RU2725506C1 - Method and system for optimizing laboratory analysis of rock samples - Google Patents

Abstract

FIELD: mining; data processing.SUBSTANCE: group of inventions relates to means of optimizing laboratory analysis of rock samples taken from an oil-and-gas bearing formation. Essence: standard laboratory studies of rock samples are performed, the results of which are used to form a database on analysis of formations. Using the obtained results of standard laboratory studies of rock samples as input data, basic predicative models are used to predict missing rock characteristics for each sample using machine learning algorithms. Using as input data obtained results of standard studies of rock samples and results of prediction of missing characteristics of rocks, obtained using basic predictive models, machine learning creates predictive models of the second group to determine corrections to formation conditions and to predict rock characteristics in thermobaric conditions. Using as input data obtained results of rock samples analysis, results of prediction of missing rock characteristics, obtained using basic predictive models, and the results of prediction of rock characteristics in thermobaric conditions, obtained using predictive models of the second group, by means of machine training create predictive models of the third group for prediction of multidimensional characteristics of reservoir rocks. Using all created predictive models, rock characteristics are predicted. All obtained results of forecasting rock characteristics are entered into a database on analysis of formations. Current degree of reservoir study is assessed based on accuracy of obtained prediction results and significance of new investigations is evaluated to increase current degree of study of formation. Based on the evaluation results of the current study of the formation, the program and the volume of laboratory tests of the samples are corrected. System for method implementation comprises base (1) of formation survey data, module (2) for rock characteristics prediction, module (3) for evaluation of formation investigation degree and module (4) for program optimization and volume of laboratory tests of samples, designed to correct program and volume of laboratory studies based on evaluation of current degree of study of formation.EFFECT: technical result: providing fast and high-quality processing of laboratory research results with possibility of scaling these results on entire section of productive deposits of oil-and-gas bearing formation.8 cl, 4 dwg

Description

The invention relates to laboratory studies of rock samples taken from an oil and gas bearing formation and, in particular, to the optimization of laboratory studies of rock samples using machine learning methods and neural networks.

Laboratory research is a key element in the exploration, development and production of hydrocarbons, requiring substantial time and financial resources. To study the core of one well with an average offset of 100 linear meters, it usually takes up to six months. Mistakes in research planning and the loss of core material when using incorrect analysis methods are very expensive for oil companies - they increase the time until the first hydrocarbon production and lead to additional costs for exploring the deposits. In this regard, the urgent task is to optimize laboratory research in terms of automation of part of routine operations and extracting more practical information from ongoing research.

A known method that provides partial optimization of core research using digital core technologies described in the patent of the Russian Federation No. 2543698. This invention relates to methods for describing the characteristics of two-dimensional and three-dimensional samples to determine the distribution of pore body sizes and pore channels, as well as the curves of the dependence of capillary pressure in a porous medium. Input includes high resolution petrographic images and laboratory measurements of porosity. The output information includes the distribution of the sizes of the pore body and pore channels and the modeling of capillary pressure curves for both the pore body and pore channels. The disadvantage of this method is that only a particular solution is proposed for the optimization of laboratory core research.

The technical result achieved by the implementation of the invention is to provide quick and high-quality processing of laboratory research results with subsequent scaling of these results over the entire section of productive deposits of the oil and gas bearing formation, which allows you to control the degree of knowledge of the formation and formulate recommendations for adjusting the complex of laboratory studies. The proposed method and system for optimizing laboratory research of rock samples makes it possible to distribute rock characteristics over the entire section of productive sediments by predicting for each core sample the entire set of characteristics determined in the laboratory research complex, providing characteristics of the core collection parts without conducting laboratory tests and reducing time to obtain the necessary information on the characteristics of the core collection.

The specified technical result is achieved by the fact that in accordance with the proposed method for the optimization of laboratory research of rock samples taken from the oil and gas reservoir, only partial laboratory research of rock samples is carried out, the results of which are used to form a database for the study of reservoirs. Using the obtained results of studies of rock samples as input using machine learning algorithms, basic predictive models are created to predict the missing rock characteristics for each sample. The obtained results of studies of rock samples and the results of predicting the missing characteristics of rocks, obtained using basic predictive models, are used as input to create predictive models of the second group through machine learning to determine corrections for reservoir conditions and predict rock characteristics in thermobaric conditions. Using the obtained results of studies of rock samples, the results of predicting the missing characteristics of rocks obtained using basic predictive models, and the results of predicting the characteristics of rocks under thermobaric conditions obtained using predictive models of the second group, using machine learning, create predictive models as input data the third group for predicting the multidimensional characteristics of reservoir rocks. Using the created predictive models of all three groups, prediction of rock characteristics is carried out and all the obtained results of prediction of rock characteristics are entered into the database for the study of reservoirs. Assess the current degree of knowledge of the reservoir based on the accuracy of the predicted results and, based on the results of the assessment of the current degree of knowledge of the reservoir, adjust the program and the volume of laboratory research of the samples.

