Offshore oilfield waterflood recovery factor prediction method
By combining geological data and machine learning models in offshore oilfields, the inadequacy of offshore oilfield recovery rate prediction has been solved, achieving more accurate and objective recovery rate prediction.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CNOOC TIANJIN BRANCH
- Filing Date
- 2022-07-12
- Publication Date
- 2026-06-02
AI Technical Summary
Existing methods for predicting oil recovery rates are not well-suited for use in offshore oil fields, particularly static methods which are highly subjective and dynamic methods which suffer from insufficient data, leading to inaccurate prediction results.
By combining drilling, logging, well logging, and seismic data to determine geological reservoir characteristics, selecting analogous oilfields for historical data fitting, constructing machine learning models for oil recovery prediction, using BP neural networks for training and testing, and combining oil recovery sensitivity analysis, a method for predicting oil recovery applicable to offshore oilfields is constructed.
It improves the applicability and objectivity of offshore oilfield recovery rate prediction, reduces subjective factors, avoids the problem of insufficient sample points, and achieves more accurate recovery rate prediction.
Smart Images
Figure CN115271182B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of oil and gas field exploration and development, and particularly relates to a method for predicting water-drive recovery rates in offshore oil fields. Background Technology
[0002] Oil recovery rate is an important parameter for evaluating the development effect of an oilfield, and it has always been a concern for reservoir workers in the design and adjustment of development plans.
[0003] Due to insufficient natural energy in the Bohai Oilfield, 86% of its reserves require water injection for energy replenishment. Furthermore, given the high investment costs of offshore oilfields, it is crucial to determine reasonable recovery rates and the effectiveness of various development measures, while also ensuring adequate engineering contingency. Therefore, researching water-drive recovery rate calculation models for offshore oilfields is of great significance in providing an evaluation basis for water-injection development.
[0004] Currently, methods for predicting oil recovery rates mainly include static and dynamic methods. Static methods are suitable for undeveloped and early-stage oilfields, and primarily include analogy and empirical formula methods. Dynamic methods are suitable for oilfields with a certain development time, abundant production dynamic data, and established development patterns, and primarily include water drive curve methods, decline curve methods, and numerical simulation methods. However, current oil recovery rate prediction methods all have certain limitations when applied offshore. For example, the analogy method within static methods is highly subjective in quantitatively predicting oil recovery rates; empirical formula methods are generally derived from onshore oilfields and have relatively more measures and higher well density compared to offshore oilfields; and dynamic methods require a large amount of production dynamic data, while offshore oilfields are constrained by cost and have relatively limited data available. Therefore, there is an urgent need to establish a method for predicting oil recovery rates in offshore oilfields to guide oil recovery rate prediction. Summary of the Invention
[0005] The purpose of this invention is to provide a method for predicting the water-drive recovery rate of offshore oilfields, so as to solve the technical problem of predicting the water-drive recovery rate of offshore oilfields.
[0006] To achieve the above objectives, the specific technical solution of the offshore oilfield waterflood recovery prediction method of the present invention is as follows:
[0007] A method for predicting the waterflood recovery rate of offshore oilfields includes the following steps:
[0008] Step 1: Determine the analogous oilfield based on the geological and reservoir characteristics of the research object;
[0009] Step 2: Perform historical fitting on the analog oilfield production dynamics data;
[0010] Step 3: Predict the recovery rate of analogous oilfields and conduct a recovery rate sensitivity analysis;
[0011] Step 4: Construct a machine learning model based on the recovery rate prediction results;
[0012] Step 5: Use the trained machine learning model to predict the recovery rate of the research object.
[0013] Furthermore, in the first step, the research object is to determine the geological reservoir characteristics of the research object by combining drilling, logging, well logging, and seismic data, and to characterize the main parameters of the geological reservoir characteristics, including: oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology, driving type, reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity. Among these, the characteristic parameters of oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology, and driving type are measured using qualitative indicators, while the other five characteristic parameters of reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity are measured using quantitative indicators.
[0014] Furthermore, in the first step, when selecting analogous oilfields, it is required that the parameters measured by qualitative indicators are the same for both the research object and the analogous oilfield. It is difficult to find identical oilfields for parameters measured by quantitative indicators, and the selection is carried out according to the following steps:
[0015] 1. First, these parameters are qualitatively characterized according to national standards, industry standards or enterprise standards. When selecting analogous oilfields, the qualitative classification results of the parameters measured by quantitative indicators should be the same for the research object and the analogous oilfield.
