Reserve growth trend prediction method based on big data analysis

Abstract

The invention provides a big data analysis-based reserve growth trend prediction method. The big data analysis-based reserve growth trend prediction method comprises the steps of 1, performing basic data acquisition and database establishment; 2, a fitting function suitable for the growth mode is judged through regression analysis; 3, correlation analysis is carried out, and parameter sensitivity detection is carried out through clustering analysis; 4, carrying out cross validation on the reserved verification data and a prediction model result; and 5, predicting the reserve growth trend. According to the big data analysis-based reserve growth trend prediction method, the requirement on the empirical data volume is reduced, batch establishment and data visualization of a big data analysis library and automation and dynamic of reserve prediction are realized, and the accuracy of short-term prediction of proved reserves is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of prediction of oil reservoir reserve growth, and particularly to a method for predicting the reserve growth trend based on big data analysis. Background Art

[0002] Currently, domestic and foreign scholars have conducted extensive research on methods for predicting the oil reservoir reserve growth trend. Zhu Jie (Journal of Xi'an Shiyou University (Natural Science Edition), 2002, P21 - 23) compared and analyzed representative models such as the Weng's cyclic life model method, Gompertz, Logistic, and Gaussian methods. Different methods have their own advantages and disadvantages. These methods mainly propose mathematical models to predict the reserve growth trend based on reserve growth trend analysis, the change trend analysis of the ratio of reserves to the number of exploration wells, etc. They cannot reflect the characteristics of the current exploration complexity of hydrocarbon-bearing basins.

[0003] In the Chinese patent application with the application number: CN201810855341.5, it involves a method for predicting oilfield exploration reserves based on a recurrent neural network. This method is based on a recurrent neural network, eliminating the drawbacks of traditional artificially designed prediction equations, and using a computer to learn the most suitable prediction model for the reserve data from a large amount of historical data to achieve accurate prediction of oilfield exploration reserves.

[0004] In the Chinese patent application with the application number: CN201810141112.7, it involves a method for predicting the annual recurrence law of the S-shaped curve of proved reserve growth. This method for predicting the annual recurrence law of the S-shaped curve of proved reserve growth establishes a prediction based on resource volume, the number of discovered oil reservoirs, and reserves of exploration wells, and is more suitable for medium - and short - term predictions of different levels of hydrocarbon systems such as basin - level, sag - level, and zone - level with a higher exploration degree.

[0005] In the Chinese patent application with the application number: CN202010698284.1, it involves a method and device for predicting oil and gas production. The method includes: obtaining characteristic data of cores collected from a target reservoir, seepage data in the target reservoir measured according to the cores, oil - bearing data in the target reservoir measured according to the cores, and engineering data for oil and gas production planned for the target reservoir; predicting the oil and gas production in the target reservoir through a spectral clustering model based on the characteristic data, seepage data, oil - bearing data, and engineering data, where the spectral clustering model is trained according to sample data corresponding to sample cores, and the sample data at least includes: sample characteristic data, sample seepage data, sample oil - bearing data, sample engineering data, and sample oil and gas production data corresponding to the sample cores. This invention solves the technical problem that the prediction of oil and gas production in the related art is time - consuming and costly.

[0006] However, the accuracy of the mathematical model methods commonly used in predicting the growth trend of proven reserves is relatively low. The premise of their application is to freeze the level of oil exploration and development and make predictions in a relatively isolated closed system at the end of the data period. That is, for a long period of time, there is no significant leap in geological theory and exploration supporting technology, no significant change in the relevant policies of the superior departments and the oil fields themselves, and no qualitative change in the understanding of resources. The more obvious the difference between this idealized simple situation and the actual complex situation, the lower the reference value of the prediction results.

[0007] Big data analysis has been widely used in oil and gas exploration and development, but it has not been effectively used in reserve prediction. In order to realize the digital oil and gas field management of big data, it needs to be combined with other applications in oil and gas exploration and development. To this end, we have invented a new method for predicting reserve growth trends based on big data analysis. Summary of the invention

[0008] The purpose of the present invention is to provide a reserve growth trend prediction method based on big data analysis, so as to reduce the requirements of traditional reserve prediction methods on empirical data and improve the accuracy of short-term and medium-term prediction of proven reserves.

[0009] The object of the present invention can be achieved by the following technical measures: a method for predicting the reserve growth trend based on big data analysis, the method for predicting the reserve growth trend based on big data analysis comprising:

[0010] Step 1: Collect basic data and build a database;

[0011] Step 2: Use regression analysis to determine the appropriate fitting function for the growth model;

[0012] Step 3: perform correlation analysis and parameter sensitivity detection through cluster analysis;

[0013] Step 4, cross-validate the prediction model results with the retained validation data;

[0014] Step 5: Predict the reserve growth trend.

