Offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering

Abstract

A method for dynamic analysis and prediction of offshore oil fields based on data-driven and knowledge engineering obtains multiple groups of historical oil field production data from offshore oil fields that have completed production. By collecting and processing the change data of oil well reservoir pressure, oil well production fluid, oil production speed and water injection volume data of water injection wells in historical offshore oil field production, the reservoir pressure difference, injection-production ratio and water injection volume data can be dynamically analyzed and evolved, so as to obtain the dynamic relationship between water injection volume and reservoir pressure difference and injection-production ratio, which is convenient for predicting and judging the water injection volume of actual offshore oil well production, taking timely measures such as fracturing and water plugging, and adjusting the water injection volume, which is conducive to achieving the effect of increasing production and injection.

Description

Technical Field

[0001] The present invention relates to a method for analyzing offshore oilfield dynamics, and more particularly to a method for analyzing and predicting offshore oilfield dynamics based on data drive and knowledge engineering. Background Art

[0002] When developing offshore oil fields, it is necessary to conduct dynamic analysis of the oil field exploitation process in order to grasp the progress of offshore oil field exploitation in real time. This includes dynamic analysis of oil field production, dynamic analysis of oil well lifting in the wellbore, and dynamic analysis of oil layers (reservoirs). When conducting dynamic analysis of the oil field exploitation process, it is necessary to use actual exploitation data as data drive and combine knowledge engineering to realize dynamic analysis of oil field development dynamics.

[0003] Production dynamics analysis is a production dynamics analysis of a single well. The existing production dynamics analysis method is to detect the production process of a single well in the oil field, but it is unable to predict and judge the production measures during the development of the oil well. At the same time, it is unable to reasonably control the daily oil production of the oil well, and may not be able to ensure that the production capacity of the oil well is in an optimal state, which will have a certain impact on the production capacity of the oil well. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an offshore oil field dynamic analysis and prediction method based on data-driven and knowledge engineering, which can predict and judge the effective water injection volume and daily oil production of actual offshore oil well exploitation in order to overcome the shortcomings of the existing technology.

[0005] The technical solution adopted by the present invention is: a method for analyzing and predicting offshore oilfield dynamics based on data-driven and knowledge engineering, comprising the following steps:

[0006] S1. Acquire multiple sets of historical single-well production data from offshore oil fields that have completed production, wherein the single-well production data includes reservoir pressure data of the single-well oil field, daily liquid production and daily oil production of the single-well oil field, and effective water injection volume data of the single-well water injection well;

[0007] S2, processing the oilfield single well production data obtained in step S1;

[0008] S3, dividing the data in the reservoir pressure difference data group R, the effective water injection data group D, the daily liquid production data group B, the single well injection-production ratio data group H, the water content data group M, and the daily oil production data group C into a modeling group, a training group, and a test group according to a set ratio;

[0009] S4. Taking the modeling group data from the reservoir pressure differential data set R, the effective water injection volume data set D, and the daily liquid production data set B, and using a scatter curve fitting method, establish a curve model between the change in reservoir pressure, the effective water injection volume data, and the liquid production. Specifically, the reservoir pressure differential in the reservoir pressure differential data set R is used as a known condition, the daily liquid production in the daily liquid production data set B is used as an independent variable, and the effective water injection volume data in the effective water injection volume data set D is used as a dependent variable. A curve model is established between the effective water injection volume data and the liquid production based on the reservoir pressure differential;

[0010] S5. Taking the training data from the reservoir pressure difference data set R, the effective water injection data set D, and the daily liquid production data set B, and comparing the scatter point values ​​of the curve model established in step S4, the curve model is adjusted according to the comparison results, i.e., the curve model is adjusted closer to the corresponding coordinate points in the training data;

