Oil reservoir similarity calculation method based on ensemble learning algorithm

Abstract

The invention provides an oil reservoir similarity calculation method based on an ensemble learning algorithm. The oil reservoir similarity calculation method based on the ensemble learning algorithm comprises the steps that 1, an evaluation index system of similar oil reservoirs is established; 2, oil reservoir data are collected, and a similar oil reservoir sample set is established; step 3, calculating similarity by using multiple machine learning algorithms; 4, comprehensively calculating the similarity by adopting an ensemble learning algorithm; and 5, recommending similar oil reservoirs. According to the oil reservoir similarity calculation method based on the integrated learning algorithm, similar oil reservoirs of the target oil reservoir can be quickly and accurately found from a large number of development examples, successful development experience and development technologies are used for reference, the success rate of a target oil reservoir development scheme is increased, the recovery efficiency is increased, and more economic benefits are created.

Description

Technical Field

[0001] The present invention relates to the technical fields of oilfield development and artificial intelligence, and particularly to a method for calculating reservoir similarity based on an ensemble learning algorithm. Background Art

[0002] Oilfield development is a job with a very strong empirical nature. When business experts design reservoir development plans, drilling engineering, and oil production process technologies, they often need to refer to and draw on the successful development experiences and advanced technologies of developed reservoirs. However, developed reservoirs with different types and large property differences have little reference value, and only the development experiences and advanced technologies of developed reservoirs similar to the target reservoir are of reference significance. Therefore, it is particularly important to accurately and quickly find developed reservoirs similar to the target reservoir from a large number of development examples. In the past, when looking for similar reservoirs, some relied on expert experience, considered a few indicators for comparative analysis, and made qualitative judgments to select similar reservoirs, lacking a scientific basis for quantitative calculation and having low accuracy. Some used a model comprehensive evaluation method, and the calculation results completely depended on the judgment criteria. Therefore, it is of great significance to comprehensively consider the static indicators affecting reservoir development effects and use data-driven artificial intelligence algorithms to quickly and accurately find the most similar developed reservoir to the target reservoir.

[0003] In the Chinese patent application with the application number 2022106662912, it relates to a reservoir analogy method and device based on a similarity calculation model, including establishing static parameters and production dynamic parameters for oilfield evaluation using the parameters of established production capacity blocks, quantitatively analyzing the static parameters and production dynamic parameters to obtain a reservoir analogy oil and gas field database; establishing a target oilfield parameter table; preliminarily screening the oilfields in the reservoir analogy oil and gas field database according to the geological reservoir characteristics of the target oilfield through multi-attribute filtering settings; taking the static parameters and production dynamic parameters corresponding to the current time series as patterns, looking for similar patterns in historical data, and performing pattern matching using similarity to obtain a pattern matching model; predicting the reservoir characteristics of the target oilfield according to the parameters of the pattern matching model. This invention patent uses a method of multi-level filtering of empirical discrimination rules to optimize similar oilfields, with strong empiricism and lacking a scientific calculation basis.

[0004] In the Chinese patent application with the application number 2016101090288, a method for screening similar oilfields is involved, including: a benchmark parameter screening step of screening benchmark parameters according to the relationship between the analogy target and its influencing factors; an oilfield screening step of calculating the similarity between the sample oilfield and the analogy oilfield based on the benchmark parameters and screening the sample oilfield based on the similarity; a screening result verification step of verifying the reliability of the screening result according to the standard deviation of the benchmark parameters in the screened sample oilfield and the analogy oilfield. This patent uses an empirical method to establish the standard deviation range and calculates the similarity in a qualitative and quantitative combination manner. The method is relatively single and relies on expert experience, making it difficult to comprehensively reflect the similarity degree from multiple perspectives and difficult to naturally reflect the logical relationship between data.

[0005] In the paper "Methods and Applications of Quantitative Screening of Similar Oilfields" published on pages 766 - 771 of Volume 11, Issue 5, 2021 of the journal "Reservoir Evaluation and Development", the grey correlation and geological factor analysis methods are comprehensively applied. According to the essential correlation between the analogy parameters and the analogy target, the analogy parameters and weights are quantitatively determined. According to the reservoir type of the object under study, parameter thresholds are introduced to calculate the similarity of the sample oilfield, and the oilfields with high similarity are screened as reference samples for technical policy research and index prediction, making the objective basic conditions for analogy analysis more reasonable and improving the screening efficiency and the reliability of the analogy result. The method described in this literature is mainly an empirical discrimination rule, relying on expert experience, making it difficult to truly reflect the data correlation relationship and lacking a scientific calculation basis.