[0015] The purpose of the present invention can also be achieved by the following technical measures:

[0016] In step 1, data is collected, sorted, cleaned, and converted from different data to generate a new data set.

[0017] In step 1, the data set sources include the annual newly added reservoir reserve parameter table, exploration well workload table, and seismic workload table database.

[0018] In step 1, a dataset is created using the MySQL database.

[0019] In step 1, an outlier recognition method is adopted to screen and denoise the database.

[0020] In step 1, the established dataset is applied to access and process relevant data: functions such as drawing specified data, panning with the left and right mouse buttons, zooming, box selection, deleting points by fuzzy selection, and displaying corresponding coordinates are realized.

[0021] In step 2, a regression analysis simulation based on big data analysis technology is carried out to judge the fitting function applicable to the growth mode.

[0022] In step 2, the model is established based on the scatter plot established from the cumulative number of oil reservoirs and the semi-logarithm of the proven reserves collected in step 1.

[0023] In step 2, the regression fitting analysis model applied is:

[0024] Positive S-shaped: y = 1 / (α + β × exp(γ × x)) (3)

[0025] S-shaped: y = 1 / (α + exp(β + γ × ln(x))) (4)

[0026] Partial S-shaped: y = 1 / (α + exp(β + γ × (x) 0.5 )) (5)

[0027] Among them, y is the semi-logarithm of the proven reserves, x is the cumulative number of oil reservoirs, and α, β, and γ are the equation coefficients.

[0028] In step 3, the correlation analysis technology in big data analysis technology is used to perform multi-parameter sensitivity detection on the regression fitting analysis model established in step 2.

[0029] In step 3, first, the most likely values of all parameters affecting reserve growth are predicted, and then the impact of all or some parameters changing simultaneously on the reserve growth objective function is investigated.

[0030] The analysis model is:

[0031]

[0032]

[0033] OF: is a function that defines the degree of mismatch between historical actual data and predicted data; historical fitting objective function; Mi: comprehensive sensitive parameter variance; w i : weight coefficient; a 0 ,a w ,a g ,a p : weight coefficients of each sensitive parameter; Predicted data of each sensitive parameter; Historical actual data of each sensitive parameter;

[0034] Through cluster analysis, the similarity of data and the characteristics of different categories are mined. The system uses the K-Means algorithm to perform cluster analysis on data samples and obtains evaluation indicators, mainly the silhouette coefficient and the nearest cluster.

[0035] In step 3, neural network technology is applied to mine the relationship between each pattern parameter and the cluster data and its macroscopic characteristics. The deep feedforward neural network DFF is used to predict the model parameters, and a broom-shaped graph can be generated.

[0036] In step 3, the sensitivity parameters of reserve growth are studied through three operations: multi-parameter sensitivity analysis, parameter sensitivity detection of the cluster analysis results, and parameter sensitivity detection of neural network and regression analysis.

[0037] In step 4, further predict the newly added proven reserves in a certain year reserved in advance by adjusting the newly added 3D seismic area and the number of exploration wells, and verify the fitting degree of the prediction model.

[0038] In step 5, based on the model selected in step 2 and considering the values of the sensitive parameters analyzed in step 3, predict the reserve growth trend in the study area for a period of time in the future.

[0039] The reserve growth trend prediction method based on big data analysis in the present invention solves the drawbacks that the traditional prediction model is closely dependent on the technical environment and development strategy in the data collection stage, reduces the requirement for the amount of empirical data, realizes the batch establishment of the big data analysis library and data visualization, the automation and dynamicization of reserve prediction, and improves the accuracy of short-term prediction of proven reserves. Brief Description of the Drawings

[0040] Figure 1 It is a flowchart of a specific embodiment of the reserve growth trend prediction method based on big data analysis of the present invention. Detailed Description of the Invention

[0041] It should be noted that the following detailed description is exemplary and is intended to provide further illustration 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.

[0042] 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.

[0043] To make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0044] The method for predicting the reserve growth trend based on big data analysis provided by the present invention, on the basis of the big data analysis method system, for example Figure 1 as shown Figure 1 is a flowchart of the method for predicting the reserve growth trend based on big data analysis of the present invention. The method for predicting the reserve growth trend based on big data analysis includes:

[0045] Step 101, collecting and building a database of basic data;

[0046] Step 102, using regression analysis to determine the fitting function applicable to the growth mode;

[0047] Step 103, correlation analysis, and performing parameter sensitivity detection through cluster analysis;

[0048] Step 104, cross-validating with the reserved verification data and the prediction model results;

[0049] Step 105, predicting the reserve growth trend.

