Inspiration profile prediction method and device in gas drive development

Abstract

The invention belongs to the technical field of carbon dioxide injection oil and gas development, and discloses an inspiration profile prediction method and device in gas drive development, and the method comprises the steps: determining the basic characteristics of a reservoir in a research area based on drilling coring and analysis test data, logging data and production dynamic data; based on the basic characteristics of the reservoir in the research area, determining a plurality of main control geological factors of the air suction capacity of the reservoir; constructing a typical geologic model based on the main control geologic factors of the reservoir gas suction capacity and the actual geologic parameters of the research area; an intelligent prediction model of an air suction profile is established based on simulated air suction profile data of a typical geologic model and actually measured air suction profile test data. According to the method, information of multiple aspects such as logging, geology and production dynamics is integrated, prediction of the gas injection well gas suction profile at the unknown time point is achieved, and therefore the purposes of accurately splitting the gas injection amount and reducing the production cost can be achieved.

Description

Technical Field

[0001] The invention belongs to the technical field of carbon dioxide injection oil and gas development, and in particular relates to a method and a device for predicting a gas suction profile in gas drive development. Background Art

[0002] In recent years, with the deepening of exploration and development, domestic old oil fields have gradually entered the middle and late stages of development. The development of oil reservoirs has exposed the contradictions of rapid production decline, rapid decline in oil production rate, and rapid increase in water content. In order to improve the efficiency of oil reservoir development, some oil reservoirs have adopted CO2 flooding development. CO2 flooding development is an effective oil recovery development method, similar to water injection development, which aims to improve the recovery rate of oil reservoirs. During the development process, the injected CO2 can improve the reservoir properties after entering the oil layer, and can reduce the pollution caused by the early development of some oil layers. CO2 flooding is not affected by high temperature and high salt. This mining method is suitable for sandstone reservoirs with large reserves and completely exhausted water flooding, and is also suitable for low permeability reservoirs with poor water flooding effect. In order to improve the CO2 flooding development effect of oil reservoirs, it is necessary to clarify the gas absorption capacity of different sand bodies and different reservoir types during the development process.

[0003] After the CO2 flooding development test, it is necessary to clarify the main controlling factors of the gas absorption capacity, and the gas absorption capacity of different reserves needs to be studied in depth. In addition, the research on the gas absorption capacity of the reservoir during CO2 flooding is also affected by human factors such as the technical level and professional experience of researchers. It is difficult to unify the research standards and the results are uncertain.

[0004] The above problems seriously restrict the dynamic evaluation and development effect of CO2 injection development. The method of combining artificial intelligence algorithms with actual test data to predict the gas absorption profile of production wells will change the current situation of difficult gas injection development and help promote the development of enhanced recovery and carbon sequestration in old oil fields. Summary of the invention

[0005] In order to overcome the defects of the above-mentioned prior art, the purpose of the present invention is to provide a method for predicting the carbon dioxide absorption amount of production wells using artificial intelligence algorithms and conventional test data in order to address the problems and shortcomings such as the numerous factors affecting the reservoir absorption capacity and the complex correlation between multiple factors during CO2 oil recovery, thereby providing a theoretical basis and data foundation for improving the CO2 oil recovery effect.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for predicting a gas suction profile in gas drive development, comprising:

[0008] Determine the basic characteristics of the reservoir in the study area based on drilling coring and analysis of test data, logging data, and production performance data;

[0009] Based on the basic characteristics of the reservoir in the study area, determine the main geological factors controlling the gas absorption capacity of the reservoir;

[0010] A typical geological model is constructed based on the main controlling geological factors of the reservoir's gas absorption capacity and the actual geological parameters of the study area;

[0011] By adopting machine learning algorithm, based on the simulated gas suction profile data of typical geological models and the actual measured gas suction profile test data, an intelligent prediction model of the gas suction profile is established to realize the prediction of the gas suction profile of gas injection wells.

[0012] In some embodiments of the present invention, the basic characteristics of the reservoir in the study area are determined based on drilling coring and analysis of test data, logging data, and production performance data, including:

[0013] The logging response anomalies of production wells are used to interpret the extent and type of reservoirs. The logging response anomalies include natural gamma and acoustic time difference.

[0014] In some embodiments of the present invention, the basic characteristics of the reservoir in the study area include rock type, pore structure and physical properties.

