibration mixtures and the obtained data on the hydrocarbon composition of calibration mixtures using the multivariate regression method, obtaining a sample of the studied fluid and determining its hydrocarbon composition, determining the proportion of each individual formation fluid from the obtained data on the hydrocarbon composition of the sample of the studied fluid and according to the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures.

The achievement of the technical result is ensured by analyzing the composition of the calibration mixtures, and not by physical and chemical properties, as well as by using the multivariate regression method for analysis and determining the contribution of each individual reservoir fluid to the mixture, which allows minimizing the error as a result of finding the relationship between the analytical signal and the value of the determined parameter.

For the preparation of calibration mixtures, one sample of each individual formation fluid may be used, and three or more samples of each individual formation fluid may be used. When using more than one sample, "averaged" samples of individual reservoir fluids are prepared, which contain fluids from all received samples of one reservoir and representatively cover the entire range of contents of each fluid, taking into account all possible variations in the contents of individual compounds. Such variations are usually associated with the natural processes of mass transfer within one reservoir, with the features of sampling and sample preparation, as well as with instrumental errors of gas chromatographic equipment.

The technical result is achieved for a method for determining the proportion of reservoir fluid in a mixture of fluids, which includes obtaining at least three samples of individual reservoir fluid from at least two reservoirs and determining their hydrocarbon composition, determining samples characterizing the composition of individual reservoir fluids, according to the hydrocarbon composition of samples of individual formation fluids for preparation of calibration mixtures, preparation of calibration mixtures by mixing samples of individual formation fluids from different formations determined at the previous stage in various ratios, determination of the hydrocarbon composition of the obtained calibration mixtures, establishing the relationship between the content of individual formation fluids in calibration mixtures and the data obtained on the hydrocarbon composition of calibration mixtures mixtures using the multivariate regression method, obtaining a sample of the studied fluid and determining its hydrocarbon composition, determining the proportion of each ind visual reservoir fluid according to the obtained data on the hydrocarbon composition of the test fluid sample and the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures.

The achievement of the technical result is achieved through the use of calibration mixtures and the use of the multivariate regression method to establish the relationship between data on the composition of calibration mixtures and data on their hydrocarbon composition. The use of calibration mixtures makes it possible to take into account changes in the composition that can be observed when mixing real samples of formation fluids.

The use of the multivariate regression method to compare data (the content of individual reservoir fluids) of calibration mixtures (samples) with data on the hydrocarbon composition of the fluid mixture when determining the proportion of individual reservoir fluids in the test sample, which may be a mixture of fluids, allows minimizing the error as a result of finding a relationship between analytical signal (for example, the height of chromatographic peaks) and the value of the determined parameter (the amount of fluid of an individual reservoir in the calibration mixture). To do this, hydrocarbon composition data (eg chromatographic peak heights) are organized in a matrix format, where each row corresponds to a different calibration sample. The values of the determined parameter in the calibration samples are entered into a separate vector (Y). Thus, the regression equation relating the value of the determined parameter with the heights of the peaks can be written in matrix form: Y = BX.

The test sample is a reservoir fluid, the content of individual fluids, in which it is necessary to establish/verify. For example, if it is necessary to determine the presence of fluid flow between reservoirs.

One of the methods of multivariate regression is the method of projection onto latent structures (PLS method), the use of which makes it possible to further improve the accuracy of determining the proportion of an individual formation fluid in a mixture of fluids by using a complete set of variables (for example, all heights of chromatographic peaks for all studied samples), correlated with a change in the target modeled parameter (for example, the proportion of the reservoir in the mixture).

When using the PLS method as a multivariate regression method, a PLS model is built.

An important step in creating a PLS model, which determines the accuracy with which the contribution of each fluid to the mixture will be determined, is the determination (selection) of formation fluid samples for creating a calibration set. If such a choice is possible, for example, when taking several samples of an individual formation fluid.

Due to the fact that the composition of samples taken from the same reservoir may differ (in some cases quite significantly), the hydrocarbon composition of at least three samples of an individual reservoir fluid is determined. Determination of the optimal reservoir fluid sample for calibration mixtures allows reducing uncertainties in the composition of an individual reservoir fluid, which are associated with the fact that in reservoirs of different permeability, the composition of individual reservoir fluid samples (taken from different wells) can vary significantly. At the same time, it is possible to avoid mixing all the received samples of an individual reservoir fluid to obtain samples for calibration mixtures.

Sample determination can be carried out by an expert as a result of analysis and comparison by any method known to a specialist of data on the hydrocarbon composition of samples of reservoir fluids with the identification of a sample that characterizes the fluid of an individual reservoir to the greatest extent.

To select (determine) a sample of each formation fluid used in the preparation of calibration mixtures, it is preferable to use the principal component method (PCA), which is an informative and widely used method for comparing individual characteristics of the hydrocarbon composition of formation fluids. This allows you to determine a sample from several samples of an individual reservoir fluid, which will characterize the composition of this fluid as much as possible (a sample that has fluid-averaged scores for the first two components is an “averaged sample”).

In the event that the samples of one individual formation fluid are characterized by a spread relative to each other on the graph of the PCM scores (i.e., the hydrocarbon compositions of these formation fluids of the same formation differ significantly, which is typical, for example, for low-permeability formations), then it is preferable to additionally obtain an “averaged” a sample of this individual formation fluid, which will include the most different samples, before preparing calibration mixtures of several formation fluids. For this, samples of this individual fluid can be used, which have the greatest distance from each other on the PCA score chart. It is these samples that will most strongly differ from each other in chemical composition and will allow the most complete description of this individual fluid.