[0011] S6. Take the test group data from the reservoir pressure differential data set R, the effective water injection volume data set D, and the daily liquid production data set B, use the reservoir pressure differential and the daily liquid production as known quantities, use the adjusted curve model to predict the effective water injection volume data, and compare and verify with the data in the effective water injection volume data set D. Based on the verification results, smooth the abnormal line segments of the curve model;

[0012] S7. During actual offshore oilfield production, the effective water injection volume data is predicted using the curve model smoothed in step S6 based on the actual oil well oil-reservoir pressure difference and daily liquid production. The predicted result is compared with the actual effective water injection volume data to timely adjust the effective water injection volume data of the individual wells in the oilfield to increase production and injection;

[0013] S8. Take the modeling group data from the water-cut data group M, the single-well injection-production ratio data group H, and the daily oil production data group C, and use the data group obtained by dividing the water cut in the water-cut data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H and the daily oil production data group C as data sources, and use a scatter curve fitting method to establish a curve model between daily oil production, water cut, and single-well injection-production ratio. Specifically, use the data obtained by dividing the water cut in the water-cut data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H as the independent variable, and the daily oil production in the daily oil production data group C as the dependent variable, and establish a curve model between daily oil production, water cut, and single-well injection-production ratio.

[0014] S9, taking the training data from the water-cut data group M, the single-well injection-production ratio data group H, and the daily oil production data group C, and comparing the scatter point values ​​of the curve model established in step S8, and adjusting the curve model based on the comparison results, that is, adjusting the curve model closer to the corresponding coordinate points in the training data;

[0015] S10, taking the test group data from the water content data group M, the single-well injection-production ratio data group H, and the daily oil production data group C, using the water content and daily oil production as known quantities, using the curve model adjusted in step S9 to predict the single-well injection-production ratio, comparing the predicted results with the data in the single-well injection-production ratio data group H for verification, and smoothing the abnormal line segments of the curve model based on the verification results;

[0016] S11. During actual offshore oilfield production, the injection-production ratio of a single well is predicted using the curve model smoothed in step S9 based on the actual water content and daily oil production of the oil wells. This is then compared with the actual injection-production ratio of the single well, so as to timely adjust the water content and injection-production ratio of the single well in the oilfield to increase production and injection.

[0017] The historical oil fields described in step S1 are all oil wells whose oil well generation capacity meets the standards.

[0018] Step S1 includes:

[0019] S101, obtaining the total amount of oil layer pressure changes and oil layer production depths in multiple sets of historical single-well production in an oil field, and establishing oil layer sub-data sets, which are respectively recorded as oil layer sub-data set A1, oil layer sub-data set A2, ..., and oil layer sub-data set AN, where N is the total number of oil wells;

[0020] S102. Obtain daily liquid production of a single well in the oil field. The daily liquid production includes oil production and water production, which are recorded as daily liquid production sub-data group B1, daily liquid production sub-data group B2, ..., and daily liquid production sub-data group BN, respectively, to form daily liquid production data group B;

[0021] S103, obtaining the daily oil production of each oil well of the single well in the oil field, recording them as daily oil production quantum data C1, daily oil production quantum data C2, ... and daily oil production quantum data CN, and forming a daily oil production data group C;

[0022] S104. Obtain the effective water injection volume data of each water injection well in the single-well exploitation of the oil field, and record them as effective water injection volume data sub-data D1, effective water injection volume data sub-data D2, ... and effective water injection volume data sub-data DN, forming an effective water injection volume data group D.

[0023] In the oilfield single well production data described in step S1, there are differences between the same type of data of the oilfield single wells.