[0015] In some embodiments of the present invention, based on the basic characteristics of the reservoir in the study area, multiple main controlling geological factors of the reservoir's gas absorption capacity are determined, including:

[0016] The Pearson correlation coefficient method, univariate feature selection, recursive feature elimination method and random forest method were used to analyze the correlation between the basic characteristics of the reservoir and the relative gas absorption capacity of the reservoir.

[0017] In some embodiments of the present invention, a typical geological model is constructed based on the main controlling geological factors of the reservoir gas absorption capacity and the actual geological parameters of the study area, including:

[0018] The single controlling geological factor is changed to perform simulation and obtain model training samples.

[0019] In some embodiments of the present invention, an intelligent prediction model of the inhalation profile is established, including:

[0020] A multi-layer model suitable for the main production layer is established based on the LSTM neural network.

[0021] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0022] The output layer of the intelligent prediction model is set to a fully connected layer with 5 neurons.

[0023] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0024] Add multiple layers of invalid data to the typical geological model to make the data shape of the typical geological model consistent with the intelligent prediction model.

[0025] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0026] The key controlling factor of the intelligent prediction model is the feature selection result.

[0027] In a second aspect, the present invention discloses a device for predicting a gas suction profile in gas drive development, comprising:

[0028] The basic reservoir characteristics determination unit is used to determine the basic reservoir characteristics of the study area based on drilling coring and analysis of test data, logging data, and production dynamic data;

[0029] A main controlling geological factor determination unit is used to determine multiple main controlling geological factors of the reservoir's gas absorption capacity based on the basic characteristics of the reservoir in the study area;

[0030] Typical geological model construction unit, used to construct a typical geological model based on the main controlling geological factors of reservoir gas absorption capacity and the actual geological parameters of the study area;

[0031] The gas suction profile prediction unit is used to establish an intelligent prediction model of the gas suction profile based on the simulated gas suction profile data of the typical geological model and the actual measured gas suction profile test data by adopting a machine learning algorithm, so as to realize the prediction of the gas suction profile of the gas injection well.

[0032] In a third aspect, the present invention discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for predicting an air intake profile in gas drive development when executing the program.

[0033] In a fourth aspect, the present invention discloses a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-mentioned method for predicting an air intake profile in gas drive development.

[0034] Technical effects and advantages of the present invention:

[0035] 1. The present invention can be used for gas drive development of oil fields in different regions and with different production histories. It aims at the problems and shortcomings of numerous factors affecting the gas absorption capacity of reservoirs and the complex correlation between multiple factors during CO2 oil recovery. Guided by theories such as big data and artificial intelligence, oil and gas geology, and reservoir engineering, the present invention systematically conducts analysis of the main controlling factors of gas absorption capacity, intelligent prediction of gas absorption profiles, and establishes a prediction model based on basic geological research, thus making up for the lack of traditional prediction of gas absorption capacity of reservoirs.

[0036] 2. The present invention integrates information from various aspects such as well logging, geology, and production dynamics to achieve the prediction of the gas injection profile of the gas injection well at an unknown time point, thereby achieving the purpose of accurately splitting the gas injection volume and reducing production costs.

[0037] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A flow chart of a method for predicting a gas suction profile in gas drive development according to the present invention;

[0039] Figure 2 This is a columnar result diagram of the fine stratigraphic grid division of the Q3-3 well in the embodiment of the present invention;

[0040] Figure 3 This is a comparative cross-section result diagram of the formations of the wells Q5-1 to Q5-3 to Q+01-3 to QD11-9 in the embodiment of the present invention;

[0041] Figure 4 This is a graph showing the analysis results of the Pearson correlation coefficient method according to an embodiment of the present invention;

[0042] Figure 5 This is a graph showing the analysis results of the univariate feature selection method according to an embodiment of the present invention;

[0043] Figure 6 This is a graph of random forest feature selection results in an embodiment of the present invention;

[0044] Figure 7 This is a feature selection result diagram of the recursive feature elimination method according to an embodiment of the present invention;

[0045] Figure 8 This is a result diagram of the correlation analysis between the perforation thickness and the relative suction volume of the Q5-1 well in the embodiment of the present invention;

[0046] Fig. 9 This is a result diagram of the analysis of the relative air intake change law of the stratum thickness in the positive arrangement of the five-layer mechanism model in the embodiment of the present invention;

[0047] Fig.10 This is a schematic diagram of the structure of the inhalation profile prediction model based on the LSTM algorithm according to an embodiment of the present invention;

[0048] Fig.11 This is a diagram showing the prediction result of the gas suction profile of the Q5-1 well based on the five-layer model according to an embodiment of the present invention;