The estimation of the distance on the graph of the PCA scores is clear and known to the expert. Obtaining "averaged" samples allows you to further improve the accuracy of the method.

Before determining the composition of reservoir fluids or from a mixture, sample preparation is preferably carried out. Sample preparation may include heating the sample for 20 to 40 minutes in a sealed vessel at a temperature of 333 ± 30 K with stirring at centrifugal acceleration from 980 to 58800 m/s ² (from 100 to 6000 g).

The determination of the hydrocarbon composition can be carried out by any method known to those skilled in the art, for example, using a gas chromatography method in combination with a flame ionization detector or a quadrupole mass analyzer or a time-of-flight mass detector. The choice of the method for determining the hydrocarbon composition of the reservoir fluid can be carried out by the value of the density of the reservoir fluid. This approach is known from the prior art (patent RU 2742651, publ. 09.02.2021) and makes it possible to improve the accuracy of determining the hydrocarbon composition of the fluid, which further improves the accuracy of determining the contribution of each individual fluid to their mixture.

The task of PLC modeling is to find the vector of regression coefficients B of the matrix Y = BX, which will relate the heights of the peaks to the fraction of a particular fluid. This problem is solved by simultaneously expanding the X and Y matrices into a space of latent variables (LV), which are determined by the maximum change in variance in the X matrix and the Y vector (Wold, S., Sjöström, M., & Eriksson, L. (2001). PLS-regression: a basic tool of chemometrics Chemometrics and intelligent laboratory systems, 58(2), 109-130). Once vector B has been determined with the calibration samples, the PLC model is ready to determine the proportion of fluid in the unknown samples. The values of the regression coefficients B depend on the number of hidden variables used to decompose the matrices X and Y. The first LVs are associated with the maximum sources of variance in X correlated with Y, i.e. are a combination of peak heights corresponding to the compounds, the content of which changes most strongly with a change in the proportion of the formation in the calibration sample. Subsequent LVs are associated with smaller contributions of compounds to the total variance - the older the LV, the smaller the variance it describes, and therefore the older LVs are discarded as an uninformative source of noise. Thus, by changing the number of latent variables in the model, the share of useful information and noise in the model is regulated, which, in turn, affects the final error in determining the fluid share. If the LV number is small, then not all useful information is used to determine the target parameter, which leads to an increase in the error. On the other hand, with a large number of LVs, noise enters the model in addition to useful information, which also leads to an increase in the error. The optimal number of LVs can be determined during a validation procedure that is known in the art to those skilled in the art and is based on minimizing model performance criteria that estimate the error in determining the proportion of fluid in the mixture. In this connection, the method may additionally include the stage of validating the PL model.

To calculate such criteria, for example, a separate set of test samples with known values of the Y parameter is required, but, unlike samples for calibration mixtures, not used to determine the vector B (to build a model). The most common measure of model quality is the root mean square prediction error (RMSEP):

where y ⁱ _real is the known content of the determined fluid in the test sample,

i , y ⁱ _pred is the fluid content in the test sample calculated using the model (at certain values of B),

i, n is the total number of test samples.

The RMSEP value is used not only to optimize the model (search for the amount of LV to calculate B, at which the RMSEP will be minimal), but also to estimate the overall error in determining the fluid content in the mixture. The main difficulty in the practical application of this criterion is the need to use additional samples in the test set, which increases the complexity of the analysis and its cost. This disadvantage can be eliminated by optimizing the model using the full cross-validation method. In this method, samples are excluded one by one from the calibration set, which are used to verify the model. The determination of the regression coefficients is carried out on the remaining samples, and the predicted value of the excluded sample is used to calculate the standard error. The excluded sample is then returned to the calibration set, from which the next calibration sample is excluded. This procedure is repeated until all samples of the calibration set have been tested. The quality criterion of the model calculated from these samples is called the Root Mean Square Error of Cross Validation (RMSECV). Its calculation does not require additional standard samples with a known content of the parameter to be determined. Those. Validation of the PLS model can be done using the rms cross-validation error or, for example, the rms prediction error.

An increase in the accuracy of the constructed model further increases with an increase in the number of calibration samples. In particular, it is preferable to use at least 5 calibration mixtures.

The optimal quantity and composition (content of individual formation fluids for preparation) of calibration mixtures from samples of individual fluids is determined depending on the number of formation fluids in the mixture. The simplest mixtures are two-component mixtures, which include only two fluids. In this case, it is preferable (but not required) to use a ten-level mix design. In accordance with this design, the calibration set will include two samples of individual fluids and nine samples of synthetic mixtures with a mixing step of 10% (i.e. the first sample of the mixture will contain 10% of the first fluid and 90% of the second, and the last one, on the contrary, will contain 90% of the first fluid and 10% of the second). From the case of a three-component mixture (a sample consists of three fluids), the use of a ten-level full-factor design significantly increases the complexity, and hence the overall efficiency of the analysis, since it requires a large number of samples of calibration mixtures (namely, 10 ³ = 1000 samples). To solve this problem - reducing the number of calibration samples with a slight deterioration in the quality of the model, several approaches have been proposed, the most preferable of which in the case of a closed system (the sum of the contents of all components in each mixture is 100%) can be a simplex-lattice experimental design (Vidmer G.M. ., Kelner R., Merme J.-M., Otto M. Analytical chemistry. Problems and approaches. Volume 2 - MIR, Moscow, 2004, 726 pages). This design uses:

- three individual fluid samples;

- one sample containing all three fluids in equal amounts (33% of each fluid);

- nine mixtures describing two-component interactions (mixtures containing only two fluids in different ratios);

- nine mixtures describing three-component interactions (mixtures containing all three fluids in various proportions, except for a mixture containing all three fluids in equal amounts).