[0016] 2. Based on reservoir depth, reservoirs are classified into five categories: Category 1 is shallow reservoirs, i.e., reservoirs with a depth of less than 500 meters; Category 2 is medium-shallow reservoirs, i.e., reservoirs with a depth of 500 meters or more but less than 2000 meters; Category 3 is medium-deep reservoirs, i.e., reservoirs with a depth of 2000 meters or more but less than 3500 meters; Category 4 is deep reservoirs, i.e., reservoirs with a depth of 3500 meters or more but less than 4500 meters; and Category 5 is ultra-deep reservoirs, i.e., reservoirs with a depth of 4500 meters or more.
[0017] 3. Based on reservoir thickness, reservoirs are classified into four categories: Category I is thin-layer reservoirs, i.e., reservoir thickness is less than 5 meters; Category II is medium-thick reservoirs, i.e., reservoir thickness is greater than or equal to 5 meters and less than 20 meters; Category III is thick-layer reservoirs, i.e., reservoir thickness is greater than or equal to 20 meters and less than 40 meters; and Category IV is extra-thick reservoirs, i.e., reservoir thickness is greater than or equal to 40 meters.
[0018] 4. Based on porosity, oil reservoirs are classified into four categories: Category 1 is ultra-low porosity reservoirs, i.e., reservoir porosity less than 10%; Category 2 is low porosity reservoirs, i.e., reservoir porosity greater than or equal to 10% and less than 15%; Category 3 is medium porosity reservoirs, i.e., reservoir porosity greater than or equal to 15% and less than 25%; and Category 4 is high porosity reservoirs, i.e., reservoir porosity greater than or equal to 25%.
[0019] 5. Based on permeability, reservoirs are classified into five categories: Category 1 is ultra-low permeability reservoirs, i.e., reservoir permeability less than 5 millidarcy; Category 2 is low permeability reservoirs, i.e., reservoir permeability greater than or equal to 5 millidarcy and less than 50 millidarcy; Category 3 is medium permeability reservoirs, i.e., reservoir permeability greater than or equal to 50 millidarcy and less than 500 millidarcy; Category 4 is high permeability reservoirs, i.e., reservoir permeability greater than or equal to 500 millidarcy and less than 1000 millidarcy; Category 5 is ultra-high permeability reservoirs, i.e., reservoir permeability greater than or equal to 1000 millidarcy.
[0020] 6. Based on the viscosity of the formation crude oil, oil reservoirs are classified into four categories: Category I is low-viscosity oil reservoir, i.e., the viscosity of the formation crude oil is less than 5 mPa·s; Category II is medium-viscosity oil reservoir, i.e., the viscosity of the formation crude oil is greater than or equal to 5 mPa·s and less than 20 mPa·s; Category III is high-viscosity oil reservoir, i.e., the viscosity of the formation crude oil is greater than or equal to 20 mPa·s and less than 50 mPa·s; Category IV is heavy oil reservoir, i.e., the viscosity of the formation crude oil is greater than or equal to 50 mPa·s.
[0021] Based on the geological and reservoir characteristics of the research object, analogous oilfields are determined, and producing oilfields with the same or similar stratigraphic position, sedimentary facies, reservoir type, reservoir lithology, driving type, reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity are selected as analogous objects.
[0022] Furthermore, in the second step, a historical fitting process is performed on the analog oilfield production dynamic data: due to the limitations of the understanding of the reservoir geological conditions during modeling, the numerical simulation model cannot truly reflect the actual situation of the reservoir. It is necessary to repeatedly adjust the static parameters of the reservoir so that the calculated values and actual values of the main dynamic indicators of oilfield development, such as pressure, oil production, and water cut, match.
[0023] Furthermore, in the second step, historical data from several oilfields are input into the corresponding geological model, and historical fitting work is carried out. Historical fitting includes two methods: oilfield fitting and single-well fitting.
[0024] Furthermore, in the third step, based on the historical fitting in the second step, recovery rate prediction is carried out, and sensitivity analysis is conducted on several oilfields by adjusting the model parameters.
[0025] Furthermore, in the third step, the recovery rate of the analog oilfield is predicted, and a recovery rate sensitivity analysis is conducted: based on historical fitting, numerical simulation studies are carried out to predict the recovery rate of the existing model, and sensitivity analysis is conducted on parameters such as horizontal permeability, horizontal to vertical permeability ratio, skin layer, formation crude oil viscosity, well network density, production pressure difference, water injection timing, and fluid extraction ratio on the recovery rate to predict the water drive recovery rate under different conditions.