[0049] Fig.12It is a structural schematic diagram of a suction profile prediction device in gas drive development according to the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] The concept of the present invention is to systematically carry out fine stratigraphic division and comparison, genetic sand body type and characteristic analysis and other research guided by the theories of sequence stratigraphy, sedimentology, oil geology, reservoir geology, etc., to reveal the rock type, pore structure, physical properties and other characteristics of the reservoir. Then, in order to reveal the gas absorption capacity of the target stratum in the study area, the basic characteristic data of the gas absorption profile of each single well was established based on the test data of the gas absorption profile, wherein the static geological characteristics include the rock type, pore structure, physical properties, porosity, permeability, maximum permeability deviation, average permeability deviation (characterizing the permeability difference within a single sand body), RT, RILD, RLLS, AC, VSH, logging interpretation conclusions, etc. of the reservoir, and the production dynamic characteristics include the production mode, cumulative gas injection volume, injection pressure, etc. Since the parameter characterizing the permeability difference between the sand bodies of a single well has only one value for a single well and there are few samples, the influence of the permeability difference between the sand bodies of a single well on the gas absorption capacity of the reservoir is directly selected through the plate method and the numerical simulation results. Feature selection refers to the process of selecting relevant feature subsets for machine learning model construction. Finally, in view of the fact that there are many factors affecting the gas profile and the mapping model has a nonlinear relationship in the process of CO2 flooding development in oil reservoirs, an intelligent prediction method for the gas profile based on the LSTM neural network model is proposed, considering the temporal problem of basic data in the gas profile prediction process. Based on the gas profile data simulated by typical geological models and the actual measured gas profile data, the LSTM neural network model architecture is built, and on this basis, an intelligent prediction model for the gas profile is trained and generated, thereby realizing the prediction of the gas profile of the injection well at an unknown time, providing a theoretical basis and technical support for accurate gas injection volume splitting.

[0052] like Figure 1 As shown, the present invention provides a method for predicting a gas suction profile in gas drive development, comprising:

[0053] Determine the basic characteristics of the reservoir in the study area based on drilling coring and analysis of test data, logging data, and production performance data;

[0054] Based on the basic characteristics of the reservoir in the study area, determine the main geological factors controlling the gas absorption capacity of the reservoir;

[0055] A typical geological model is constructed based on the main controlling geological factors of the reservoir's gas absorption capacity and the actual geological parameters of the study area;

[0056] By adopting machine learning algorithm, based on the simulated gas suction profile data of typical geological models and the actual measured gas suction profile test data, an intelligent prediction model of the gas suction profile is established to realize the prediction of the gas suction profile of gas injection wells.

[0057] In some embodiments of the present invention, the basic characteristics of the reservoir in the study area are determined based on drilling coring and analysis of test data, logging data, and production performance data, including:

[0058] The logging response anomalies of production wells are used to interpret the extent and type of reservoirs. The logging response anomalies include natural gamma and acoustic time difference.

[0059] In some embodiments of the present invention, the basic characteristics of the reservoir in the study area include rock type, pore structure and physical properties.

[0060] In some embodiments of the present invention, based on the basic characteristics of the reservoir in the study area, multiple main controlling geological factors of the reservoir's gas absorption capacity are determined, including:

[0061] The Pearson correlation coefficient method, univariate feature selection, recursive feature elimination method and random forest method were used to analyze the correlation between the basic characteristics of the reservoir and the relative gas absorption capacity of the reservoir.

[0062] In some embodiments of the present invention, a typical geological model is constructed based on the main controlling geological factors of the reservoir gas absorption capacity and the actual geological parameters of the study area, including:

[0063] The single controlling geological factor is changed to perform simulation and obtain model training samples.

[0064] In some embodiments of the present invention, an intelligent prediction model of the inhalation profile is established, including:

[0065] A multi-layer model suitable for the main production layer is established based on the LSTM neural network.

[0066] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0067] The output layer of the intelligent prediction model is set to a fully connected layer with 5 neurons.

[0068] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0069] Add multiple layers of invalid data to the typical geological model to make the data shape of the typical geological model consistent with the intelligent prediction model.

[0070] In some embodiments of the present invention, establishing an intelligent prediction model for the inhalation profile further includes:

[0071] The key controlling factor of the intelligent prediction model is the feature selection result.