Such a calibration set will contain the optimal number of samples, which will allow one to determine the content of each of them with a sufficiently high accuracy, i.e. building good quality 3-component PLS models will also be provided.

In the case of a larger number of individual fluids in a mixture (more than 3), the use of a simplex-lattice experimental design leads to a significant increase in calibration samples (for example, 69 samples are required for a four-component mixture, and 251 samples for a five-component mixture). There are various approaches to create calibration kits with the minimum possible number of samples that can be used in multicomponent mixtures. For example, the cyclic permutation design (Brereton, R. G. (1997). Multilevel multifactor designs for multivariate calibration. Analyst, 122(12), 1521-1529.), which is based on the concept of a cyclic generator, which allows obtaining a small number of calibration samples for multicomponent mixtures. In this case, the changes in the contents for all components of the mixture are strictly orthogonal, which makes it possible to independently take into account various interactions between individual factors. However, in this design, the number of calibration samples is strictly fixed. For example, a gauge design is known (Kirsanov, D., Panchuk, V., Agafonova-Moroz, M., Khaydukova, M., Lumpov, A., Semenov, V., & Legin, A. (2014). A sample -effective calibration design for multiple components Analyst, 139(17), 4303-4309), based on a uniform distribution of N points in a space of arbitrary dimension n, where each measurement corresponds to a separate component in the mixture, and a point corresponds to a sample. The volume of this space is fixed by the upper limits of the concentration of components. For uniform filling, the entire n-dimensional volume is divided into m identical equilateral subvolumes in such a way that when the entire volume is filled with dots, each subvolume contains at least one point, i.e. m≤N. In this case, the number of these subvolumes should be maximum, for example, for a two-dimensional space containing 9-15 points, m will be equal to 9, if the number of points is 16-25, then m=16, etc.

Additionally, a preliminary assessment of the reliability of the PLC model can be carried out by the ratio of variations in the composition (dispersion) of samples within an individual reservoir fluid and between them in the space of the main components using the PCA, respectively. Quantification of the possibility of separating samples into layers can be carried out on the basis of the classic multivariate Hotelling test (Esbensen, KH, Guyot, D., Westad, F., & Houmoller, LP (2002). Multivariate data analysis: in practice: an introduction to multivariate data analysis and experimental design. In this case, the accounts of the first two components obtained by modeling using the PCA are used as features. The Hotelling test (T2) is a generalization of Student's t-statistics used to detect significant differences in samples described by several features. The T2 test calculated from the Hotelling statistics for two groups of samples (for example, from two different fluids) is converted into an F-test and compared with the critical (tabular) values of the Fisher F-distribution. If the obtained value of F is less than the critical one (F _table ), then the differences between the groups are random, otherwise the differences between the groups are statistically significant and samples from these layers can be used to create calibration mixtures of the PLC model.

The technical result is also achieved when using in the methods a system (computer) for determining the proportion of formation fluid in a mixture of fluids, which contains at least one processor and a program code and is configured to be executed by the processor under the control of the program code: establishing a relationship between the content of individual formation fluids in calibration mixtures, which are prepared by mixing individual reservoir fluids from different reservoirs in various ratios, and obtained data on the hydrocarbon composition of calibration mixtures using the multivariate regression method, determining the proportion of each individual reservoir fluid in the test fluid from the obtained data on the hydrocarbon composition of the test fluid sample and established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures.

The hydrocarbon composition of the calibration mixtures can be determined by any method known to the specialist for determining the composition, and also qualitatively established through the properties of the calibration mixtures and the test sample.

The achievement of the technical result is also ensured when using in the methods a computer-readable medium for determining the proportion of reservoir fluid in a mixture of fluids, on which a computer program is stored, having a program code, when executed on a computer, the processor performs establishing the relationship between the content of individual reservoir fluids in the calibration mixtures that are prepared by mixing individual reservoir fluids from different reservoirs in various ratios, and the obtained data on the hydrocarbon composition of the calibration mixtures using the multivariate regression method, determining the proportion of each individual reservoir fluid in the test fluid from the obtained data on the hydrocarbon composition of the sample of the test fluid and the established relationship between the content of individual fluids in calibration mixtures and hydrocarbon composition of calibration mixtures.

The achievement of the technical result is ensured by the implementation by the processor of establishing the specified relationship and using it to determine the share of the contribution of each individual formation fluid to the mixture according to the obtained data on the hydrocarbon composition of the mixture of fluids.

The method of multivariate regression can be the method of projection onto latent structures (PLS) and the processor can additionally build a PLS model according to the established relationship, or another well-known algorithm.

The processor can additionally determine the sample that best characterizes the composition of the individual reservoir fluid using the principal component method (PCA) and data on the hydrocarbon composition of samples of individual reservoir fluids.

Also, the processor can additionally carry out the determination of samples of individual fluids for the preparation of averaged samples according to their greatest distance from each other on the graph of PCA scores.