[0026] Furthermore, in the fourth step, the recovery rate prediction results and sensitivity analysis results of several oilfield basic models from the third step are input into the machine learning model, and the selected machine learning model is a BP neural network. The machine learning model is constructed based on the recovery rate prediction results: a sample set is formed from the recovery rate prediction results of all analog oilfields and the recovery rates predicted by sensitivity analysis. A portion of the samples are randomly selected as the training set, and the remaining samples are used as the test set. The parameters of the sensitivity analysis from the third step are selected as the feature values for recovery rate prediction. The machine learning model is used to train the training set, and the trained machine learning model is used to predict the recovery rate on the test set. The prediction accuracy of the test set is calculated. If the test accuracy meets the expected requirements, training is stopped. If the test accuracy does not meet the expected requirements, the relevant parameters of the machine learning model are adjusted until the prediction accuracy of the test set meets the expected requirements.
[0027] Furthermore, in the fifth step, the recovery rate of the research object is predicted using the trained machine learning model: the feature parameters of the research object are input into the machine learning model trained in the fourth step, and the feature parameters are the parameters of the sensitivity analysis in the third step, so as to obtain the recovery rate prediction result.
[0028] The offshore oilfield waterflood recovery prediction method of the present invention has the following advantages:
[0029] 1. This invention selects offshore oilfields as analogous oilfields, fully considering the special characteristics of offshore oilfield development, and its applicability is stronger compared with empirical formulas for predicting recovery rates;
[0030] 2. In the quantitative prediction of oil recovery rate, this invention constructs a machine learning model. The machine learning samples are derived from analogous oil fields. However, compared with the conventional analogy method, it avoids subjective factors to a certain extent and has strong objectivity.
[0031] 3. When using machine learning, this invention adopts the results of recovery rate sensitivity analysis, thus avoiding the problem of insufficient sample points in conventional machine learning. Attached Figure Description
[0032] Figure 1 This is a schematic diagram of the process of the present invention;
[0033] Figure 2 This is a schematic diagram of oilfield pressure fitting according to the present invention; (it is the actual graphic on the screen).
[0034] Figure 3 This is a schematic diagram of oilfield production fitting according to the present invention; (it is the actual graphic on the screen).
[0035] Figure 4 This is a schematic diagram of oilfield water cut fitting according to the present invention; (it is the actual graphic on the screen).
[0036] Figure 5 This is a schematic diagram of oilfield single-well pressure fitting according to the present invention; (it is the actual graphic on the screen).
[0037] Figure 6 This is a schematic diagram of oilfield single-well production fitting according to the present invention; (it is the actual graphic on the screen).
[0038] Figure 7 This is a schematic diagram of water cut fitting for a single well in an oilfield according to the present invention. (It is the actual graphic on the screen.) Detailed Implementation
[0039] To better understand the purpose, structure, and function of this invention, the following detailed description of a method for predicting waterflood recovery rate in offshore oilfields, in conjunction with the accompanying drawings, is provided.
[0040] like Figures 1-7 As shown, the present invention employs the following steps:
[0041] Step 1: Determine the analogous oilfield based on the geological and reservoir characteristics of the research object;
[0042] Specifically, it involves combining drilling, logging, well logging, and seismic data to conduct geological research in order to determine the geological reservoir characteristics of the X oilfield under study. The main parameters used to characterize the geological reservoir characteristics include: oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology, driving type, reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity. Among these, five characteristic parameters—oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology, and driving type—are measured using qualitative indicators, while the other five characteristic parameters—reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity—are measured using quantitative indicators.
[0043] 1) When selecting analog oilfields, it is required that the parameters measured by qualitative indicators are the same for the research object and the analog oilfield. It is difficult to find the same oilfield for the parameters measured by quantitative indicators. First, these parameters are qualitatively characterized according to national standards, industry standards or enterprise standards. When selecting analog oilfields, it is required that the qualitative classification results of the parameters measured by quantitative indicators are the same for the research object and the analog oilfield. The qualitative classification results of each quantitative parameter in this embodiment are as follows;
[0044] 2) Based on reservoir depth, reservoirs are classified into five categories: Category 1 is shallow reservoirs, i.e., reservoir depth less than 500 meters; Category 2 is medium-shallow reservoirs, i.e., reservoir depth greater than or equal to 500 meters and less than 2000 meters; Category 3 is medium-deep reservoirs, i.e., reservoir depth greater than or equal to 2000 meters and less than 3500 meters; Category 4 is deep reservoirs, i.e., reservoir depth greater than or equal to 3500 meters and less than 4500 meters; and Category 5 is ultra-deep reservoirs, i.e., reservoir depth greater than or equal to 4500 meters.