[0072] In a second aspect, the present invention also discloses a device for predicting the suction profile in gas drive development, such as Fig.12 As shown, including:

[0073] The basic reservoir characteristics determination unit is used to determine the basic reservoir characteristics of the study area based on drilling coring and analysis of test data, logging data, and production dynamic data;

[0074] A main controlling geological factor determination unit is used to determine multiple main controlling geological factors of the reservoir's gas absorption capacity based on the basic characteristics of the reservoir in the study area;

[0075] Typical geological model construction unit, used to construct a typical geological model based on the main controlling geological factors of reservoir gas absorption capacity and the actual geological parameters of the study area;

[0076] The gas suction profile prediction unit is used to establish an intelligent prediction model of the gas suction profile based on the simulated gas suction profile data of the typical geological model and the actual measured gas suction profile test data by adopting a machine learning algorithm, so as to realize the prediction of the gas suction profile of the gas injection well.

[0077] In a third aspect, the present invention further discloses an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the above-mentioned method for predicting an intake profile in gas drive development when executing the program.

[0078] In a fourth aspect, the present invention further discloses a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned method for predicting an air intake profile in gas drive development.

[0079] In order to better illustrate the present solution, the following embodiments are also provided.

[0080] Example

[0081] The method of the present invention is used to study the intelligent prediction model of the gas absorption profile in the CO2 flooding test area of ​​the Hei 46 block of Jilin Oilfield, to study the factors and methods affecting the gas absorption capacity of the reservoir during the CO2 flooding process, and to establish an intelligent prediction model.

[0082] Step 1: Determine the basic characteristics of the reservoir in the study area based on the core sampling and analysis of laboratory data, logging data, and production dynamic data in the Heilongjiang 46 area of ​​Jilin Oilfield.

[0083] Specifically, it includes detailed stratigraphic division and comparison of the basic geological research part of Heilongjiang Oilfield Block 46. For stratigraphic units of different scales, the analytical methods and technical means of stratigraphic division and comparison often differ to a certain extent. In order to achieve the comparison of strata of different scales in the study area, the technical methods of "base level cycle control, marker layer constraint, and lithofacies difference analysis" are adopted. According to the geological conditions of the study area and the characteristics of well logging data, a detailed stratigraphic framework of the study area is established.

[0084] Figure 2 This is a columnar result diagram of the fine stratigraphic framework division of Well Q3-3 in the study area. According to the principle of consistency of stratigraphic superposition style, the cyclic changes in the stratigraphic base level control the sedimentary style, different sedimentary microfacies types and spatial relationship distribution of the strata. During the period of rising stratigraphic base level, the accommodative space increases, the relative water body becomes deeper, and the strata present a retrograde superposition style; when the stratigraphic base level drops, the accommodative space decreases, the relative water body becomes shallower, and the strata present a prograde superposition style. Therefore, based on the cyclicity of stratigraphic units and the characteristics of sand body thickness changes, isochronous stratigraphic correlation can be performed. For base level cycles of the same order, the superposition style of the strata is the same. It should be noted that this superposition style of stratigraphic progradation, retrogradation or aggradation is different from the rhythmicity of stratigraphic units caused by cycles. The strata on a single well are divided according to this principle.

[0085] Figure 3 This is the result of the stratigraphic correlation profile of Q5-1~Q5-3~Q+01-3~QD11-9 wells. Following the principle of stratigraphic thickness consistency and marker layer constraint, the so-called "stratigraphic thickness consistency" refers to the regular changes in the stratigraphic thickness of the sedimentary units developed under the control of a single stratigraphic base level cycle in the vertical and along-source directions, such as similar thickness, monotonically increasing or monotonically decreasing. In the direction of the provenance, the stratigraphic thickness of each small layer of the target layer Qingyi is basically consistent, with no obvious thickness changes. The marker layer refers to a stratum with special lithology, stable thickness, wide distribution range, and typical logging response characteristics. It has isochronous sedimentary properties and is often an important basis for stratigraphic division and comparison within the block. In the stratigraphic correlation process of the study area, the main marker layer used is mudstone, among which the mudstone at the top and bottom of Qingyi is stably distributed and is an obvious marker layer. During the stratigraphic division and comparison process, the marker layer is the gray-black mudstone at the top of the first section of Qing, which is a lake phase. It has a stable distribution in the plane and can be distinguished from the bottom strata of the second section of Qing; the 16th layer at the bottom of the first section of Qing is a stable mudstone section, which is electrically characterized by high natural gamma, high natural potential and high acoustic time difference, which can be distinguished from the top strata of the fourth section of the underlying Quan; the electrical characteristics of the 3rd, 9th and 13th layers are also relatively obvious, which can be used as a reference interface for internal layer comparison. In this way, the stratigraphic division and comparison of the wells were carried out. Then, the types and characteristic analysis of the genetic sand bodies were studied to reveal the rock type, pore structure, physical properties and other characteristics of the reservoir, and statistics were performed.