[0024] Step S2 includes:

[0025] S201, dividing the total amount of oil well reservoir pressure changes in each of the oil layer sub-data group A1, oil layer sub-data group A2, ..., and oil layer sub-data group AN by the oil layer production depth, to obtain the unit depth change of the oil well reservoir pressure in each oil layer sub-data group, denoted as reservoir pressure difference r1, reservoir pressure difference r2, ..., and reservoir pressure difference rN, respectively, to form the reservoir pressure difference data group R, where N is the total number of oil wells;

[0026] S202, respectively dividing the effective water injection volume data sub-data D1, the effective water injection volume data sub-data D2, ..., and the effective water injection volume data sub-data DN by the daily liquid production sub-data group B1, the daily liquid production sub-data group B2, ..., and the daily liquid production sub-data group BN, to obtain the single-well injection-production ratio of the oil well, which are recorded as the single-well injection-production ratio of the oil well h1, the single-well injection-production ratio of the oil well h2, ..., and the single-well injection-production ratio of the oil well hN, respectively, to form the single-well injection-production ratio data group H;

[0027] S203. Divide the water production in each daily liquid production sub-data group B1, daily liquid production sub-data group B2, ... and daily liquid production sub-data group BN by the corresponding daily liquid production to obtain the water content of the corresponding oil well, which is recorded as water content m1, water content m2, ... and water content mN respectively, forming the water content data group M.

[0028] The division into modeling group, training group and test group according to the set ratio described in step S3 includes: division into modeling group, training group and test group according to the ratio of 7:2:1; or division into modeling group, training group and test group according to the ratio of 6:2:2; or division into modeling group, training group and test group according to the ratio of 5:3:2.

[0029] The modeling group, training group, and test group divided in step S3 are randomly selected from the data in the same group to ensure that the data changes between the divided data groups in the same group are discontinuous.

[0030] The comparison of the scatter values ​​described in step S5 and step S9 is specifically as follows:

[0031] (1) Take the training set data corresponding to the curve model and divide the corresponding data in the training set data into independent variables and dependent variables;

[0032] (2) Mark the coordinates of the independent variable and dependent variable data in the same group in the training set data on the curve model between the effective water injection volume data and the liquid production volume based on the oil layer pressure difference established in step S4, or on the curve model between the daily oil production and the water cut and the single well injection-production ratio established in step S8;

[0033] (3) The curve model established in step S4 or step S8 is adjusted closer to the position of the coordinate point in the training set data so that the curve model coincides with the coordinate point of the training set data.

[0034] Step S5 specifically involves taking the training data from the reservoir pressure difference data set R, the effective water injection data set D, and the daily liquid production data set B, taking the daily liquid production in the daily liquid production data set B as the independent variable, and the effective water injection data in the effective water injection data set D as the dependent variable;

[0035] For step S9, specifically, the training group data from the water-containing data group M, the single-well injection-production ratio data group H, and the daily oil production data group C are taken, and the data obtained by dividing the water content in the water-containing data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H is used as the independent variable, and the daily oil production in the daily oil production data group C is used as the dependent variable.

[0036] The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering of the present invention has the following beneficial effects:

[0037] 1. The method of the present invention can dynamically analyze and evolve the reservoir pressure difference, injection-production ratio, and effective water injection volume data by collecting and processing data on changes in reservoir pressure of oil wells, daily liquid production and oil production of oil wells, and effective water injection volume of water injection wells in historical offshore oil field exploitation, thereby deriving a dynamic relationship between the effective water injection volume and the reservoir pressure difference and injection-production ratio. This facilitates the prediction and judgment of the effective water injection volume in actual offshore oil well exploitation, and allows timely adoption of measures such as fracturing and water plugging to adjust the effective water injection volume, thereby facilitating the effect of increasing production and injection.