[0006] Through comprehensive analysis, it can be seen that the above existing technologies are quite different from the present invention and fail to solve the technical problems we want to solve. Therefore, we have invented a method for calculating reservoir similarity based on the ensemble learning algorithm. Summary of the Invention

[0007] The object of the present invention is to provide a method for calculating reservoir similarity based on the ensemble learning algorithm, which can improve the success rate of the development plan of the target reservoir, increase the recovery rate, and create more economic benefits.

[0008] The object of the present invention can be achieved by the following technical measures: A method for calculating reservoir similarity based on the ensemble learning algorithm, which includes:

[0009] Step 1, establish an evaluation index system for similar reservoirs;

[0010] Step 2, collect reservoir data and establish a sample set of similar reservoirs;

[0011] Step 3, calculate the similarity using multiple machine learning algorithms;

[0012] Step 4, comprehensively calculate the similarity using the ensemble learning algorithm;

[0013] Step 5, recommend similar reservoirs.

[0014] The object of the present invention can also be achieved by the following technical measures:

[0015] In step 1, the analytic hierarchy process is used to refine layer by layer to select evaluation indicators; the expert scoring method is used for comprehensive ranking to determine the weights.

[0016] In step 1, nine index parameters are selected from three aspects of reservoir properties, fluid properties, and temperature-pressure system. Among them, the selected reservoir physical property parameters are reservoir lithology, reservoir thickness, porosity, and permeability; the selected fluid property parameters are the density, viscosity, and original oil saturation of underground crude oil; the selected temperature-pressure system parameters are the original formation pressure and the original formation temperature.

[0017] In step 2, from different reservoir schemes, collect the index data of nine similar reservoir evaluation indicators to form a similar sample set.

[0018] In step 2, by establishing an oilfield word segmentation thesaurus and adopting document deep parsing technology, extract reservoir parameters from the scheme report to form a similar reservoir sample set.

[0019] Step 3 includes:

[0020] Step 31, use the Euclidean distance, that is, the absolute distance between two points in space, to calculate the similarity;

[0021] Step 32, use the Pearson correlation coefficient, that is, the linear correlation degree of two vectors, to calculate the similarity;

[0022] Step 33, use the Gaussian distance, that is, the standard Euclidean distance, to calculate the similarity;

[0023] Step 34, use the cosine angle, the smaller the vector angle, the more similar, to calculate the similarity;

[0024] Step 35, use the Jaccard distance, that is, the proportion of the number of different dimensions, to calculate the similarity.

[0025] In step 3, each algorithm uses a weight coefficient to multiply the index contribution to comprehensively consider the differences in the influence degrees of different indicators on the determination of similarity size, and comprehensively calculate the similarity.

[0026] In step 4, for the sorting results of each machine learning, an average fusion strategy and a bagging integration method are used for integration to obtain the final sorting result.

[0027] In step 5, according to the similarity sorting result, recommend reservoir development examples with higher similarity and draw on successful development experiences.

[0028] The object of the present invention can also be achieved by the following technical measures: A reservoir similarity calculation system based on an ensemble learning algorithm, characterized in that the reservoir similarity calculation system based on an ensemble learning algorithm uses a reservoir similarity calculation method based on an ensemble learning algorithm to find similar reservoirs of a target reservoir.

[0029] In the reservoir similarity calculation method based on an ensemble learning algorithm of the present invention, first, the similarities between the target reservoir and the developed reservoirs are calculated respectively through a variety of machine learning algorithms, and then an ensemble learning algorithm is used for comprehensive calculation to give a final conclusion, so as to quickly, accurately and intelligently recommend the developed reservoir most similar to the target reservoir, and refer to the development experience and advanced technologies of the developed reservoirs.