[0045] 3) Based on reservoir thickness, reservoirs are classified into four categories: Category I is thin-layer reservoir, i.e., reservoir thickness is less than 5 meters; Category II is medium-thick reservoir, i.e., reservoir thickness is greater than or equal to 5 meters and less than 20 meters; Category III is thick-layer reservoir, i.e., reservoir thickness is greater than or equal to 20 meters and less than 40 meters; Category IV is extra-thick reservoir, i.e., reservoir thickness is greater than or equal to 40 meters.
[0046] 4) Based on porosity, oil reservoirs are classified into four categories: Category 1 is ultra-low porosity reservoirs, i.e., reservoir porosity is less than 10%; Category 2 is low porosity reservoirs, i.e., reservoir porosity is greater than or equal to 10% and less than 15%; Category 3 is medium porosity reservoirs, i.e., reservoir porosity is greater than or equal to 15% and less than 25%; and Category 4 is high porosity reservoirs, i.e., reservoir porosity is greater than or equal to 25%.
[0047] 5) Based on permeability, reservoirs are classified into five categories: Category 1 is ultra-low permeability reservoirs, i.e., reservoir permeability is less than 5 millidarcy; Category 2 is low permeability reservoirs, i.e., reservoir permeability is greater than or equal to 5 millidarcy and less than 50 millidarcy; Category 3 is medium permeability reservoirs, i.e., reservoir permeability is greater than or equal to 50 millidarcy and less than 500 millidarcy; Category 4 is high permeability reservoirs, i.e., reservoir permeability is greater than or equal to 500 millidarcy and less than 1000 millidarcy; Category 5 is ultra-high permeability reservoirs, i.e., reservoir permeability is greater than or equal to 1000 millidarcy.
[0048] 6) Based on the viscosity of the formation crude oil, oil reservoirs are classified into four categories: Category I is low-viscosity oil reservoirs, i.e., the viscosity of the formation crude oil is less than 5 mPa·s; Category II is medium-viscosity oil reservoirs, i.e., the viscosity of the formation crude oil is greater than or equal to 5 mPa·s and less than 20 mPa·s; Category III is high-viscosity oil reservoirs, i.e., the viscosity of the formation crude oil is greater than or equal to 20 mPa·s and less than 50 mPa·s; and Category IV is heavy oil reservoirs, i.e., the viscosity of the formation crude oil is greater than or equal to 50 mPa·s. The oil-bearing stratum of the X oilfield under study is: N1m. L The sedimentary facies is shallow-water delta; the reservoir type is lithological-structural; the reservoir lithology is sandstone; the driving type is artificial water injection; the reservoir depth is 1672–1753 meters, belonging to a shallow to medium-depth reservoir; the effective thickness is 7.8 meters, belonging to a medium to thick reservoir; the porosity is 31.5%, belonging to a high-porosity reservoir; the permeability is 1471.4 millidarcy, belonging to an ultra-high permeability reservoir; and the formation crude oil viscosity is 158 millipascals, belonging to a heavy oil reservoir.
[0049] Based on the geological and reservoir characteristics of the research object, four analogous oilfields were identified: A, B, C, and D.
[0050] The oil-bearing strata in oilfield A are: N1m LThe sedimentary facies is shallow-water delta; the reservoir type is lithological-structural; the reservoir lithology is sandstone; the driving type is artificial water injection; the reservoir depth is 1203–1263 meters, belonging to a shallow to medium-depth reservoir; the effective thickness of oilfield A is 6 meters, 8 meters, 10 meters, 11 meters, 11.9 meters, 12 meters, 14 meters, 16 meters, 18 meters, and 20 meters, with the preferred thickness in this embodiment being 11.9 meters, i.e., a reservoir thickness greater than or equal to 5 meters and less than 20 meters; it belongs to a Class II medium-thick reservoir; porosity The permeability is 27.7%, classifying it as a high-porosity reservoir. Oilfield A has permeability values of 1000 mDarcy, 1200 mDarcy, 1400 mDarcy, 1600 mDarcy, 1691.0 mDarcy, 1692.0 mDarcy, 1800 mDarcy, and 2000 mDarcy. In this embodiment, the preferred value is 1691.0 mDarcy. Reservoir A has a permeability greater than or equal to 1000 mDarcy, classifying it as a Class V ultra-high permeability reservoir. The formation crude oil viscosity is 135 mPa·s, classifying it as a heavy oil reservoir.