[0038] 2. The method of the present invention can dynamically analyze and evolve the data on water content, injection-production ratio and daily oil production by collecting and processing data on changes in oil layer pressure of oil wells, daily liquid production and daily oil production of oil wells and effective water injection volume of water injection wells in historical offshore oil field exploitation, thereby deriving a dynamic relationship between daily oil production and water content and injection-production ratio, which is convenient for predicting and judging the daily oil production of actual offshore oil well exploitation, and timely adjusting the daily oil production, which is conducive to achieving the effect of increasing production and injection. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 This is a curve model diagram showing the relationship between the change in oil layer pressure, the effective water injection volume, and the daily fluid production;

[0040] Figure 2 It is a curve model diagram of the relationship between daily oil production, water content and injection-production ratio of the present invention. DETAILED DESCRIPTION

[0041] The following detailed description of the data-driven and knowledge-engineered offshore oilfield dynamic analysis and prediction method of the present invention is provided in conjunction with the following embodiments and accompanying drawings. It is apparent that the described embodiments are merely a portion of the embodiments of the present invention, and not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0042] The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering of the present invention comprises the following steps:

[0043] S1. Obtain multiple sets of historical oilfield single-well production data from completed offshore oilfields. The historical oilfields are all oilfields with wells that meet oil well production capacity standards. The oilfield single-well production data contains differences between the same type of data in the oilfield single-well production data. The oilfield single-well production data includes single-well oil reservoir pressure data, single-well oil production and daily fluid production, and single-well water injection volume; specifically, it includes:

[0044] S101, obtaining the total amount of oil layer pressure changes and oil layer production depths in multiple sets of historical single well production in oil fields, and establishing oil layer sub-data sets, which are respectively recorded as oil layer sub-data set A1, oil layer sub-data set A2, ... and oil layer sub-data set AN, the total number of oil wells N;

[0045] S102. Obtain daily liquid production of a single well in the oil field. The daily liquid production includes oil production and water production, which are recorded as daily liquid production sub-data group B1, daily liquid production sub-data group B2, ..., and daily liquid production sub-data group BN, respectively, to form daily liquid production data group B;

[0046] S103, obtaining the daily oil production of each oil well of the single well in the oil field, recording them as daily oil production quantum data C1, daily oil production quantum data C2, ... and daily oil production quantum data CN, and forming a daily oil production data group C;

[0047] S104. Obtain the effective water injection volume of each water injection well in the single-well production of the oil field, and record them as effective water injection quantum data D1, effective water injection quantum data D2, ... and effective water injection quantum data DN, forming an effective water injection volume data group D.

[0048] S2, processing the oilfield single well production data obtained in step S1; including:

[0049] S201, dividing the total amount of oil well reservoir pressure changes in each of the oil layer sub-data group A1, oil layer sub-data group A2, ..., and oil layer sub-data group AN by the oil layer production depth, to obtain the unit depth change of the oil well reservoir pressure in each oil layer sub-data group, denoted as reservoir pressure difference r1, reservoir pressure difference r2, ..., and reservoir pressure difference rN, respectively, to form the reservoir pressure difference data group R, where N is the total number of oil wells;

[0050] S202, respectively divide the effective water injection quantum data D1, the effective water injection quantum data D2, ..., and the effective water injection quantum data DN by the daily liquid production sub-data group B1, the daily liquid production sub-data group B2, ..., and the daily liquid production sub-data group BN, to obtain the injection-production ratio of the oil well, which are recorded as the injection-production ratio of the oil well h1, the injection-production ratio of the oil well h2, ..., and the injection-production ratio of the oil well hN, respectively, to form the injection-production ratio data group H;

[0051] S203. Divide the water production in each daily liquid production sub-data group B1, daily liquid production sub-data group B2, ... and daily liquid production sub-data group BN by the corresponding daily liquid production to obtain the water content of the corresponding oil well, which is recorded as water content m1, water content m2, ... and water content mN respectively, forming the water content data group M.

[0052] S3, dividing the data in the reservoir pressure difference data group R, the effective water injection data group D, the daily liquid production data group B, the injection-production ratio data group H, the water content data group M, and the daily oil production data group C into a modeling group, a training group, and a test group according to a set ratio;

[0053] The dividing into modeling group, training group and test group according to the set ratio includes: dividing into modeling group, training group and test group according to the ratio of 7:2:1; or dividing into modeling group, training group and test group according to the ratio of 6:2:2; or dividing into modeling group, training group and test group according to the ratio of 5:3:2.