[0030] In the reservoir similarity calculation method based on an ensemble learning algorithm of the present invention, with the help of big data and data mining technologies, based on a similar sample set, through a variety of machine learning algorithms and an ensemble learning algorithm, it is possible to quickly and accurately find the developed reservoirs similar to the target reservoir and draw on successful development experiences. This intelligent recommendation method for similar reservoirs based on an ensemble learning algorithm uses the analytic hierarchy process to screen evaluation parameters, and adopts document deep parsing and key parameter automatic extraction technologies to automatically extract reservoir parameters, avoiding the huge manual workload brought by manual collection of reservoir parameters. Five machine learning algorithms are used to calculate the similarity, making full use of the advantages of various methods, and an ensemble learning algorithm is used for comprehensive calculation to intelligently recommend similar reservoirs. The recommended results are considered reliable after being reviewed by business experts.

[0031] The present invention can quickly and accurately find similar reservoirs of the target reservoir from a large number of development examples, draw on successful development experiences and technologies, improve the success rate of the development plan of the target reservoir, increase the recovery rate, and create more economic benefits.

[0032] Compared with the traditional empirical judgment method, this result has a more scientific basis for quantitative calculation. By searching from a large number of similar sample sets, the range of comparable similar reservoirs is wider, rather than just finding similar reservoirs in the reservoirs familiar to the experts themselves. The found similar reservoirs are more accurate and have greater reference significance. Brief Description of the Drawings

[0033] Figure 1 It is a flowchart of a specific embodiment of the reservoir similarity calculation method based on an ensemble learning algorithm of the present invention. Detailed Description of the Invention

[0034] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0035] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular forms are also intended to include the plural forms. In addition, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they specify the presence of features, steps, operations, and / or combinations thereof.

[0036] As Figure 1 shown, Figure 1 is a flowchart of the reservoir similarity calculation method based on the ensemble learning algorithm of the present invention. The reservoir similarity calculation method based on the ensemble learning algorithm includes:

[0037] Step 1: Establish an evaluation index system for similar reservoirs;

[0038] Step 2: Collect reservoir data and establish a similar reservoir sample set;

[0039] Step 3: Calculate the similarity using multiple machine learning algorithms;

[0040] Step 4: The ensemble learning algorithm comprehensively calculates the similarity;

[0041] Step 5: Intelligently recommend similar reservoirs.

[0042] In Step 1, the analytic hierarchy process is used to refine layer by layer and select evaluation indicators. The expert scoring method is used for comprehensive ranking to determine the weights. Nine index parameters are selected from three aspects of reservoir properties, fluid properties, and temperature-pressure systems. Among them, the selected reservoir physical property parameters are reservoir lithology, reservoir thickness, porosity, and permeability. The selected fluid properties are the density, viscosity, and original oil saturation of underground crude oil. The selected temperature-pressure system parameters are the original formation pressure and the original formation temperature.

[0043] In Step 2, from different reservoir schemes, the index data of nine similar reservoir evaluation indicators are automatically collected to form a similar sample set. By establishing an oilfield word segmentation thesaurus and adopting document deep parsing technology, reservoir parameters are automatically extracted from the scheme report to form a similar reservoir sample set.

[0044] Step 3 includes:

[0045] Step 31: Calculate the similarity using the Euclidean distance (i.e., the absolute distance between two points in space);

[0046] Step 32: Calculate the similarity using the Pearson correlation coefficient (i.e., the linear correlation degree between two vectors);

[0047] Step 33: Calculate the similarity using the Gaussian distance (standard Euclidean distance);

[0048] Step 34, calculate the similarity using the cosine angle (the smaller the vector angle, the more similar).

[0049] Step 35, calculate the similarity using the Jaccard distance (the proportion of the number of different dimensions).

[0050] In steps 31, 32, 33, 34, and 35, for each algorithm, the weight coefficient is multiplied by the index contribution to comprehensively consider the differences in the influence degrees of different indicators on the determination of similarity size, and the similarity is calculated comprehensively.

[0051] In step 4, for the sorting results of each machine learning, an average fusion strategy and a bagging integration method are used for integration to obtain the final sorting result.

[0052] In step 5, according to the similarity sorting result, reservoir development examples with higher similarity are intelligently recommended to draw on successful development experiences.