Additionally, the results of specialized studies, as well as the results of downhole studies, can be used as input data for creating predictive models. Adjustment of the program and volume of laboratory tests can be carried out in real time.

The system for optimizing laboratory research of rock samples taken from an oil and gas bearing formation, in accordance with the invention, contains a database for researching strata intended for storing standard results of laboratory studies of rock samples taken from an oil and gas bearing formation, the results of profile and downhole studies, two-dimensional and three-dimensional images of rocks and the results of their analysis, as well as the results of predicting rock characteristics, a module for predicting rock characteristics, intended for predicting rock characteristics, a module for assessing the degree of reservoir knowledge, designed to assess the current degree of knowledge of the reservoir based on the accuracy of the forecast results, and optimization of the program and volume of laboratory research of samples, designed to adjust the program and volume of laboratory research according to the results of the assessment of the current degree of knowledge layer.

The system for optimizing laboratory research of samples can additionally contain a preprocessor of the results of laboratory research, designed to form a single coordinated database of laboratory research from various sources of information.

The invention is illustrated by drawings, where in FIG. 1 shows a block diagram of a system for optimizing the results of laboratory studies of rock samples, FIG. 2 is a flowchart of a method for optimizing laboratory test results; FIG. 3 shows an example of the operation of the system for predicting certain rock characteristics, and in FIG. Figure 4 shows an example of a system for optimizing laboratory research of one of the characteristics of rocks.

Conducting a full range of laboratory tests takes a large amount of resources (financial and time). As a result, there are few lengthy and expensive definitions, but they are of primary importance for development design. Moreover, mass studies are often performed on different core samples and are not linked to each other. For example, the determination of capillary pressure curves, experiments on relative phase permeabilities (RPPs) and reservoir studies can be performed on various collections of core samples. At present, typification of reservoir rocks is usually used to relate data from various studies. Each collector type is assigned specific ranges or characteristic values of filtration-capacitive and other properties. This approach is simplified and significantly increases the level of uncertainties in the description of the reservoir system.

The proposed system for optimizing laboratory research of rock samples is a combination of hardware and software that provide the characteristics of parts of the core collection without laboratory tests and reduce the time to obtain the necessary information on the characteristics of the core collection. Reducing the time to obtain the necessary information on the characteristics of the core collection is due to the constant determination of the degree of knowledge of the reservoir and the formation of recommendations for adjusting the complex of laboratory studies. The system receives the results of part of the laboratory tests as input and generates new data on the reservoir system, and also evaluates the degree of knowledge of the productive section based on which the core research is optimized.

The system also performs automated control of the laboratory research cycle by constantly optimizing research programs and schedules. The system also issues recommendations for terminating studies in accordance with predefined criteria.

As shown in FIG. 1, the system contains a database 1 of reservoir research data intended for storing laboratory test results from various information sources. Database 1 contains the results of studies of rock samples taken from an oil and gas bearing formation, the results of profile and intra-well testing, two-dimensional and three-dimensional images of rocks and the results of their analysis, as well as the results of predicting the characteristics of rocks obtained by the proposed method.

The system also includes a module 2 for predicting rock characteristics. This module is designed to predict rock characteristics. He makes a forecast of some rock characteristics over others and forms a complete set of characteristics of each core sample selected for a complex of laboratory studies. In the process, module 2 uses machine learning algorithms, trained on the results of previous laboratory studies of core samples from various terrigenous and carbonate deposits. Module 2 allows you to determine the characteristics of rocks throughout the section of productive deposits by predicting for each sample the entire set of characteristics determined in the process of laboratory research.

Module 3 for assessing the degree of knowledge of the reservoir is intended to assess the current degree of knowledge of the reservoir based on the accuracy of the forecast obtained in module 2 of the forecasting results.