[0054] The divided modeling group, training group and test group are randomly selected from the data within the same group to ensure that the data changes between the divided data groups within the same group are discontinuous.

[0055] S4. Take the modeling group data from the reservoir pressure difference data group R, the effective water injection volume data group D, and the daily liquid production data group B, and use the scatter curve fitting method to establish a curve model between the change in reservoir pressure, the effective water injection volume, and the daily liquid production. Specifically, the reservoir pressure difference in the reservoir pressure difference data group R is used as a known condition, the daily liquid production in the daily liquid production data group B is used as an independent variable, and the effective water injection volume in the effective water injection volume data group D is used as a dependent variable. A curve model based on the reservoir pressure difference between the effective water injection volume and the liquid production volume is established, as shown in FIG. Figure 1 As shown;

[0056] S5. Taking the training data from the reservoir pressure difference data set R, the effective water injection volume data set D, and the daily liquid production data set B, and comparing the scatter point values ​​of the curve model established in step S4, the curve model is adjusted according to the comparison results, i.e., the curve model is adjusted closer to the corresponding coordinate points in the training data;

[0057] S6. Take the test data from the reservoir pressure differential data set R, the effective water injection volume data set D, and the daily liquid production data set B, use the reservoir pressure differential and the daily liquid production as known quantities, use the adjusted curve model to predict the effective water injection volume, and compare and verify the data with the effective water injection volume data set D. Based on the verification results, smooth the abnormal line segments of the curve model;

[0058] S7. During actual offshore oilfield production, the effective water injection volume is predicted using the smoothed curve model from step S6 based on the actual oil well reservoir pressure difference and daily liquid production. The predicted result is compared with the actual effective water injection volume to timely adjust the effective water injection volume of a single well in the oilfield to increase production and injection;

[0059] S8. Take the modeling group data from the water-cut data group M, the single-well injection-production ratio data group H, and the daily oil production data group C, and use the data group obtained by dividing the water cut in the water-cut data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H and the daily oil production data group C as data sources. Use the scatter curve fitting method to establish a curve model between daily oil production, water cut, and injection-production ratio. Specifically, use the data obtained by dividing the water cut in the water-cut data group M by the single-well injection-production ratio in the injection-production ratio data group H as the independent variable, and the daily oil production in the daily oil production data group C as the dependent variable, and establish a curve model between daily oil production, water cut, and single-well injection-production ratio, as shown in FIG. Figure 2 As shown;

[0060] S9, taking the training data from the water-cut data group M, the injection-production ratio data group H, and the daily oil production data group C, and comparing the scatter point values ​​of the curve model established in step S8, and adjusting the curve model according to the comparison results, that is, adjusting the curve model closer to the corresponding coordinate points in the training data;

[0061] S10, taking the test group data from the water content data group M, the injection-production ratio data group H, and the daily oil production data group C, using the water content and daily oil production as known quantities, using the curve model adjusted in step S9 to predict the injection-production ratio, comparing the predicted results with the data in the injection-production ratio data group H for verification, and smoothing the abnormal line segments of the curve model based on the verification results;

[0062] S11. During actual offshore oilfield production, the injection-production ratio is predicted using the curve model smoothed in step S9 based on the actual water content and daily oil production of the oil wells, and compared with the injection-production ratio of the individual wells, so as to timely adjust the water content and injection-production ratio of the individual wells in the oilfield to increase production and injection.

[0063] The scatter value comparison described in step S5 and step S9 of the present invention includes:

[0064] (1) Take the training set data corresponding to the curve model and divide the corresponding data in the training set data into independent variables and dependent variables; specifically:

[0065] Step S5 specifically involves taking the training data from the reservoir pressure difference data set R, the effective water injection volume data set D, and the daily liquid production data set B, taking the daily liquid production in the daily liquid production data set B as the independent variable, and the effective water injection volume in the effective water injection volume data set D as the dependent variable;

[0066] For step S9, specifically, the training group data in the water-content data group M, the injection-production ratio data group H and the daily oil production data group C are taken, and the data obtained by dividing the water content in the water-content data group M by the injection-production ratio in the injection-production ratio data group H is used as the independent variable, and the daily oil production in the daily oil production data group C is used as the dependent variable.