[0051] The oil-bearing strata in Oilfield B are: N1m L The sedimentary facies is shallow-water delta; the reservoir type is lithological-structural; the reservoir lithology is sandstone; the driving type is artificial water injection; and the reservoir depth is 1420–1485 meters, classifying it as a shallow to medium-depth reservoir. For Oilfield B, effective thicknesses are 6 meters, 8 meters, 9.7 meters, 10 meters, 11 meters, 12 meters, 14 meters, 16 meters, and 18 meters. In this embodiment, the preferred thickness is 9.7 meters, meaning the reservoir thickness is greater than or equal to 5 meters and less than 20 meters; this classifies it as a Class II medium-thick reservoir. The porosity is 28.8%, classifying it as a high-porosity reservoir; the permeability is 1332.0 millidarcy, classifying it as an ultra-high-permeability reservoir; the viscosity of the formation crude oil from Oilfield B is: 50 mPa·s, 80 mPa·s, 100 mPa·s, 120 mPa·s, 140 mPa·s, 159 mPa·s, 180 mPa·s, 200 mPa·s, and 220 mPa·s; in this embodiment, the preferred viscosity is 159 mPa·s, meaning the formation crude oil viscosity is greater than or equal to 50 mPa·s, classifying it as a Class IV heavy oil reservoir. The oil-bearing stratum in Oilfield C is N1m. L The sedimentary facies is shallow-water delta; the reservoir type is lithological-structural; the reservoir lithology is sandstone; the driving type is artificial water injection; the reservoir depth is 1455–1483 meters, belonging to a medium-shallow reservoir; the effective thickness of oilfield C is 6 meters, 8 meters, 8.1 meters, 8.2 meters, 10 meters, 11 meters, 12 meters, 14 meters, 16 meters, and 18 meters, with the preferred thickness in this embodiment being 8.1 meters, i.e., the reservoir thickness is greater than or equal to 5 meters and less than 20 meters, belonging to a Class II medium-thick reservoir; the porosity is 29.9%, belonging to a high-porosity reservoir; the permeability is 1502.9 millidarcy, belonging to an ultra-high permeability reservoir; the formation crude oil viscosity is 120 millipascals, belonging to a heavy oil reservoir.
[0052] The oil-bearing strata in Oilfield D are: N1mL The sedimentary facies is shallow-water delta; the reservoir type is lithological-structural; the reservoir lithology is sandstone; the driving type is artificial water injection; the reservoir depth is 1150–1263 meters, belonging to a medium-shallow reservoir; the effective thickness of the oilfield is 6 meters, 8 meters, 10 meters, 11 meters, 11.2 meters, 11.6 meters, 12 meters, 14 meters, 16 meters, and 18 meters, with the preferred thickness in this embodiment being 11.2 meters, i.e., the reservoir thickness is greater than or equal to 5 meters and less than 20 meters; it belongs to a Class II medium-thick reservoir; the porosity is 29.3%, belonging to a high-porosity reservoir; the permeability is 1104.0 millidarcy, belonging to an ultra-high permeability reservoir; the formation crude oil viscosity is 130 mPa·s, belonging to a heavy oil reservoir.
[0053] The geological reservoir characteristics of the research object X oilfield and analogous oilfields A, B, C, and D are shown in Table 1 below:
[0054] Table 1: Geological and Reservoir Characteristic Parameters of the Research Object and Analogous Oilfields
[0055]
[0056]
[0057] In the first step above, based on the geological reservoir characteristics of the research object, analogous oilfields are determined: producing oilfields with the same or similar stratigraphic position, sedimentary facies, reservoir type, reservoir lithology, driving type, reservoir depth, effective thickness, porosity, permeability, and formation crude oil viscosity are selected as analogous objects.
[0058] Step 2: Perform historical fitting on the analog oilfield production dynamics data;
[0059] The process of historical fitting of analog oilfield production dynamic data: Due to the limitations of the understanding of reservoir geology during modeling, the numerical simulation model cannot truly reflect the actual situation of the reservoir. It is necessary to repeatedly adjust the static parameters of the reservoir to match the calculated and actual values of the main dynamic indicators of oilfield development, such as pressure, oil production, and water cut.