The processor can additionally perform a preliminary assessment of the reliability of the PLC model, for example, by the ratio of variations in the composition of samples within the fluid and between them, while the hydrocarbon composition of the samples of the formation fluid is analyzed by the processor using the PCA in the space of the main components, determines the T2 criterion calculated by Hotelling statistics, for groups of samples of individual formation fluids, recalculated into F-criterion and compared with the calculated critical values of the Fisher F-distribution, if the F-criterion is greater than the critical value of the Fisher F-distribution, the processor fixes that the LLS model of the calibration samples of formation fluids is reliable.

The processor may further validate the PLM, for example using a cross-validation rms error method.

The invention is illustrated by the following figures.

The figure 1 shows the scheme of joint production of reservoir fluids from two reservoirs, where 1 - reservoirs.

The figure 2 shows the plot of accounts obtained by the principal component method for estimating the parameters of the composition of individual reservoir fluids 2 - N1, 3 - N2, 4 - N3.

Figures 3 - 5 show plots "introduced - found" of the PLC model, reflecting the relationship between the known content of individual reservoir fluids in calibration mixtures and determined by the PLC model for 5 - calibration and 6 - test samples.

The figure 6 shows the scoring plot obtained by the principal component method for estimating the parameters of the composition of individual reservoir fluids 7 - N4, 8 - N5.

Figure 7 shows the plot "introduced - found" PLC-model, reflecting the relationship between the known content of individual formation fluids in calibration mixtures and determined by the model for 5 - calibration and 6 - test samples.

Figure 8 shows the scoring plot obtained by the principal component method for estimating the composition parameters of individual formation fluids 9 - N6, 10 - N7, 11 - N8, 12 - N9, 13 - N10, 14 - N11.

Figures 9 - 11 show plots of "introduced - found" PLC models, reflecting the relationship between the known content of individual reservoir fluids in calibration mixtures and determined by the PLC model for 5 - calibration and 6 - test samples.

At least one individual formation fluid sample is obtained from at least two formations. Then, calibration mixtures are prepared by mixing individual formation fluids from different formations in various ratios and the hydrocarbon composition of the resulting calibration mixtures is determined. The relationship between the composition of the calibration mixtures and the obtained data on their hydrocarbon composition is established using the multivariate regression method.

A sample of a mixture of individual reservoir fluids is obtained and its hydrocarbon composition is determined. The share of each individual reservoir fluid is determined according to the obtained data on the hydrocarbon composition of the reservoir fluid sample and according to the established relationship between the composition of the calibration mixtures and their hydrocarbon composition.

The following are private examples of implementation that illustrate the claimed invention, but do not limit it.

Example 1. Samples of individual fluids N1 (11 samples), N2 (4 samples), N3 (9 samples), as well as a sample of a mixture of these fluids were obtained. Sample preparation was carried out. All samples of oil (hydrocarbon fluid) were divided into two parts: the first were thermostated for 30 min at a temperature of 65°C and separated from mechanical impurities and water by centrifugation at the same temperature and relative centrifugal acceleration of 3430 m/s ² (350 g) , the latter were centrifuged at room temperature. Group analysis of oil revealed the following results, which are presented in table 1.

Table 1—Group composition of reservoir fluid samples at different sample preparation.

Oil No. Weight
saturated fraction, g Weight
aromatic fraction, g Resin weight, g Mass of asphaltenes, g Σ masses of all fractions, g No. 1 (hot) 0.4619 0.2499 0.0592 0.0203 0.7913 No. 1 (cold) 0.4568 0.2601 0.0384 0.0126 0.7679 No. 2 (hot) 0.5013 0.2009 0.0495 0.0092 0.7609 No. 2 (cold) 0.3940 0.2063 0.0110 0.0065 0.6178 No. 3 (hot) 0.5582 0.1999 0.0458 0.0080 0.8119 No. 3 (cold) 0.4639 0.2101 0.0284 0.0051 0.7075 No. 4 (hot) 0.4544 0.2187 0.0487 0.0148 0.7366 No. 4 (cold) 0.4207 0.2517 0.0359 0.0106 0.7189

The results in Table 1 indicate a change in the ratio of oil components during the separation of water and mechanical impurities without heating: relative to samples after hot centrifugation, samples obtained by separation at room temperature are depleted in tar-asphaltene compounds, while they contain a larger amount of aromatic compounds. The decrease in the proportion of the saturated fraction, apparently, can be associated with the precipitation of paraffin emulsion during centrifugation. Also, in all samples, a decrease in the sum of fractions was revealed, which may be due to the inefficient separation of water and mechanical impurities during conventional centrifugation, since they do not contribute to the mass of fractions as a result of their removal during the toluene extraction of asphaltenes and the distillation of solvents from all four fractions: insoluble impurities remain on the filter in the Soxhlet nozzle, water is distilled off in the form of an azeotrope.

Thus, it should be concluded that the proposed version of the sample preparation method not only allows one to effectively separate impurities that are undesirable for subsequent chromatographic study of the fluid, but also to preserve the initial group composition of oil, which is clearly important for quantitative analysis. The results presented confirm the additional increase in accuracy during sample preparation at elevated temperature. This sample preparation option was used to prepare samples for other examples, but at the same time it does not limit the claimed method for determining the proportion of formation fluid in a mixture of fluids, but only allows to further improve the accuracy of data on the composition of the samples, which in turn will further increase the accuracy of determining the proportion of the contribution individual reservoir fluids in the mix.