The system also contains module 4 for optimizing the program and volume of laboratory research of samples, intended to adjust the program and volume of laboratory research based on the results of evaluating the current degree of knowledge of the reservoir. The main objective of module 4 is the automated management of the laboratory research cycle from the development of primary programs and research schedules to their constant optimization and issuing recommendations for terminating research. Module 4 allows you to plan laboratory tests in automatic mode, which eliminates the influence of the human factor and speeds up the process of laboratory research.

The system for optimizing laboratory research of samples can additionally contain a preprocessor 5 of the results of laboratory research, designed to form a single coordinated database of laboratory research from various sources of information. The data entering database 1 can be a set of files having a different format, structure and degree of information content or detailing of research results. For example, the results of determining capillary pressure in one embodiment may include five pairs of capillary pressure-saturation values, and in another embodiment, seven pairs. Among other problems of data quality, one can distinguish the presence of a large number of omissions in the data (when a particular rock sample is characterized only by a part of the parameters), typos, errors in the recording formats of numerical values. To solve the problems described, the preprocessor 5 is designed - a module for computer processing the collected data, which performs computer processing of the collected data. Preprocessor 5 aggregates data from measurements of various formats into a single database using intelligent algorithms for approximate matching and error correction. Preprocessor 2 uses artificial intelligence methods to perform one or more actions selected from the group:

- Bringing the collected data to a single format, including alignment of dimensions and units;

- Identification using machine learning methods of data anomalies: gaps, outliers, incorrect values, facts of recalibration of sensors during the experiment.

Next, a laboratory optimization method described in FIG. 2

At the input of the system of FIG. 1, a primary research program is submitted. Here, taking into account the specified limitations (budget, time, equipment, etc.), research is planned and research schedules and roadmaps are generated, which are sent to the laboratory for implementation. In accordance with block 6, as laboratory tests of rock samples taken from the oil and gas bearing layer are performed, their results are continuously or with a certain frequency transferred to the formation study database 1 (database 1 in Fig. 1). In accordance with one embodiment of the invention, a preprocessor (preprocessor 5 in FIG. 1) prepares the results of laboratory tests (block 7 in FIG. 2) - forms a single consistent database of laboratory tests from various sources of information (files). Preprocessing also includes uploading and pairing data from various databases. Data preprocessing can be performed by searching and analyzing mnemonics in the data headers, semantic analysis of the text in the headers using artificial neural networks and using an interactive learning platform when the decision is made to correlate new data of a particular rock characteristic for the first time by a competent user system.

One of the functions of the unit for preprocessing laboratory research results 7 is to bring the collected data to a single format: determining units of measurement (based on statistical tests of the correspondence of the distribution of values to standard samples with known units), aligning dimensions and units of measurement (using translation tables of known dimensions and rules on based on trained decision trees in the general case), etc.

Another function of block 7 “detection of data anomalies” uses machine learning algorithms to determine the presence in the laboratory results of anomalies corresponding to gaps, outliers, incorrect values (for example, values many tens of times higher than physically reasonable values) and marks them in accordance with identified anomaly.

Then, according to block 8, the results of laboratory tests are fed to the input of a module for predicting rock characteristics (module 2 in Fig. 1). This module forms a cascade of basic predictive models based on machine learning algorithms (linear regressions, decision trees, gradient boosting, support vector method, artificial neural networks, etc.). Machine learning uses advanced statistical models to analyze existing data and implement the forecasting mechanism. Models based on various machine learning algorithms are trained on the provided sample data, which should include various features on the basis of which the forecast is made, and relevant relevant data (predicted characteristics) (see, for example, Luger D.F. Artificial Intelligence: strategies and Methods for Solving Complex Problems, 4th Edition, Williams Publishing House, 2003, pp. 369-482 or V. Vyugin. Mathematical Foundations of Machine Learning and Prediction. - Litres, 2017, pp. 295-306).

The first group of basic predictive models forming a cascade includes models that allow you to get the most accurate prediction of certain rock characteristics, using only information about core removal and the results of mass standard laboratory tests. As the data of standard studies can be:

- The depth of the roof and the sole of the reservoir;

- Porosity (by gas volumetric and hydrostatic methods),

- Permeability (absolute and according to Klinkenberg);

- Place of sampling;

- Density of the sample (mineralogical and volumetric);

- Cytological description (packing of grains, lithotype, color);

- Stratigraphy.