[0060] First, historical data from oilfields A, B, C, and D are input into the corresponding geological models, and historical data fitting is performed. Historical data fitting includes two methods: oilfield fitting and single-well fitting. Taking oilfield A as an example, the process and results of historical data fitting are illustrated below:
[0061] 1. Input the development parameters of oilfield A into the numerical simulation model, and fit the model by continuously optimizing and changing the static parameters and surface parameters. The fitting results are as follows: Figures 2 to 7As shown, the horizontal permeability of the fitted numerical simulation model of oilfield A is 1691.0 millidarcy, the horizontal to vertical permeability ratio is 10, the skin permeability is 5, and the formation crude oil viscosity is 135 mPa·s.
[0062] 2. After fitting using a similar method, the horizontal permeability of the numerical simulation model for oilfield B is 1332.0 millidarcy, the horizontal to vertical permeability ratio is 20, the skin permeability is 6, and the formation crude oil viscosity is 159 millipascals.
[0063] The horizontal permeability of the numerical simulation model for oilfield 3C is 1502.9 millidarcy, the horizontal to vertical permeability ratio is 22, the skin permeability is 7, and the formation crude oil viscosity is 120 millipascals.
[0064] The horizontal permeability of the numerical simulation model of oilfield 4.D is 1104.0 millidarcy, the horizontal to vertical permeability ratio is 18, the skin permeability is 5, and the formation crude oil viscosity is 130 millipascals.
[0065] Step 3: Predict the recovery rate of analogous oilfields and conduct a recovery rate sensitivity analysis;
[0066] Based on the historical data fitting in the second step, recovery rate prediction was carried out, and sensitivity analysis was conducted on oilfields A, B, C, and D by adjusting the model parameters. The recovery rate prediction results and sensitivity analysis results of the basic model for oilfield A are shown in Table 2 below:
[0067] Table 2: Recovery Rate Prediction Results and Sensitivity Analysis Results of Basic Model for Oilfield A
[0068]
[0069] The recovery rate prediction results and sensitivity analysis results of the basic model for oilfield B are shown in Table 3 below:
[0070] Table 3: Recovery Rate Prediction Results and Sensitivity Analysis Results of Basic Model for Oilfield B
[0071]
[0072] The recovery rate prediction results and sensitivity analysis results of the basic model for oilfield C are shown in Table 4 below:
[0073] Table 4: Recovery Rate Prediction Results and Sensitivity Analysis Results of Basic Model for Oilfield C
[0074]
[0075] The recovery rate prediction results and sensitivity analysis results of the basic model for oilfield D are shown in Table 5 below:
[0076] Table 5: Recovery Rate Prediction Results and Sensitivity Analysis Results of the Basic Model for Oilfield D
[0077]
[0078] In the third step above, the recovery rate of the analog oilfield is predicted, and a sensitivity analysis of the recovery rate is carried out: based on historical fitting, numerical simulation studies are conducted to predict the recovery rate of the existing model, and sensitivity analysis of parameters such as horizontal permeability, horizontal to vertical permeability ratio, skin layer, formation crude oil viscosity, well network density, production pressure difference, water injection timing, and fluid extraction ratio on the recovery rate is carried out to predict the water drive recovery rate under different conditions.
[0079] Step 4: Construct a machine learning model based on the recovery rate prediction results;
[0080] The recovery rate prediction results and sensitivity analysis results of the basic model for oilfields A, B, C, and D in step three are input into the machine learning model. In this embodiment, the machine learning model selected is a backpropagation (BP) neural network. Each oilfield in this BP neural network involves 33 sample points, totaling 132 sample points. 112 sample points are selected as the training set, and the remaining 20 sample points are used as the test set. The model output node number is 1, representing the recovery rate. The hidden layer nodes are 5. The input and hidden layer transfer functions are logsig, and the output layer transfer function is pure. The training function is `traingd`, the learning function is `learngd`, the learning rate is 0.05, and the maximum number of iterations is 20000. The accuracy of the interpolation model is measured using root mean square error (RMSE), mean relative error (MREE), and mean absolute error (MAE). The test accuracy requirement is that the RMSE, MRE, and MAE are all less than 10%. Based on the prediction results of the machine learning model, the RMSE, MRE, and MAE are 7.25%, 6.28%, and 2.86%, respectively, and the prediction accuracy of the test set meets the expected requirements.