Then, for samples prepared at elevated temperature, determine the hydrocarbon composition using the method of gas chromatography, combined with a quadrupole mass analyzer. The resulting chromatograms are analyzed using PCA. In FIG. Figure 2 shows the graph of the PCA accounts for the first two components. The share of the explained dispersion of these components is 71%. The calculated F values for each pair of all three reservoirs are significantly higher than the tabular values of the Fisher criterion: the F value for the N1 and N2 reservoirs was 93.26 (F _table = 3.88), for the N1 and N3 reservoirs the F value was 129.14 (F _table = 3.59), for layers N2 and N3 the value of F was 136.85 (F _table = 4.10), respectively, the quality of the PLS models obtained in the future will be high.

Figure 2 shows that the samples of individual formation fluids below have the largest differences in composition (with the greatest distance relative to each other) in the formation (N1: samples #1, #3, #4, #11; N2: samples #12 , No. 13, No. 14, No. 15; N3: samples No. 16, No. 18, No. 20, No. 22).

From these samples, by mixing in equal amounts, "averaged" samples were obtained for each individual reservoir fluid for the preparation of calibration samples. This made it possible to reduce the number of samples for the preparation of "averaged" samples of individual reservoir fluids. At the same time, a specialist can choose a variant of the implementation of the method, in which "averaged" samples are prepared by mixing in equal quantities all samples of an individual reservoir fluid.

The composition (amount of individual reservoir fluids for preparation) of the calibration mixtures was determined from the simplex-lattice experimental plan (Table 2).

Table 2—Content of individual formation fluids in calibration mixtures, %.

Sample No. N1 N2 N3 PLS1 100 0 0 pls 2 0 100 0 pls 3 0 0 100 pls 4 33 33 33 pls 5 25 75 0 pls 6 75 25 0 pls 7 0 25 75 pls 8 0 75 25 pls 9 25 0 75 pls 10 75 0 25 pls 11 twenty 40 40 pls 12 40 twenty 40 pls 13 40 40 twenty pls 14 ten 70 twenty pls 15 twenty 70 ten pls 16 70 twenty ten pls 17 70 ten twenty pls 18 ten twenty 70 pls 19 twenty ten 70 pls 20 fifty fifty 0 pls 21 fifty 0 fifty pls 22 0 fifty fifty

Determine the hydrocarbon composition of the calibration mixtures using the method of gas chromatography, combined with a quadrupole mass analyzer. In this case, for additional accuracy in determining the contribution, the PLC method was chosen as one of the multivariate regression methods.

Data on the heights of chromatographic peaks and data on the content of individual reservoir fluids in calibration mixtures are transferred to the latent space using the PLC method, the PLC model is built to establish the relationship between the composition of the calibration mixtures and their hydrocarbon composition (curve 5 in figures 3-5, shown in black). The principle of constructing a PLC model is known to those skilled in the art.

Determination of the formation fluid mixture composition is carried out by determining the hydrocarbon composition of the mixture using the same method that was used to determine the hydrocarbon composition of the calibration mixtures. The obtained data of chromatograms - the heights of chromatographic peaks are converted into the space of latent variables using the PLS method and a PLS model is obtained. Figures 3 - 5 show the graphs of the "introduced-found" PLC models for determining the proportion of reservoir fluid: N1 - Fig.3, N2 - Fig.4, N3 - Fig.5.

As a result of comparing the data of the generated PLC model and data on the hydrocarbon composition of the mixture of formation fluids (the heights of the chromatographic peaks obtained when determining the hydrocarbon composition of the mixture of fluids), the content of individual formation fluids in the mixture is determined from the heights of the chromatographic peaks (curve 6 in figures 3-5, shown in blue). Table 3 shows an example of the result of determining the composition of the mixture according to the PLC models presented in figures 3 - 5 for one of the mixtures of hydrocarbon fluids, the composition of which was determined (which was used for verification).

Table 3—Proportion of the contribution of individual formation fluids.

Reservoir fluid Content in the mixture, % N1 39.5 N2 13.1 N3 45.2

The optimization of the model and estimation of its error were carried out using full cross-validation. All constructed PLS models for a three-component mixture are of satisfactory quality - the correlation coefficients of the models take values not lower than 0.8, and the RMSECV values do not exceed 13%.

Example 2 Samples of individual fluids N4 (6 samples) and N5 (5 samples) were obtained, as well as a sample of a mixture of these fluids. Sample preparation was carried out at elevated temperature as in example 1. The hydrocarbon composition of all samples was determined. Further, the chromatograms were processed by the principal component method. Figure 6 is a graph of the PCA scores for the first two components, which together explain 88% of all variance in the data. The scores of the first two components were used to calculate the F criterion calculated according to the Hotteling test, which amounted to 38.35, which significantly exceeds the tabular value (4.26).

The following were selected for mixing: sample No. 3 of reservoir fluid N4 and sample No. 15 of reservoir fluid N5, which had the average values of the first two principal components over the entire reservoir. In this case, no average samples of individual fluids were obtained. Calibration samples were obtained by mixing sample samples according to a ten level design. The content of individual reservoir fluids in the calibration mixtures is given in Table 4. Further, for all calibration samples, the hydrocarbon composition is determined, the heights of the chromatographic peaks are determined, which are used to build a PLC model designed to determine the fraction of the N4 formation.