[0053] The following are several specific embodiments of applying the present invention

[0054] Embodiment 1

[0055] In a specific embodiment 1 of applying the present invention, the reservoir similarity calculation method based on the ensemble learning algorithm includes the following steps:

[0056] 1. Establish an evaluation index system for similar reservoirs, including 9 indicators such as reservoir lithology, reservoir thickness, porosity, permeability, density of underground crude oil, viscosity, original oil saturation, original formation pressure, and original formation temperature.

[0057] 2. Collect the index data of developed reservoirs and construct a similar sample data set. Automatically extract the 9 index data of each reservoir from a large number of project reports and merge them together to form a similar sample data set.

[0058] 3. Calculate the similarity using 5 machine learning algorithms respectively. Extract the 9 index parameters of the target reservoir and calculate the similarity between the target reservoir and each reservoir in the similar sample data set using five machine learning algorithms such as Euclidean distance, Pearson correlation coefficient, Gaussian distance, cosine angle, and Jaccard distance. After standardizing the data when calculating the similarity, in order to consider the differences in the influence degrees of different indicators, multiply the weight of each indicator by the contribution of each indicator.

[0059] 4. The ensemble learning algorithm comprehensively calculates the similarity. Use the average algorithm to comprehensively calculate the similarity between the target reservoir and each developed reservoir and perform similarity sorting.

[0060] 5. Intelligently recommend similar reservoirs. According to the similarity sorting, intelligently recommend the top 5 similar reservoirs for reference by the target reservoir.

[0061] Example 2

[0062] In a specific Example 2 of the application of the present invention, as Figure 1 shown, Figure 1 is a flowchart of the intelligent recommendation method for similar reservoirs based on the ensemble learning algorithm of the present invention. The reservoir similarity calculation method based on the ensemble learning algorithm includes:

[0063] Step 110, establish an evaluation index system for similar reservoirs. The present invention selects 9 indicators including reservoir lithology, reservoir thickness, porosity, permeability, density of underground crude oil, viscosity, original oil saturation, original formation pressure, and original formation temperature.

[0064] Step 120, collect data to establish a similar sample set.

[0065] In the case of the present invention, 200 developed reservoirs of water - drive sandstone reservoirs are selected. After extracting the index parameters, 198 reservoirs are determined to be effective samples, forming a similar reservoir sample set. One test target reservoir is selected. The present invention can be implemented using sample data sets and target reservoirs different from the examples, and should not be construed as limited to the implementation examples presented here.

[0066] Step 130, calculate the similarity using 5 methods.

[0067] In order to consider the differences in the influence degrees of different parameters on the judgment results of similar reservoirs, a weight coefficient is set for each index. The weight coefficients are determined by calling reservoir business experts in the field using the analytic hierarchy process and finally obtained through comprehensive scoring. The comprehensive calculated weight coefficients of the 9 indicators of reservoir lithology, reservoir thickness, porosity, permeability, density of underground crude oil, viscosity, original oil saturation, original formation pressure, and original formation temperature are 0.105, 0.045, 0.245, 0.135, 0.045, 0.1125, 0.1125, 0.14, and 0.06 in sequence.

[0068] Step 140, calculate the similarity using the ensemble learning algorithm.

[0069] The average method is used to integrate and calculate the ranking scores, and the final ranking scores are obtained through integration. The ranking score of each algorithm is equal to 198 minus the similarity order. The final ranking scores are re - ranked to obtain the similarity order.

[0070] Step 150, intelligently recommend similar reservoirs. The top five reservoirs in the ranking are preferably selected to automatically recommend the best reference similar reservoirs.

[0071] Example 3

[0072] In a specific Example 3 of the application of the present invention, as Figure 1 shown, Figure 1This is a flowchart of the intelligent recommendation method for similar reservoirs based on the ensemble learning algorithm of the present invention. The reservoir similarity calculation method based on the ensemble learning algorithm includes:

[0073] Step 110, establish an evaluation index system for similar reservoirs. The present invention selects 9 indicators including reservoir lithology, reservoir thickness, porosity, permeability, density of underground crude oil, viscosity, original oil saturation, original formation pressure, and original formation temperature.

[0074] Step 120, collect data to establish a similar sample set.

[0075] In the case of the present invention, 168 developed heavy oil reservoirs are selected. After extracting the index parameters, 159 reservoirs are determined to be valid samples, forming a similar reservoir sample set. A test target reservoir is selected. The present invention can be implemented using a sample data set and a target reservoir different from the examples, and should not be construed as being limited to the implementation examples presented here.