Using basic predictive models, rock characteristics such as:

- Residual water saturation (by centrifugation or by the method of a semipermeable membrane);

- Electrical resistivity (resistivity);

- The parameter of porosity;

- Effective permeability (for gas, water, kerosene);

- Velocities (interval time) of longitudinal and transverse elastic waves;

- Residual oil and gas saturation;

- Clay and carbonate;

- Specific surface of the pore space;

- The limits of compressive and tensile strength;

- Angle of internal friction and cohesion;

- Thermophysical characteristics (heat capacity, thermal conductivity, thermal diffusivity).

The second group of predictive models that form the cascade includes models that use not only the results of mass standard laboratory tests as input features, but also the forecast results of the models of the first group. The models of the second group are primarily used to determine corrections for reservoir conditions (PU) and allow to predict the following characteristics in thermobaric conditions:

- Residual water saturation in PU;

- Electrical resistivity (resistivity) in PU;

- The parameter of porosity in PU;

- Effective permeability (for gas, water, kerosene) in PU;

- Velocities (interval time) of longitudinal and transverse elastic waves in PU;

- Residual oil and gas saturation in PU.

The third group of predictive models forming a cascade includes models that use not only the results of mass standard laboratory tests, but also the forecast results of models of the first and second groups as input features. Models of the third group are primarily used to determine multidimensional characteristics of reservoir rocks, when for each lithological difference it is necessary to predict not one value, but a set of data (functional dependence of one characteristic on another, parameter distribution, parameter tensor, etc.). Among the multidimensional characteristics of reservoir rocks determined by the models of the third group, we can distinguish:

- Capillary pressure curve (pore size distribution) by a semipermeable membrane method;

- Capillary pressure curve (pore size distribution) by mercury porosimetry;

- Relative phase permeability;

- Distribution of grains by size (granulometry);

- Distribution of minerals and organic matter (x-ray diffractometry, pyrolysis).

A variety of predictive models is determined by the presence of a significant amount of data sufficient for learning machine learning algorithms. Another requirement for building accurate models is data representativeness. The high predictive ability of models is limited by the range of attributes in which they were trained.

Additionally, to increase the accuracy of predicting rock characteristics, as additional input data for models, the following specialized studies can be used: depth distribution of natural radioactivity, permeability, electrical resistivity, and acoustic characteristics; scratch test; thermophysical profiling; IR spectrometry.

As additional input data, the data of downhole studies can also be used: data of the downhole imager (FMI); data of geological and technological research (GTI) and data of geographic information systems (GIS).

As additional input data, 2D and 3D images of rocks and the results of their analysis can also be used:

- Photos of a full-sized core in daylight and UV light;

- Images and description of thin sections;

- Tomography of a full-sized core;

- Micro and nonotomography;

- Photographs of a scanning electron microscope (SEM).

All the predicted rock characteristics replenish the reservoir 1 database.

Then, in accordance with block 9 in FIG. 2, by means of the module for assessing the degree of knowledge of the formation (module 3 in Fig. 1), the current degree of knowledge of the formation and the significance of new studies are evaluated to increase the current degree of knowledge on the basis of the accuracy of prediction of the results of various types of studies.

VOI (Value of Information) or “value of information” is one of the important properties of information, the evaluation of which depends on the goals of the processes of its generation and processing. The VOI concept is considered by special information theories and decision theory (see, for example, A. Kharkevich, On the value of information, Problems of Cybernetics, 1960, Issue 4, p. 53, or Howard RA Information value theory, IEEE Transactions on systems science and cybernetics, 1966, v. 2, No. 1, pp. 22-26).

If the goal is achievable in several possible ways, VOI can be determined by the reduction of resource costs (material, time) brought by this information. The more information leads to the achievement of the goal, the more useful it is, the more VOI. Evaluation of VOI occurs in module 3 for assessing the degree of knowledge of the reservoir (module 3 in Fig. 1).

As the VOI, the accuracy of prediction of rock characteristics can be selected by the module for predicting the results of laboratory tests (module 2 in Fig. 1). As metrics for forecast accuracy, the following can be used: determination coefficient (R ² ), mean square error of the forecast (MSE); the average absolute forecast error (MAE), etc. In this option, VOI can be estimated independently for each type of research. The quantitative expression of VOI will depend on the requirements for accuracy in predicting performance. For example, let the necessary forecast accuracy of any characteristic be the value of R ₂ ² (R ₂ ² cannot be better than for a trained model). Then, upon reaching this condition, the VOI for any subsequent measurement will be zero. At the initial moment of time, the VOI value is maximum - any new information will allow us to evaluate the accuracy metrics of forecasting algorithms trained on historical data on analogues. Thus, VOI at the initial moment of time can be equated to unity at some initial value of R ₀ ² . Then for each i-th type of core research VOI at each step of the research optimization cycle can be estimated as

If for each rock sample selected in the current research cycle, the forecast of the characteristic under consideration satisfies the accuracy criterion, then VOI≤0, and in the new research program the corresponding type of research will be absent. In the case when VOI> 0, further training of the predictive model from the module for predicting the results of labs occurs. research on new data and issuing recommendations to expand the research program with additional experiments to determine the characteristics under consideration.