[0050] The following are specific embodiments of applying the present invention

[0051] Embodiment 1

[0052] In a specific Embodiment 1 of applying the present invention, the method for predicting the reserve growth trend based on big data analysis includes the following steps:

[0053] In Step 1, data is collected. After collecting, sorting, cleaning, and transforming data from different sources, a new data set is generated.

[0054] The data set sources include the annual new oil reservoir reserve parameter table, the exploration well workload table, and the seismic workload table database.

[0055] Considering that the MySQL database has advantages such as being open-source, high-speed, cross-platform, high security, and high reliability, the present invention uses MySQL to establish the data set.

[0056] The outlier recognition method (Isolation Forest Method) is used to screen and denoise the database.

[0057] With the established dataset, relevant data can be accessed and processed: functions such as drawing specified data, panning with the left and right mouse buttons, zooming in and out, box selection, deleting points by fuzzy selection, and displaying corresponding coordinates can be realized.

[0058] In step 2, based on the regression analysis simulation of big data analysis technology, the fitting function applicable to the growth mode is judged.

[0059] The model is established based on the scatter plot established by the cumulative number of oil reservoirs and the semi-logarithm of proven reserves collected in step 1.

[0060] The regression fitting analysis model applied is:

[0061] Positive "S" type: y = 1 / (α + β × exp(γ × x)) (3)

[0062] "S" type: y = 1 / (α + exp(β + γ × ln(x))) (4)

[0063] Partial "S" type: y = 1 / (α + exp(β + γ × (x) 0.5 )) (5)

[0064] Among them, y is the semi-logarithm of proven reserves, x is the cumulative number of oil reservoirs, and α, β, and γ are the equation coefficients.

[0065] In step 3, the multi-parameter sensitivity detection of the regression fitting analysis model established in step 2 is carried out by using the correlation analysis technology in big data analysis technology.

[0066] First, predict the most likely values of all parameters affecting reserve growth, and then examine the impact of the simultaneous change of all or part of the parameters on the reserve growth objective function.

[0067] The analysis model is:

[0068]

[0069]

[0070] OF: is a function that defines the degree of mismatch between historical actual data and predicted data; historical fitting objective function; Mi: comprehensive sensitive parameter variance; w i : weight coefficient; a 0 ,a w ,a g ,a p : weight coefficients of each sensitive parameter; Predicted data of each sensitive parameter; Historical actual data of each sensitive parameter.

[0071] Through cluster analysis, the similarity of data and the characteristics of different classes are mined. The system uses the K-Means algorithm to perform cluster analysis on data samples and obtains evaluation indicators, mainly the silhouette coefficient and the nearest cluster.

[0072] Apply neural network technology to mine the relationship between each pattern parameter and the cluster data and its macroscopic characteristics. Use a deep feedforward neural network (DFF) to predict model parameters and generate a broom-shaped graph. The implementation of the DFF algorithm is as follows:

[0073] The data of each field is normalized before input to improve the efficiency of neural network gradient descent calculation. Automatic mini-batch normalization is used, and the random dropout method is adopted in multiple hidden layers. For each unit, any one of the feature activation units in the previous layer may be discarded, so as not to overly rely on a certain one or several feature activation units. Therefore, the weights can be evenly distributed to each activation unit to avoid overfitting; use Adaptive Moment Estimation and learning rate decay to optimize the model.

[0074] Conduct research on the sensitivity parameters of reserve growth through three operations: multi-parameter sensitivity analysis, parameter sensitivity detection of the cluster analysis results, and parameter sensitivity detection of neural networks and regression analysis.

[0075] In step 4, cross-validate with the reserved validation data and the prediction model results.

[0076] Further predict the newly added proven reserves in a certain year reserved in advance by adjusting the newly added 3D seismic area and the number of exploration wells, and verify the fitting degree of the prediction model.

[0077] In step 5, predict the reserve growth trend in the study area.

[0078] Based on the model selected in step 2 and considering the values of the sensitive parameters analyzed in step 3, predict the reserve growth trend in the study area for a period of time in the future.

[0079] Embodiment 2

[0080] In a specific embodiment 2 of applying the present invention, the reserve growth trend prediction method based on big data analysis of the present invention includes:

[0081] Step 1, collect the reserve, exploration well foundation, exploration well oil layer, exploration well oil testing, reserve-exploration well database, and seismic workload data in the JY area. After collecting, sorting, cleaning, and converting the collected data using MySQL, generate a new data set.