[0067] (2) Mark the coordinates of the independent variable and dependent variable data in the same group in the training set data on the curve model between the change of oil layer pressure and the effective water injection rate and liquid production rate established in step S4, or on the curve model between daily oil production and water cut and injection-production ratio established in step S8;

[0068] (3) The curve model established in step S4 or step S8 is adjusted closer to the position of the coordinate point in the training set data so that the curve model coincides with the coordinate point of the training set data.

[0069] Next, a data set was collected to conduct an experimental test on the offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering of the present invention:

[0070] Multiple sets of historical oilfield production data were obtained, as shown in Table 1.

[0071] Table 1

[0072]

[0073] The data in Table 1 are processed to obtain a data set of related data, as shown in Table 2.

[0074] Table 2

[0075]

[0076]

[0077] The data in Tables 1 and 2 were combined and grouped, with the modeling group data shown in Table 3.

[0078] Table 3

[0079]

[0080] The training group data are shown in Table 4.

[0081] Table 4

[0082]

[0083] The test group data is shown in Table 5.

[0084] Table 5

[0085]

[0086] By adopting the scatter point fitting method, the data in Table 3 are used to establish a curve model between the effective water injection rate and the daily liquid production when the reservoir pressure is a known condition. At the same time, the established curve model is adjusted using the data in Table 4 to obtain a curve model between the change in reservoir pressure, the effective water injection rate and the daily liquid production.

[0087] Take the reservoir pressure difference data, effective water injection volume data, and daily liquid production data in Table 5, take the reservoir pressure difference and daily liquid production as known quantities, use the established curve model to predict the effective water injection volume, and compare and verify the prediction results with the collected effective water injection volume data. According to the verification results, the adjusted curve model is smoothed to obtain the final curve model, as shown in the figure. Figure 1 In actual offshore oilfield production, the effective water injection volume can be predicted based on the actual oil well oil layer pressure difference and liquid production, using a curve model that establishes the relationship between the change in oil layer pressure, the effective water injection volume, and the daily liquid production. This can then be compared with the actual effective water injection volume, allowing timely measures to increase production and injection.

[0088] By adopting the scatter point fitting method, the curve model between daily oil production, water cut and injection-production ratio is established using the data in Table 3. At the same time, the established curve model is adjusted using the data in Table 4 to obtain the curve model between daily oil production, water cut and injection-production ratio, as shown in the figure: Figure 2 shown.

[0089] Take the water content data, injection-production ratio data and daily oil production data in Table 5, take the water content data and injection-production ratio data as known quantities, and use Figure 2 The daily oil production is predicted using the curve model in the figure, and the prediction results are compared and verified with the collected daily oil production data. According to the verification results, the adjusted curve model is smoothed to obtain the final curve model, as shown in the figure. Figure 2 Therefore, in actual offshore oilfield production, the daily oil production can be predicted based on the actual effective water injection volume, liquid production volume, and water production data using a curve model between daily oil production, water cut, and injection-production ratio. This can then be compared with the actual daily oil production, allowing for timely implementation of measures to increase production and injection.