[0081] In the fourth step above, a machine learning model is constructed based on the recovery rate prediction results: the recovery rate prediction results of all analog oilfields and the sensitivity analysis results are used to form a sample set. A portion of the samples are randomly selected as the training set, and the remaining samples are used as the test set. The parameters of the sensitivity analysis in the third step are selected as the feature values for the recovery rate prediction. The machine learning model is used to train the training set. The trained machine learning model is then used to predict the recovery rate of the test set. The prediction accuracy of the test set is calculated. If the test accuracy meets the expected requirements, training is stopped. If the test accuracy does not meet the expected requirements, the relevant parameters of the machine learning model are adjusted until the prediction accuracy of the test set meets the expected requirements.
[0082] Step 5: Use the trained machine learning model to predict the recovery rate of the research object;
[0083] The recovery rate prediction parameters for the X oilfield were input into the machine learning model constructed in the fourth step for prediction. The basic parameters for the X oilfield were set as follows: horizontal permeability: 1471.4 millidarcy; horizontal to vertical permeability ratio: 15; skin depth: 5; formation crude oil viscosity: 158 mPa·s; well density: 3.1 wells / km²; production pressure differential: 1.2 MPa; water injection timing: simultaneous water injection; and fluid extraction ratio: 4.5 times. Ultimately, the recovery rate of the X oilfield was determined to be 28.2%.
[0084] In the fifth step above, the recovery rate of the research object is predicted using the trained machine learning model: the feature parameters of the research object are input into the machine learning model trained in the fourth step. The feature parameters are the parameters of the sensitivity analysis in the third step, and the recovery rate prediction results are obtained.
[0085] The technologies not described above are existing technologies and will not be elaborated further.
[0086] It is understood that the present invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of the invention. Furthermore, under the teachings of the present invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of the invention. Therefore, the present invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are within the protection scope of the present invention.
Claims
1. A method for predicting waterflood recovery rate in offshore oilfields, characterized in that, Including: the following step: Step 1: Determine the analogous oilfield based on the geological and reservoir characteristics of the research object; Step 2: Perform historical fitting on the analog oilfield production dynamics data; Step 3: Predict the recovery rate of analogous oilfields and conduct a recovery rate sensitivity analysis; Step 4: Construct a machine learning model based on the recovery rate prediction results; Step 5: Use the trained machine learning model to predict the recovery rate of the research object; In the third step, the recovery rate of the analog oilfield is predicted and a recovery rate sensitivity analysis is conducted: based on historical fitting, numerical simulation studies are carried out to predict the recovery rate of the existing model, and sensitivity analysis is conducted on parameters such as horizontal permeability, horizontal to vertical permeability ratio, skin layer, formation crude oil viscosity, well network density, production pressure difference, water injection timing, and fluid extraction ratio on the recovery rate to predict the water drive recovery rate under different conditions. In the fourth step, the recovery rate prediction results and sensitivity analysis results of several oilfield basic models from the third step are input into the machine learning model, and the selected machine learning model is a BP neural network. The machine learning model is constructed based on the recovery rate prediction results: the recovery rate prediction results of all analog oilfields and the recovery rates predicted by sensitivity analysis constitute a sample set, a portion of the samples are randomly selected as the training set, and the remaining samples are used as the test set. The parameters of the sensitivity analysis in the third step are selected as the feature values for recovery rate prediction. The machine learning model is used to train the training set, and the trained machine learning model is used to predict the recovery rate of the test set. The prediction accuracy of the test set is calculated. If the test accuracy meets the expected requirements, training is stopped. If the test accuracy does not meet the expected requirements, the relevant parameters of the machine learning model are adjusted until the prediction accuracy of the test set meets the expected requirements.
2. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1, characterized in that, The research object in the first step is to determine the geological reservoir characteristics of the research object by combining drilling, logging, well logging and seismic data. The main parameters used to characterize the geological reservoir characteristics include: oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology, driving type, reservoir depth, effective thickness, porosity, permeability and formation crude oil viscosity. Among them, the characteristic parameters oil-bearing strata, sedimentary facies, reservoir type, reservoir lithology and driving type are measured by qualitative indicators, while the other five characteristic parameters reservoir depth, effective thickness, porosity, permeability and formation crude oil viscosity are measured by quantitative indicators.
3. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1 or 2, characterized in that, In the first step, the analog oil field is selected according to the following steps: (1) First, the reservoir types of the research object are classified according to the relevant national, industry or enterprise standards for reservoir classification; when selecting analogous oilfields, the qualitative indicators of the characteristic parameters of the research object and the analogous oilfield are the same, and the reservoir types classified by the quantitative indicators of the characteristic parameters of the research object and the analogous oilfield are the same. (2) Based on the burial depth, oil reservoirs are divided into 5 categories: Category 1 is shallow oil reservoir, that is, the burial depth of the oil reservoir is less than 500 meters; Category 2 is medium-shallow oil reservoir, that is, the burial depth of the oil reservoir is greater than or equal to 500 meters and less than 2000 meters; Category 3 is medium-deep oil reservoir, that is, the burial depth of the oil reservoir is greater than or equal to 2000 meters and less than 3500 meters; Category 4 is deep oil reservoir, that is, the burial depth of the oil reservoir is greater than or equal to 3500 meters and less than 4500 meters; Category 5 is ultra-deep oil reservoir, that is, the burial depth of the oil reservoir is greater than or equal to 4500 meters. (3) Based on the thickness of the reservoir, the reservoirs are divided into four categories: Category 1 is thin reservoir, that is, the reservoir thickness is less than 5 meters; Category 2 is medium-thick reservoir, that is, the reservoir thickness is greater than or equal to 5 meters and less than 20 meters; Category 3 is thick reservoir, that is, the reservoir thickness is greater than or equal to 20 meters and less than 40 meters; Category 4 is extra-thick reservoir, that is, the reservoir thickness is greater than or equal to 40 meters. (4) Based on porosity, oil reservoirs are divided into four categories: Category 1 is ultra-low porosity oil reservoir, that is, the oil reservoir porosity is less than 10%; Category 2 is low porosity oil reservoir, that is, the oil reservoir porosity is greater than or equal to 10% and less than 15%; Category 3 is medium porosity oil reservoir, that is, the oil reservoir porosity is greater than or equal to 15% and less than 25%; Category 4 is high porosity oil reservoir, that is, the oil reservoir porosity is greater than or equal to 25%. (5) Based on permeability, oil reservoirs are divided into 5 categories: Category 1 is ultra-low permeability oil reservoir, that is, the permeability of the oil reservoir is less than 5 millidarcy; Category 2 is low permeability oil reservoir, that is, the permeability of the oil reservoir is greater than or equal to 5 millidarcy and less than 50 millidarcy; Category 3 is medium permeability oil reservoir, that is, the permeability of the oil reservoir is greater than or equal to 50 millidarcy and less than 500 millidarcy; Category 4 is high permeability oil reservoir, that is, the permeability of the oil reservoir is greater than or equal to 500 millidarcy and less than 1000 millidarcy; Category 5 is ultra-high permeability oil reservoir, that is, the permeability of the oil reservoir is greater than or equal to 1000 millidarcy. (6) Based on the viscosity of the formation crude oil, the reservoirs are divided into four categories: Category 1 is low viscosity oil reservoir, that is, the viscosity of the formation crude oil is less than 5 mPa·s; Category 2 is medium viscosity oil reservoir, that is, the viscosity of the formation crude oil is greater than or equal to 5 mPa·s and less than 20 mPa·s; Category 3 is high viscosity oil reservoir, that is, the viscosity of the formation crude oil is greater than or equal to 20 mPa·s and less than 50 mPa·s; and Category 4 is heavy oil reservoir, that is, the viscosity of the formation crude oil is greater than or equal to 50 mPa·s.
4. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1, characterized in that, In the second step, a historical fitting process is performed on the analog oilfield production dynamic data: Due to the limitations of the understanding of reservoir geology during modeling, the numerical simulation model cannot truly reflect the actual situation of the reservoir. It is necessary to repeatedly adjust the static parameters of the reservoir so that the calculated values of the main dynamic indicators of oilfield development, such as pressure, oil production, and water cut, match the actual values.
5. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1 or 4, characterized in that, In the second step, historical data from several oilfields are input into the corresponding geological model, and historical fitting is carried out. Historical fitting includes two methods: oilfield fitting and single-well fitting.
6. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1, characterized in that, In the third step, based on the historical fitting in the second step, recovery rate prediction is carried out, and sensitivity analysis is conducted on several oilfields by adjusting the model parameters.
7. The method for predicting waterflood recovery rate in offshore oilfields according to claim 1, characterized in that, In the fifth step, the recovery rate of the research object is predicted using the trained machine learning model: the feature parameters of the research object are input into the machine learning model trained in the fourth step, and the feature parameters are the parameters of the sensitivity analysis in the third step, so as to obtain the recovery rate prediction result.