[0082] The established data set can be used to access and process relevant data, realizing functions such as drawing specified data, panning with the left and right mouse buttons, zooming in and out, box selection, deleting points by fuzzy selection, and displaying corresponding coordinates.

[0083] Step 2: Establish a scatter plot of the cumulative number of reservoirs and the semi-logarithm of the proven reserves collected in Step 1, use regression analysis to simulate and determine the fitting function applicable to the growth pattern, and output a broom-shaped diagram.

[0084] Step 3: Use the correlation analysis technology in big data analysis technology to perform multi-parameter sensitivity detection on the regression fitting analysis model established in Step 2.

[0085] The research on the sensitivity parameters of reserve growth is mainly carried out through three operations: multi-parameter sensitivity analysis, parameter sensitivity detection of the clustering analysis results, and parameter sensitivity detection of neural networks and regression analysis. By analyzing the seismic work area and the number of exploration wells in the JY area, it shows a positive feedback to the objective function.

[0086] Table 1 Parameter Sensitivity Analysis Table for JY Area

[0087]

[0088] Step 4: Cross-validate with the reserved validation data and the results of the prediction model.

[0089] Furthermore, by adjusting the newly added 3D seismic area and the number of exploration wells, the newly added proven reserves for a certain year are predicted. Table 2 shows the fitting prediction for 2019. The fitting combination shows that the fitting result of the fifth group of parameters is 2680×10 4 t, which is in good agreement with the actual newly added proven reserves of 2450×10 4 t in 2019.

[0090] Table 2 Fitting Result Table of Different Sensitive Parameters

[0091]

[0092]

[0093] Step 5: Based on the model selected in Step 2, considering the values of the sensitive parameters analyzed in Step 3, predict the reserve growth trend in the JY area for the next ten years.

[0094] Table 3 Prediction Result Data Table Using the Fifth Group of Data (2021 - 2030)

[0095] Year <![CDATA[Newly proven reserves (×10 4 t)]]> 2021 4569 2022 4751 2023 4753 2024 4801 2025 4883 2026 4894 2027 4987 2028 5038 2029 5242 2030 5325

[0096] Example 3

[0097] In Specific Embodiment 3 of applying the present invention, the method for predicting the reserve growth trend based on big data analysis of the present invention includes:

[0098] Step 1: Collect the reserve, exploration well foundation, exploration well oil layer, exploration well oil testing, reserve-exploration well database, and seismic workload data of the DY Sag. After collecting, sorting, cleaning, and converting the collected data using MySQL, generate a new data set.

[0099] The established data set can realize the access and processing of relevant data: realize functions such as drawing specified data, panning with the left and right mouse buttons, zooming, box selection, fuzzy point selection and deletion, and corresponding coordinate display.

[0100] Step 2: Establish a scatter plot of the cumulative number of reservoirs and the semi-logarithm of the proven reserves collected in Step 1, use regression analysis to simulate and judge the fitting function applicable to the growth mode, and output a broom-shaped diagram.

[0101] Step 3: Use the correlation analysis technology in big data analysis technology to perform multi-parameter sensitivity detection on the regression fitting analysis model established in Step 2.

[0102] The research on the reserve growth sensitivity parameters is mainly carried out through three operations: multi-parameter sensitivity analysis, parameter sensitivity detection of the clustering analysis results, and parameter sensitivity detection of neural networks and regression analysis. By analyzing the seismic work area and the number of exploration wells in the DY area, it is found that they tend to have a positive feedback on the objective function.

[0103] Table 4 Parameter Sensitivity Analysis Table of the DY Sag

[0104]

[0105] Step 4: Cross-validate with the reserved verification data and the results of the prediction model.

[0106] Furthermore, by adjusting the newly added 3D seismic area and the number of exploration wells, predict the newly added proven reserves in a certain year. The fitting combination shows that the fitting result of the sixth group of parameters is 820×10 4 t, which is relatively consistent with the actual newly added proven reserves of 845×10 4 t in 2019.

[0107] Table 5 Fitting Result Table of Different Sensitive Parameters

[0108]

[0109] Step 5: Based on the model selected in Step 2, considering the values of the sensitive parameters analyzed in Step 3, predict the reserve growth trend in the DY area in the next ten years.

[0110] Table 3 Prediction Result Data Table Using the Sixth Group of Data (2021 - 2030)

[0111]

[0112]

[0113] 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.

[0114] Except for the technical features described in the specification, the rest are well-known technologies to those skilled in the art.