In another embodiment of the system, a confidence interval range may be selected as the VOI. In this case, the condition for the equality of VOI to zero will be a decrease in the range of the confidence interval after the cycle of retraining the predictive model to a given level.

In another embodiment of the system, the analysis of the VOI value in each optimization cycle is carried out separately for each type of rock. Rock typing can be performed according to various criteria. The most common typification is based on lithological differentiation of rocks. The adjustment of the research program in this case also occurs independently for each type, which allows additional studies to be performed more precisely and efficiently.

In accordance with block 10 in FIG. 2, depending on the current degree of knowledge of the reservoir, the laboratory research optimization system through the program optimization module and the volume of laboratory research of samples (module 4 in Fig. 1) continuously monitors and, if necessary, adjusts the laboratory research program and scope, as well as the work schedule. Schedules can be generated based on factors such as

- Expected research budget;

- Expected timelines for research;

- Available laboratory facilities;

- Scale of priorities for research;

- Schedules of core selection and delivery;

- The quality of the core;

- Scale of priorities for related facilities (formations, deposits);

- History of laboratory research on the object (if the program is not primary).

The system operates in an iterative mode: the adjustment of the program and research schedule occurs cyclically as new laboratory data arrives until a certain level of knowledge of the reservoir system is reached (VOI> 0). If VOI≤0, then the system gives a stop signal to terminate research in view of achieving a given level of knowledge of the reservoir system. If VOI> 0, then the system adjusts the research program taking into account the significance of each subsequent study. Based on the adjusted program, new research plans are also being formed.

Thus, the main data generated by the system during the optimization of laboratory tests include:

- The value of VOI separately as each type of research and each type of breed;

- An adjusted study program of the studied formation or a stop signal to stop research.

The proposed system for optimizing laboratory research of rock samples can be implemented using hardware and software. For the hardware of the implementation, the device must contain at least a processor and read-only memory (ROM). For the software part of the implementation, the proposed method for assessing rock characteristics can be a set of functional modules (functions or classes) written in programming languages (for example, C ++, Python). The program code may be stored in ROM blocks and retrieved by the processor for execution.

As an example of the implementation of the laboratory research optimization method, the results of laboratory research optimization of the following characteristics of core samples from the Achimov deposits of one of the fields in Western Siberia are presented:

- residual water saturation (Quo),

- electrical resistivity (resistivity),

- porosity parameter (Pn).

As the initial data transferred to block 6 as laboratory tests of rock samples were performed, the following were used:

- core removal,

- the roof of the reservoir,

- the bottom of the reservoir,

- open porosity by hydrostatic method,

- absolute gas permeability,

- coring depth,

- the density of the core sample,

- grit (from lithography),

- lithotype

- layer (age of deposits).

A large database was used to predict the Quo, resistivity, and RP, including the results of studies of more than 1200 samples of Achimov deposits. The dataset contained a large number of outliers and missing values. Data preprocessing in block 7 in FIG. 2 and the use of the gradient boosting algorithm (J. Friedman, "Greedy function approximation: A gradient boosting machine," vol. 29, pp. 1189-1232, 102001) allowed us to achieve high quality predictive models in block 8 in FIG. 2 Data preprocessing was performed by searching and analyzing mnemonics in the data headers and semantic analysis of the text in the headers.

The results of the prediction of Quo, resistivity and RP for selected core from a new well in one of the fields in Western Siberia are shown in FIG. 3, where 3a is the dependence of the experimental values of Quo on the predicted values, 3b is the dependence of the experimental values of resistivity on the predicted values, 3c is the dependence of the experimental values of Pn on the predicted values). The values of R ² for each of the three types of studies without special laboratory tests for the new core were equal:

- R ² = 0.903 for residual water saturation (Quo),

- R ² = 0.741 for electrical resistivity (resistivity),

- R ² = 0.879 for the porosity parameter (Pn).