[0076] Step 130, calculate the similarity using 5 methods.

[0077] In order to consider the differences in the influence degrees of different parameters on the judgment results of similar reservoirs, a weight coefficient is set for each indicator. The weight coefficient is determined by convening reservoir business experts in the field using the analytic hierarchy process and finally obtained through comprehensive scoring. The weight coefficients of the 9 indicators of reservoir lithology, reservoir thickness, porosity, permeability, density of underground crude oil, viscosity, original oil saturation, original formation pressure, and original formation temperature are 0.105, 0.045, 0.235, 0.145, 0.035, 0.1225, 0.1125, 0.13, and 0.07 in sequence.

[0078] Step 140, calculate the similarity using the ensemble learning algorithm.

[0079] The average method is used to integrate and calculate the ranking scores, and the final ranking scores are obtained through integration. The ranking score of each algorithm is equal to 159 minus the similarity order. The final ranking scores are reordered to obtain the similarity order.

[0080] Step 150, intelligently recommend similar reservoirs. Select the top five reservoirs in the ranking and automatically recommend the best reference similar reservoirs.

[0081] Finally, it should be noted that the above are only the preferred embodiments of the present invention and are not used to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

[0082] Except for the technical features described in the specification, all are well-known technologies to those skilled in the art.

Claims

1. A method for calculating reservoir similarity based on an ensemble learning algorithm, characterized in that The reservoir similarity calculation method based on the ensemble learning algorithm includes: Step 1: Establish an evaluation index system for similar reservoirs; Step 2: Collect reservoir data and establish a similar reservoir sample set; Step 3: Calculate the similarity using multiple machine learning algorithms; Step 4: Comprehensively calculate the similarity using the ensemble learning algorithm; Step 5: Recommend similar reservoirs.

2. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 1, wherein, In Step 1, the analytic hierarchy process is used to refine layer by layer to select evaluation indicators; the expert scoring method is used for comprehensive ranking to determine the weights.

3. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 2, wherein In Step 1, nine index parameters are selected from three aspects of reservoir properties, fluid properties, and temperature-pressure systems. Among them, the selected reservoir physical property parameters are reservoir lithology, reservoir thickness, porosity, and permeability; the selected fluid property parameters are the density, viscosity, and original oil saturation of underground crude oil; the selected temperature-pressure system parameters are the original formation pressure and the original formation temperature.

4. The reservoir similarity calculation method based on the integrated learning algorithm according to claim 1, characterized in that In Step 2, from different reservoir schemes, collect the index data of nine similar reservoir evaluation indicators to form a similar sample set.

5. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 4, characterized in that In Step 2, by establishing an oilfield word segmentation thesaurus and adopting document deep parsing technology, extract reservoir parameters from the scheme report to form a similar reservoir sample set.

6. The method for calculating reservoir similarity based on the ensemble learning algorithm according to claim 1, wherein Step 3 includes: Step 31: Calculate the similarity using the Euclidean distance, that is, the absolute distance between two points in space; Step 32: Calculate the similarity using the Pearson correlation coefficient, that is, the linear correlation degree of two vectors; Step 33: Calculate the similarity using the Gaussian distance, that is, the standard Euclidean distance; Step 34: Calculate the similarity using the cosine angle. The smaller the vector angle, the more similar; Step 35: Calculate the similarity using the Jaccard distance, that is, the proportion of the number of different dimensions.

7. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 6, characterized in that, In Step 3, each algorithm comprehensively calculates the similarity by multiplying the weight coefficient by the index contribution to comprehensively consider the differences in the influence degrees of different indicators on the similarity size discrimination.

8. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 1, wherein In Step 4, for the sorting results of each machine learning, an average fusion strategy and the bagging ensemble method are used for integration to obtain the final sorting result.

9. The method for calculating reservoir similarity based on the integrated learning algorithm according to claim 1, wherein In Step 5, according to the similarity sorting result, recommend reservoir development examples with higher similarity and draw on successful development experiences.

10. A reservoir similarity calculation system based on an ensemble learning algorithm, characterized in that, The reservoir similarity calculation system based on the ensemble learning algorithm uses the reservoir similarity calculation method described in any one of claims 1-9 to find similar reservoirs of the target reservoir.

Images