[0090] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for analyzing and predicting offshore oilfield dynamics based on data-driven and knowledge engineering, characterized in that: The steps include: S1. Acquire multiple groups of historical oilfield single-well production data from offshore oilfields that have completed production, wherein the oilfield single-well production data includes reservoir pressure data of single-well oil wells, daily liquid production and daily oil production of single-well oil wells, and effective water injection volume data of single-well water injection wells; S2, processing the oil field single well production data obtained in step S1; S3, respectively dividing the data in the reservoir pressure difference data group R, the effective water injection data group D, the daily liquid production data group B, the single well injection-production ratio data group H, the water content data group M and the daily oil production data group C into a modeling group, a training group and a test group according to a set ratio; S4, taking the modeling group data in the reservoir pressure difference data group R, the effective water injection volume data group D, and the daily liquid production data group B, and adopting the scatter curve fitting method to establish a curve model between the change of reservoir pressure, the effective water injection volume data and the liquid production, specifically, taking the reservoir pressure difference in the reservoir pressure difference data group R as a known condition, taking the daily liquid production in the daily liquid production data group B as an independent variable, and taking the effective water injection volume data in the effective water injection volume data group D as a dependent variable, and establishing a curve model between the effective water injection volume data and the liquid production based on the reservoir pressure difference; S5, taking the training group data from the oil layer pressure difference data group R, the effective water injection data group D, and the daily liquid production data group B, and comparing the scatter point values ​​of the curve model established in step S4, and adjusting the curve model according to the comparison result, that is, adjusting the curve model closer to the corresponding coordinate point in the training group data; S6, taking the test group data from the reservoir pressure difference data group R, the effective water injection volume data group D, and the daily liquid production data group B, taking the reservoir pressure difference and the daily liquid production as known quantities, using the adjusted curve model to predict the effective water injection volume data, and comparing and verifying with the data in the effective water injection volume data group D, and smoothing the abnormal line segments of the curve model according to the verification results; S7, when actually exploiting an offshore oil field, according to the actual oil well oil layer pressure difference and daily liquid production, the effective water injection volume data is predicted using the curve model after smoothing in step S6, and the predicted result is compared with the actual effective water injection volume data, so as to timely adjust the effective water injection volume data of a single well in the oil field and make measures to increase production and injection; S8, taking the modeling group data in the water-cut data group M, the single-well injection-production ratio data group H and the daily oil production data group C, taking the data group obtained by dividing the water content in the water-cut data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H and the daily oil production data group C as data sources, adopting the scatter curve fitting method to establish a curve model between daily oil production, water content and single-well injection-production ratio, specifically, taking the data obtained by dividing the water content in the water-cut data group M by the single-well injection-production ratio in the single-well injection-production ratio data group H as the independent variable, and the daily oil production in the daily oil production data group C as the dependent variable, and establishing a curve model between daily oil production, water content and single-well injection-production ratio; S9, taking the training group data in the water-cut data group M, the single-well injection-production ratio data group H and the daily oil production data group C, and comparing the scatter point values ​​of the curve model established in step S8, and adjusting the curve model according to the comparison result, that is, adjusting the curve model closer to the corresponding coordinate point in the training group data; S10, taking the test group data in the water content data group M, the single well injection-production ratio data group H and the daily oil production data group C, taking the water content and the daily oil production as known quantities, using the curve model adjusted in step S9 to predict the single well injection-production ratio, comparing and verifying the prediction results with the data in the single well injection-production ratio data group H, and smoothing the abnormal line segments of the curve model according to the verification results; S11. During actual offshore oilfield exploitation, the single well injection-production ratio is predicted using the curve model after smoothing in step S9 according to the actual water content of the oil well and the daily oil production, and compared with the actual single well injection-production ratio, so as to timely adjust the water content and injection-production ratio of the single well in the oilfield and take measures to increase production and injection.

2. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: The historical oil fields described in step S1 are all oil wells whose oil well generation capacity meets the standards.

3. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: Step S1 includes: S101, obtaining the total amount of oil layer pressure changes and oil layer production depths of multiple groups of historical oil field single well production, and establishing oil layer sub-data groups, which are respectively recorded as oil layer sub-data group A1, oil layer sub-data group A2, ... and oil layer sub-data group AN, where N is the total number of oil wells; S102, obtaining the daily liquid production of the single well in the oil field, the daily liquid production includes the oil production and the water production, which are respectively recorded as daily liquid production sub-data group B1, daily liquid production sub-data group B2, ... and daily liquid production sub-data group BN, forming a daily liquid production data group B; S103, obtaining the daily oil production of each oil well of the single well in the oil field, recording them as daily oil production quantum data C1, daily oil production quantum data C2, ... and daily oil production quantum data CN, and forming a daily oil production data group C; S104, obtaining the effective water injection volume data of each water injection well in the single well exploitation of the oil field, and recording them as effective water injection volume data sub-data D1, effective water injection volume data sub-data D2, ... and effective water injection volume data sub-data DN, to form effective water injection volume data group D.

4. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: In the oilfield single well production data described in step S1, there are differences between the same type of data of the oilfield single wells.

5. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: Step S2 includes: S201, respectively divide the total amount of oil well reservoir pressure changes in each of the reservoir sub-data groups A1, A2, ... and AN by the reservoir production depth, to obtain the unit depth change of reservoir pressure of the oil wells in each reservoir sub-data group, which are recorded as reservoir pressure difference r1, reservoir pressure difference r2, ... and reservoir pressure difference rN, to form a reservoir pressure difference data group R, where N is the total number of oil wells; S202, respectively divide the effective water injection volume data sub-data D1, the effective water injection volume data sub-data D2, ... and the effective water injection volume data sub-data DN by the daily liquid production sub-data group B1, the daily liquid production sub-data group B2, ... and the daily liquid production sub-data group BN, to obtain the single-well injection-production ratio of the oil well, which are recorded as the single-well injection-production ratio of the oil well h1, the single-well injection-production ratio of the oil well h2, ... and the single-well injection-production ratio of the oil well hN, to form the single-well injection-production ratio data group H; S203. Divide the water production in each daily liquid production sub-data group B1, daily liquid production sub-data group B2, ... and daily liquid production sub-data group BN by the corresponding daily liquid production to obtain the water content of the corresponding oil well, which are recorded as water content m1, water content m2, ... and water content mN, respectively, to form water content data group M.

6. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: The division into modeling group, training group and test group according to the set ratio described in step S3 includes: division into modeling group, training group and test group according to the ratio of 7:2:1; or division into modeling group, training group and test group according to the ratio of 6:2:2; or division into modeling group, training group and test group according to the ratio of 5:3:

2.

7. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: The modeling group, training group and test group divided in step S3 are randomly selected from the data in the same group to ensure that the data changes between the divided data groups in the same group are discontinuous.

8. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 1 is characterized in that: The comparison of the scatter values ​​described in step S5 and step S9 is specifically as follows: (1) Taking the training set data corresponding to the curve model, and dividing the corresponding data in the training set data into independent variables and dependent variables; (2) Marking the coordinates of the independent variables and dependent variables in the same group of the training group data in the curve model between the effective water injection volume data and the liquid production volume based on the reservoir pressure difference established in step S4, or in the curve model between the daily oil production and the water content and the single well injection-production ratio established in step S8; (3) The curve model established in step S4 or step S8 is adjusted to be closer to the position of the coordinate point in the training set data, so that the curve model coincides with the coordinate point of the training set data.

9. The offshore oilfield dynamic analysis and prediction method based on data-driven and knowledge engineering according to claim 8 is characterized in that: For step S5, specifically, the training group data in the reservoir pressure difference data set R, the effective water injection data set D, and the daily liquid production data set B are taken, the daily liquid production in the daily liquid production data set B is taken as the independent variable, and the effective water injection data in the effective water injection data set D is taken as the dependent variable; For step S9, specifically, the training group data in the water content data group M, the single well injection-production ratio data group H and the daily oil production data group C are taken, and the data obtained by dividing the water content in the water content data group M by the single well injection-production ratio in the single well injection-production ratio data group H is used as the independent variable, and the daily oil production in the daily oil production data group C is used as the dependent variable.

Images