To estimate the VOI, the required prediction accuracy of all three characteristics was the value of R ₂ ² = 0.8. Thus, upon reaching this condition, the VOI for any subsequent measurement will be less than or equal to zero. Let VOI at the initial moment of time be equal to unity at a value of R ₀ ² = 0. Then we get the following expression for evaluating the VOI (block 9 in Fig. 2):

The VOI values for subsequent special laboratory determinations of Quo, resistivity, and Pn according to the new core data in accordance with equation (2) were equal to:

- VOI = -0.13 for residual water saturation (Quo),

- VOI = 0.07 for electrical resistivity (resistivity),

- VOI = -0.1 for the porosity parameter (Pn).

Thus, for the new core it is recommended to conduct laboratory tests only for resistivity, for which VOI> 0. The accuracy of the prediction of Quo and RP using the optimization system of laboratory research of rock samples is high enough to abandon their special laboratory research.

Due to the fact that the accuracy of the resistivity forecast is also quite high, but less than the specified level (R ₂ ² = 0.8), it was recommended that only 10% of standard core samples from a new well be laboratory tested (block 10 in Fig. 2). The results were used to retrain the predictive model of resistivity. After retraining the gradient boosting algorithm on new data, the accuracy of the resistivity forecast increased to R ² = 0.801, and the value VOI = -0.001. In FIG. Figure 4 shows the dependences of the experimental resistivity values predicted without additional training (4a) and after additional training on the results of experimental determination of resistivity in 10% of samples from a new well (4b). Therefore, recommendations for additional laboratory definitions of resistivity were not issued, and the procedure for optimizing laboratory tests was terminated.

Claims (19)

1. A method for optimizing laboratory research of rock samples taken from an oil and gas bearing formation, in accordance with which: - carry out standard laboratory research of rock samples, the results of which are used to form a database for the study of reservoirs, - using as the input data the results of standard laboratory studies of rock samples using machine learning algorithms create basic predictive models to predict the missing rock characteristics for each sample, - using as input data the results of standard studies of rock samples and the results of forecasting the missing rock characteristics obtained using basic predictive models, using machine learning create predictive models of the second group to determine corrections for reservoir conditions and predict rock characteristics in thermobaric conditions , - using as input data the results of studies of rock samples, the results of predicting the missing characteristics of rocks, obtained using basic predictive models, and the results of predicting the characteristics of rocks in thermobaric conditions, obtained using predictive models of the second group, by means of machine learning create predictive third group models for predicting multidimensional characteristics of reservoir rocks, - using all the created predictive models carry out the prediction of rock characteristics and all the obtained results of prediction of rock characteristics are entered into the database for the study of strata, - evaluate the current degree of knowledge of the reservoir based on the accuracy of the obtained forecasting results and evaluate the significance of new studies to increase the current degree of knowledge of the reservoir, - according to the results of assessing the current degree of reservoir knowledge, the program and the volume of laboratory research of the samples are adjusted. 2. The method according to p. 1, according to which the results of specialized studies are additionally used as input data for creating predictive models. 3. The method according to p. 1, according to which the results of downhole studies are additionally used as input to create predictive models. 4. The method according to p. 1, according to which two-dimensional and three-dimensional images of rocks and the results of their analysis are additionally used as input to create predictive models. 5. The method according to p. 1, according to which, before creating the basic predictive models for predicting the missing rock characteristics, preprocessing of the results of laboratory studies is carried out to form a single consistent database for the study of reservoirs. 6. The method according to p. 1, according to which the adjustment of the program and the volume of laboratory tests is performed in real time. 7. A system for optimizing laboratory research of rock samples taken from an oil and gas bearing formation, comprising: - a reservoir research database intended for storing laboratory test results of rock samples taken from an oil and gas bearing reservoir, the results of profile and well testing, two-dimensional and three-dimensional images of rocks and the results of their analysis, as well as the results of predicting rock characteristics, - rock forecasting module, designed to predict rock characteristics, - a module for assessing the degree of knowledge of the reservoir, designed to assess the current degree of knowledge of the reservoir based on the accuracy of the predicted results, and assess the significance of new studies to increase the current degree of knowledge of the reservoir, - a module for optimizing the program and volume of laboratory research of samples, intended to adjust the program and volume of laboratory research based on the results of assessing the current degree of knowledge of the reservoir. 8. The system of claim 7, further comprising a preprocessor of the results of laboratory tests, designed to form a single consistent database of laboratory tests from various sources of information.

Images