The hydrocarbon composition of the mixture is determined, the proportion of the contribution of individual fluids must be determined using the same gas chromatography method as for determining the hydrocarbon composition of the calibration mixtures. The heights of the chromatographic peaks and the generated PLC model are used to determine the contribution of the individual fluid. In this case (two-component mixture), the contribution of formation fluid N4 was determined. The N5 reservoir fluid fraction was defined as 100 N4 reservoir fraction.

Table 4—Content of individual reservoir fluids in calibration mixtures, %.

Sample N4 N5 Y01 0 100 Y02 twenty 80 Y03 40 60 Y04 60 40 Y05 80 twenty Y06 100 0 Y07 ten 90 Y08 thirty 70 Y09 fifty fifty Y10 70 thirty Y11 90 ten

Figure 7 shows the "introduced-found" plot of the resulting PLS model. Black color shows the samples involved in the construction of the model. Blue - samples in the test. The contribution of the N4 formation fluid was 75.6%. The optimization of the model and estimation of its error were carried out using full cross-validation. In the ideal case, the straight lines drawn on the calibration and test samples should have a slope and a correlation coefficient close to unity, and the bias should be near zero, which is observed for the resulting PLC model - the slopes in the calibration and test are 0.99 and 0, 98; correlation coefficients 0.99 and 0.98; and an offset of 0.6 and 0.9, respectively. The overall error of the model, estimated from the value of RMSECV is 5.2%. All parameters of the model (slope of the graph entered-found, its shift and correlation coefficient), as well as the total error, indicate the high quality of the obtained LLS model for determining the fraction of two-component mixture reservoirs.

Example 3 Individual fluid samples were taken from six reservoirs. Sample preparation was carried out at elevated temperature as in example 1. The hydrocarbon composition of all samples was determined. Further, the chromatograms were processed by the principal component method. Figure 8 is a plot of PCA scores for samples of these fluids. It can be seen from the graph that reservoirs N6, N7, N8, N9, N10 do not have a clear separation with the exception of reservoir N11, however, six-component models were built for these reservoirs to determine the proportion of each individual reservoir fluid in the mixture. Due to the fact that it is difficult to establish the exact boundaries of the areas on the PCA graph for layers, the cluster centers corresponding to individual layers on the graph of accounts were selected, and sample samples corresponding to the centers of the clusters were selected for the preparation of calibration mixtures. The calibration mixtures were obtained by mixing the samples in accordance with the design of a uniform distribution of points in n-dimensional space, which is described in detail above. The contents of individual samples in mixtures of calibration samples are shown in Table 5.

The hydrocarbon composition of the calibration mixtures is determined. Using data on the composition of the calibration mixtures and their hydrocarbon composition (characterized by the heights of the chromatographic peaks), PLC models are built to determine the proportion of the contribution of fluids from all six reservoirs to the mixture. In FIG. 9-11 shows the plots of "introduced-found" PLS models with the best quality - N9 (Fig. 9), N10 (Fig. 10), N11 (Fig. 11). As expected after analysis of the PCG graph, the best quality model is designed to determine the proportion of the reservoir N11 - the correlation coefficient in cross-validation was 0.97, and the value of RMSECV 4.5%. The PLC models for the N9 and N10 formations were characterized by lower quality, which nevertheless remained satisfactory (the correlation coefficient is greater than 0.8 for both models, and the error in determining the proportion of the formations did not exceed 11%).

Table 5—Content of individual formation fluids in calibration mixtures, %.

No. N6 N7 N8 N9 N10 N11 C1 78 3 2 6 5 6 C2 22 ten 34 eleven 2 21 C3 39 26 eight 0 6 21 C4 54 eleven twenty 3 12 0 C5 97 0 0 2 one 0 C6 one 70 four 12 0 13 C7 22 40 2 four 23 9 C8 four 24 2 one 5 64 C9 3 92 one one 0 3 C10 9 80 3 four 3 one C11 12 54 7 9 ten eight C12 four one 56 13 19 7 C13 0 5 74 6 3 12 C14 0 0 96 2 one one C15 twenty four 44 four 19 9 C16 7 34 eleven fifteen 2 31 C17 2 four 3 90 one 0 C18 one one 2 71 four 21 C19 9 0 21 55 fourteen one C20 3 9 four 42 31 eleven C21 5 eight 24 26 eight 29 C22 2 9 one four 78 6 C23 one 0 0 four 95 0 C24 ten four fifteen 6 63 2 C25 fifteen 9 16 eight 47 5 C26 0 0 0 four 0 96 C27 four 7 3 one 2 83 C28 9 2 6 5 ten 68 C29 9 6 28 3 one 53 C30 25 24 13 3 29 6 C31 12 23 29 19 12 5 C32 28 19 36 13 2 2 C33 ten 17 13 23 22 fifteen C34 19 16 25 5 fourteen 21 C35 eight 9 46 28 5 four C36 four 5 7 9 6 69

The remaining models for determining the proportion of the N6, N7, N8 formations were of insufficiently satisfactory quality (the correlation coefficient of these models did not exceed 0.5, and the average error was above 30%). Despite the low quality of the PLC models for these three formations, the proposed method demonstrated high efficiency, as it allowed determining the content of formations that differ slightly from each other in composition (N9 and N10).

Table 6 presents the results of determining the proportion of the contribution of fluids from the formations N9, N10 and N11 to the fluid mixture.

Table 6—Proportion of the contribution of individual formation fluids.