[0086] Step 2: Based on the basic characteristics of the reservoir in the study area, determine the multiple main geological factors that control the reservoir's gas absorption capacity.

[0087] Specifically, based on the study of the basic characteristics of the reservoir in the study area, the Pearson coefficient method, univariate feature selection method, random forest algorithm, and recursive feature elimination method were used to analyze the correlation between the basic characteristics of the reservoir and the relative gas absorption of the reservoir through static geological parameters, dynamic factors and other data, and the characteristic correlation coefficient was normalized. Finally, after a comprehensive analysis by the voting method, it was concluded that the main controlling factors affecting the initial gas absorption profile of drilling were the interlayer heterogeneity characteristics of reservoir thickness, reservoir physical properties, and permeability.

[0088] Figure 4 This is the result of the Pearson correlation coefficient analysis. The Pearson coefficient measures the strength and direction of the linear relationship between two continuous variables. It calculates a value between -1 and 1, which is statistical evidence of whether there is a linear relationship between the two features, where 0 indicates no correlation, 1 indicates a perfect positive correlation, and -1 indicates a perfect negative correlation. The Pearson coefficient method can be used to quickly identify features with obvious linear correlations. The main controlling geological factors of the air intake profile obtained by the Pearson correlation coefficient method are layer thickness, Kh, maximum permeability deviation, and permeability.

[0089] Figure 5 Analysis result graph for univariate feature selection. Another method of feature selection is to pre-screen predictors using simple univariate statistical indicators, and then only use predictors that pass certain criteria in subsequent model steps. Similar to the subsequent recursive selection, the cross-validation of the subsequent model will have a certain bias because the remaining predictors have been evaluated on the data set, and this bias needs to be reduced by resampling. The main controlling geological factors of the gas-inhalation profile obtained by the univariate feature selection method are layer thickness, Kh, maximum permeability deviation, and shale content VSH.

[0090] Figure 6This is the result of random forest feature selection. In essence, random forest feature selection is a bagging algorithm composed of a bunch of basic estimators (decision trees). Each tree is built on the basis of randomly extracting observations and features from the data set, but not every tree can see all the features or all the observations, which also ensures that the trees are uncorrelated and therefore not prone to overfitting. Among them, each tree is a series of yes or no questions based on a single or combined feature. In each split, the question divides the data set into two buckets, each of which contains observations that are different from the other bucket but more similar to each other. Therefore, the importance of each feature depends on the "purity" of each bucket. For classification problems, the measure of "purity" is usually the Gini index or information gain / entropy. For regression problems, the measure of error is variance. Therefore, when training a tree, you can calculate how much "purity" each feature has reduced. The less "purity" the feature, the more important it is. In random forest, the impurity reduction of each feature can be used to determine the final importance of the variable by tree averaging. The main controlling geological factors of the gas absorption profile obtained by the random forest model feature selection method are layer thickness, minimum permeability deviation, Kh, and mud content.

[0091] Figure 7 This is the result of feature selection by recursive feature elimination. Recursive feature elimination is a package-based feature selection algorithm that also uses filter-based feature selection internally. Different models are adapted and the features with the weakest correlation with the target in the data are removed until the specified number of features is reached. This method is achieved by fitting the given machine learning algorithm used in the core of the model, sorting the features by importance, discarding the least important features, and refitting the model. This process is repeated until the specified number of features are retained. This recursive feature selection method can eliminate the dependencies and collinearity that may exist in the model to a certain extent. The main controlling geological factors of the gas suction profile obtained by the recursive feature elimination method are layer thickness, Kh, minimum permeability deviation, and average permeability deviation.

[0092] By analyzing the specific results of the feature selection of the above machine learning algorithm, the feature correlation coefficient is normalized for easy comparison. Finally, after comprehensive analysis by the voting method, it is concluded that the main controlling geological factors affecting the initial gas absorption profile of drilling are the interlayer heterogeneity characteristics of reservoir thickness, reservoir physical properties, and permeability.

[0093] Step 3: Construct a typical geological model based on the main controlling geological factors of the reservoir's gas absorption capacity and the actual geological parameters of the study area.