Claims

1. A method for predicting the trend of reserve growth based on big data analysis, characterized in that, the method for predicting the trend of reserve growth based on big data analysis includes: Step 1, collect and build a database of basic data; Step 2, use regression analysis to judge the fitting function applicable to the growth pattern; Step 3, conduct correlation analysis, and through cluster analysis, conduct parameter sensitivity detection; Step 4, cross-validate with the reserved verification data and the results of the prediction model; Step 5, predict the trend of reserve growth.

2. The method for predicting the trend of reserve growth based on big data analysis according to claim 1, characterized in that, in Step 1, collect data, and after collecting, sorting, cleaning, and transforming data from different sources, generate a new data set.

3. The method for predicting the trend of reserve growth based on big data analysis according to claim 2, characterized in that, in Step 1, the data set sources include the parameter table of newly added oil reservoir reserves over the years, the exploration well workload table, and the seismic workload table database.

4. The method for predicting the trend of reserve growth based on big data analysis according to claim 3, characterized in that, in Step 1, use the MySQL database to establish the data set.

5. The method for predicting the trend of reserve growth based on big data analysis according to claim 4, characterized in that, in Step 1, adopt an outlier identification method to screen and denoise the database.

6. The method for predicting the trend of reserve growth based on big data analysis according to claim 5, characterized in that, in Step 1, apply the established data set to access and process relevant data: realize functions such as drawing specified data, panning with the left and right mouse buttons, zooming, box selection, deleting points by fuzzy selection, and displaying corresponding coordinates.

7. The method for predicting the trend of reserve growth based on big data analysis according to claim 1, characterized in that, in Step 2, conduct regression analysis simulation based on big data analysis technology to judge the fitting function applicable to the growth pattern.

8. The method for predicting the trend of reserve growth based on big data analysis according to claim 7, characterized in that, in Step 2, the establishment of the model is based on the scatter plot established by the cumulative number of oil reservoirs and the semi-logarithm of the proven reserves collected in Step 1.

9. The method for predicting the trend of reserve growth based on big data analysis according to claim 8, characterized in that, in Step 2, the regression fitting analysis model applied is: Positive S-shaped: y = 1 / (α + β × exp(γ × x)) (3) S-shaped: y = 1 / (α + exp(β + γ × ln(x))) (4) Partial S-shaped: y = 1 / (α + exp(β + γ×(x) 0.5 ))(5) where y is the semi-logarithm of the proven reserves, x is the cumulative number of oil reservoirs, and α, β, and γ are the equation coefficients.

10. The method for predicting the trend of reserve growth based on big data analysis according to claim 1, characterized in that, in Step 3, use the correlation analysis technology in big data analysis technology to conduct multi-parameter sensitivity detection on the regression fitting analysis model established in Step 2.

11. The method for predicting the trend of reserve growth based on big data analysis according to claim 10, characterized in that, In step 3, first predict the most likely values of all parameters affecting reserve growth, and then examine the impact of simultaneous changes in all or some of the parameters on the reserve growth objective function. The analysis model is: OF: A function that defines the degree of mismatch between historical actual data and predicted data; historical fitting objective function. Mi: Variance of comprehensive sensitivity parameter; w i : Weight coefficient; a 0 , a w , a g , a p : Weight coefficients of each sensitivity parameter; Predicted data of each sensitivity parameter; Historical actual data of each sensitivity parameter; Through cluster analysis, the similarity of data and the characteristics of different classes are mined. The system uses the K-Means algorithm to perform cluster analysis on data samples and obtains evaluation indicators, mainly the silhouette coefficient and the nearest cluster.

12. The reserve growth trend prediction method based on big data analysis according to claim 11, characterized in that in step 3, neural network technology is applied to mine the relationship between each pattern parameter and the clustered data and its macroscopic characteristics. The deep feedforward neural network DFF is used to predict the model parameters, and a broom-shaped diagram can be generated.

13. The reserve growth trend prediction method based on big data analysis according to claim 12, characterized in that in step 3, a reserve growth sensitivity parameter study is carried out through three operations: multi-parameter sensitivity analysis, parameter sensitivity detection of the cluster analysis results, and parameter sensitivity detection of neural network and regression analysis.

14. The reserve growth trend prediction method based on big data analysis according to claim 1, characterized in that in step 4, further predict the newly added proven reserves reserved in a certain year by adjusting the newly added 3D seismic area and the number of exploration wells, and verify the fitting degree of the prediction model.

15. The reserve growth trend prediction method based on big data analysis according to claim 1, characterized in that in step 5, based on the model selected in step 2, considering the values of the sensitive parameters analyzed in step 3, predict the reserve growth trend in a future period of time in the study area.

Images