Reservoir fluid Content in the mixture, % N9 2.7 N10 14.6 N11 61.6

Thus, the above examples confirm the achievement of the technical result - an increase in the accuracy of determining the content of the proportion of each reservoir fluid in a mixture of two or more fluids, as well as an increase in the reliability of the distinguishability of hydrocarbon fluids in a mixture that belong to different reservoirs / fields, which is provided by determining the hydrocarbon composition of the calibration mixtures and establishing the relationship between the content of individual reservoir fluids in calibration mixtures and the hydrocarbon composition of calibration mixtures using multivariate regression methods.

Establishing the relationship between the content of individual reservoir fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures and determining the contribution of individual reservoir fluids can be carried out using the claimed system and computer-readable media.

Claims (68)

1. Method for determining the proportion of formation fluid in a mixture of fluids, which includes - obtaining at least one sample of an individual formation fluid from at least two different formations; - preparation of calibration mixtures by mixing individual reservoir fluids from different reservoirs in various ratios; - determination of the hydrocarbon composition of the obtained calibration mixtures; - establishment of the relationship between the content of individual reservoir fluids in the calibration mixtures and the obtained data on the hydrocarbon composition of the calibration mixtures using the multivariate regression method; - obtaining a sample of the studied fluid and determining its hydrocarbon composition; - determination of the proportion of each individual reservoir fluid based on the obtained data on the hydrocarbon composition of the test fluid sample and on the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures. 2. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 1, in which the multivariate regression method is a projection onto latent structures (PLS) method. 3. A method for determining the proportion of formation fluid in a mixture of fluids according to claim 2, in which a PLC model is built. 4. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 1, in which three samples of each individual formation fluid are taken. 5. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 4, which further includes preparing averaged samples of individual fluids of selected samples of each individual formation fluid before preparing calibration mixtures. 6. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 1, which additionally includes sample preparation of the selected samples. 7. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 6, in which sample preparation includes heating the sample from 20 to 40 min in a sealed vessel at a temperature of 333±30 K with stirring at centrifugal acceleration from 980 to 58800 m/s ² . 8. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 1, in which the hydrocarbon composition is determined using a gas chromatography method in combination with a flame ionization detector, or with a quadrupole mass analyzer, or with a time-of-flight mass detector. 9. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 8, in which the choice of a method for determining the hydrocarbon composition is carried out according to the density value of the formation fluid. 10. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 1, in which the test sample includes two individual reservoir fluids. 11. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 10, in which the test sample includes two individual reservoir fluids, while the preparation of calibration mixtures is carried out in accordance with a ten-level design, in which the calibration set includes two samples of individual fluids and nine samples of synthetic mixtures with a mixing step of 10%. 12. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 1, in which the number of calibration mixtures is at least 5. 13. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 3, in which a preliminary assessment of the reliability of the PLC model is additionally carried out. 14. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 13, in which a preliminary assessment of the reliability of the PLC model is carried out by the ratio of variations in the composition of samples within the fluid and between them, while the hydrocarbon composition of the samples of the reservoir fluid is analyzed using PCA in the space of the main components , determine the T2 criterion calculated by Hotelling statistics for groups of samples of individual reservoir fluids, recalculate into the F-criterion and compare with the calculated critical values of the Fisher F-distribution, if the F-criterion is greater than the critical value of the Fisher F-distribution, fix that The PLC model of reservoir fluid calibration samples is reliable. 15. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 3, which further includes the step of validating the PLC model. 16. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 15, in which the validation of the PLC model is carried out using the rms error of the cross-validation. 17. Method for determining the proportion of formation fluid in a mixture of fluids, which includes - obtaining at least three samples of individual formation fluid from at least two different formations and determining their hydrocarbon composition; - determination of samples characterizing the composition of individual reservoir fluids, according to the hydrocarbon composition of samples of individual reservoir fluids for the preparation of calibration mixtures; - preparation of calibration mixtures by mixing samples of individual reservoir fluids from different reservoirs determined at the previous stage in various ratios; - determination of the hydrocarbon composition of the obtained calibration mixtures; - establishment of the relationship between the content of individual reservoir fluids in the calibration mixtures and the obtained data on the hydrocarbon composition of the calibration mixtures using the multivariate regression method; - obtaining a sample of the studied fluid and determining its hydrocarbon composition; - determination of the proportion of each individual reservoir fluid based on the obtained data on the hydrocarbon composition of the test fluid sample and on the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures. 18. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 17, in which the multivariate regression method is a projection onto latent structures (PLS) method. 19. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 18, in which a PLC model is built. 20. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 17, in which the determination of a sample characterizing the composition of an individual formation fluid is carried out using the principal component method (PCA). 21. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 17, which further includes preparing average samples of individual fluids before preparing calibration mixtures. 22. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 21, in which samples of this individual fluid are used to prepare averaged samples of individual fluids, which have the greatest distance from each other on the PCA score chart. 23. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 17, in which sample preparation is carried out before determining the hydrocarbon composition of the samples. 24. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 23, in which sample preparation includes heating the sample from 20 to 40 minutes in a sealed vessel at a temperature of 333±30 K with stirring at centrifugal acceleration from 980 to 58800 m/s ² . 25. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 17, in which the hydrocarbon composition of the samples is determined using a gas chromatography method in combination with a flame ionization detector, or with a quadrupole mass analyzer, or with a time-of-flight mass detector. 26. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 25, in which the choice of a method for determining the hydrocarbon composition of the formation fluid is carried out by the value of the density of the formation fluid. 27. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 17, in which the number of calibration mixtures is at least 5. 28. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 18, in which the test sample includes two individual reservoir fluids. 29. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 29, in which the test sample includes two individual reservoir fluids, while the preparation of calibration mixtures is carried out in accordance with a ten-level design, in which the calibration set includes two samples of individual fluids and nine samples of synthetic mixtures with a bias step of 10%. 30. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 19, in which a preliminary assessment of the reliability of the PLS model is additionally carried out. 31. The method for determining the proportion of reservoir fluid in a mixture of fluids according to claim 30, in which a preliminary assessment of the reliability of the PLC model is carried out by the ratio of variations in the composition of samples within the fluid and between them, while the hydrocarbon composition of the samples of the reservoir fluid is analyzed using PCA in the space of the main components , determine the T2 criterion calculated by Hotelling statistics for groups of samples of individual reservoir fluids, recalculate into the F-criterion and compare with the calculated critical values of the Fisher F-distribution, if the F-criterion is greater than the critical value of the Fisher F-distribution, fix that The PLC model of reservoir fluid calibration samples is reliable. 32. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 19, which further includes the step of validating the PLC model. 33. The method for determining the proportion of formation fluid in a mixture of fluids according to claim 32, wherein the validation of the model is performed using the rms error of the cross-validation. 34. A system for determining the proportion of formation fluid in a mixture of fluids, which contains at least one processor and program code and is configured to be executed by the processor under the control of the program code: - establishing the relationship between the content of individual reservoir fluids in calibration mixtures, which are prepared by mixing individual reservoir fluids of different reservoirs in various ratios, and the obtained data on the hydrocarbon composition of calibration mixtures using the multivariate regression method; - determining the proportion of each individual reservoir fluid in the test fluid according to the obtained data on the hydrocarbon composition of the sample of the test fluid and the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures. 35. The system for determining the proportion of reservoir fluid in a mixture of fluids according to claim 34, in which the multivariate regression method is a method of projection onto latent structures (PLS). 36. The system for determining the proportion of reservoir fluid in a mixture of fluids according to claim 35, in which the processor additionally builds a PLC model. 37. The system for determining the proportion of formation fluid in a mixture of fluids according to claim 34, in which the processor additionally determines the sample that best characterizes the composition of an individual formation fluid using the principal component method (PCA) and data on the hydrocarbon composition of samples of individual formation fluids. 38. The system for determining the proportion of formation fluid in a mixture of fluids according to claim 37, in which the processor additionally determines samples of individual fluids for preparing averaged samples by their greatest distance from each other on the graph of PCM scores. 39. The system for determining the proportion of formation fluid in a mixture of fluids according to claim 35, in which the processor additionally conducts a preliminary assessment of the reliability of the PLC model. 40. The system for determining the proportion of reservoir fluid in a mixture of fluids according to claim 39, in which the processor performs a preliminary assessment of the reliability of the PLC model by the ratio of variations in the composition of samples within the fluid and between them, while the hydrocarbon composition of the samples of the reservoir fluid is analyzed by the processor using the PCA in space principal components, determines the T2 criterion calculated by Hotelling statistics for groups of samples of individual formation fluids, recalculates into an F-criterion and compares with the calculated critical values of the Fisher F-distribution, if the F-criterion is greater than the critical value of the Fisher F-distribution, the processor fixes that the PLC model of the reservoir fluid calibration samples is reliable. 41. The system for determining the proportion of formation fluid in a mixture of fluids according to claim 35, in which the processor additionally validates the PLC model. 42. The system for determining the proportion of formation fluid in a mixture of fluids according to claim 41, in which the processor validates the model using a cross-validation rms error method. 43. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids, on which a computer program is stored, having a program code, when executed on a computer, the processor performs: - establishment of the relationship between the content of individual reservoir fluids in calibration mixtures, which are prepared by mixing individual reservoir fluids of different reservoirs in various ratios, and the obtained data on the hydrocarbon composition of calibration mixtures using the multivariate regression method; - determination of the proportion of each individual reservoir fluid in the test fluid according to the obtained data on the hydrocarbon composition of the sample of the test fluid and the established relationship between the content of individual fluids in the calibration mixtures and the hydrocarbon composition of the calibration mixtures. 44. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 43, in which the multivariate regression method is a method of projection onto latent structures (PLS). 45. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 44, in which the processor additionally builds a PLC model. 46. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 43, in which the processor additionally determines the sample that best characterizes the composition of an individual formation fluid using the principal component method (PCA) and data on the hydrocarbon composition of samples of individual formation fluids. 47. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 46, in which the processor additionally determines samples of individual fluids to prepare averaged samples by their greatest distance from each other on the PCA score chart. 48. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 44, in which the processor additionally performs a preliminary assessment of the reliability of the PLC model. 49. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 48, in which the processor performs a preliminary assessment of the reliability of the PLC model by the ratio of variations in the composition of samples within the fluid and between them, while the hydrocarbon composition of the samples of the formation fluid is analyzed by the processor using the PCA in the space of principal components, determines the T2 criterion calculated by Hotelling statistics for groups of samples of individual reservoir fluids, recalculates into an F-criterion and compares with the calculated critical values of the Fisher F-distribution, if the F-criterion is greater than the critical value of the Fisher F-distribution , the processor fixes that the PLC model of the reservoir fluid calibration samples is reliable. 50. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 44, in which the processor additionally validates the PLC model. 51. A computer-readable medium for determining the proportion of formation fluid in a mixture of fluids according to claim 50, in which the processor validates the model using a cross-validation rms error method.

Images