[0094] Figure 8This is the result of the correlation analysis between the perforation thickness and relative suction volume of the Q5-1 well. Because the work area is a general gas injection, the relative suction capacity of the reservoir is affected by the injection position. If some sand bodies are not perforated, they will not be able to effectively absorb gas. Therefore, the perforation thickness of the reservoir is more correlated with the relative suction capacity of the reservoir. And previous studies have shown that the factor most correlated with the suction capacity of the reservoir is the perforation thickness of the reservoir. The Q5-1 well has been tested for water and gas absorption profiles many times. The research shows that the reservoir thickness is linearly positively correlated with the relative suction volume of the reservoir. The greater the perforation reservoir thickness of a single small layer, the higher the relative suction volume of gas.

[0095] On the basis of the selection of gas absorption profile control factors and the test data of gas absorption profile, the influence of reservoir perforation thickness, reservoir physical properties, reservoir heterogeneity, reservoir production mode, water drive and CO2 drive alternating injection on the gas absorption capacity was analyzed.

[0096] Fig. 9 This is the analysis result diagram of the change law of relative air intake volume in the positive arrangement of the formation thickness of the five-layer mechanism model. On the basis of the analysis of the main controlling factors of the air intake profile, a typical geological model was constructed based on the actual geological parameters of the study area. The numerical simulation technology was used to simulate the air intake profile data under different well spacings, different permeabilities and grade differences, different development methods and other parameters, and random noise was inserted to construct an enhanced data set to provide training samples for the subsequent intelligent prediction model training.

[0097] Step 4: Using machine learning algorithms, based on the simulated gas suction profile data of typical geological models and the actual measured gas suction profile test data, an intelligent prediction model of the gas suction profile is established to predict the gas suction profile of the gas injection well.

[0098] Fig.10 This is a schematic diagram of the structure of the air intake profile prediction model based on the LSTM algorithm. When designing the model architecture, in order to further extract the implicit information contained in the time series, this application sets the time step of a single data to 20, and designs a 4-layer composite LSTM layer after the input layer, and sets the number of neurons in each layer to 91, 55, 11, and 99 respectively. The output layer is set to a fully connected layer with 5 neurons to output the air intake profile data predicted by the 5-layer typical geological model. For the 3-layer typical geological model, two layers of invalid data are filled to make its data shape consistent with the five-layer intelligent prediction model.

[0099] Fig.11In order to predict the gas absorption profile of the Q5-1 well based on the five-layer intelligent prediction model, some gas injection wells were selected as the target well group in each block, and the four geological factors corresponding to the development layer of the target well, such as cumulative gas injection volume, permeability, Kh, and formation thickness, were used as input features to establish the water absorption profile sample points of the corresponding wells. In order to verify the accuracy and reliability of the method of this application, five connected sample layers were selected as an input for each well to obtain the corresponding predicted gas absorption profile. The gas absorption profile results of the model fitting were compared with the actual gas absorption profile data, which basically conformed to the relative gas absorption volume of the actual reservoir. The relative gas absorption volume prediction results of a specific layer are shown in the figure. Fig.11 In addition, overall, the average fitting error is about 7.5%. Due to the limitation of the model data set, there are fewer data in the high-value area of ​​the gas suction volume, resulting in the interpretation space of the high-value area not being fully opened. Therefore, the model fitting effect in the low-value area of ​​the reservoir gas suction volume is better than the model fitting effect in the high-value area of ​​the reservoir gas suction volume. Fig.11 This phenomenon is characterized in detail.

[0100] In summary, the present invention can target gas drive development in oil fields in different regions and with different production histories, and target the problems and shortcomings of numerous factors affecting reservoir gas absorption capacity and complex correlation between multiple factors during CO2 oil recovery. Guided by theories such as big data and artificial intelligence, oil and gas geology, and reservoir engineering, on the basis of basic geological research, the present invention systematically conducts analysis of main controlling factors of gas absorption capacity, intelligent prediction of gas absorption profile, and other research, and establishes a prediction model, which makes up for the lack of traditional prediction of reservoir gas absorption capacity. On the other hand, the present invention integrates information from various aspects such as well logging, geology, and production dynamics to realize the prediction of gas absorption profiles of gas injection wells at unknown time points, thereby achieving the purpose of accurately splitting the gas injection volume and reducing production costs.

[0101] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for predicting the air suction profile in gas drive development, characterized in that: include: Determine the basic characteristics of the reservoir in the study area based on drilling coring and analysis of test data, logging data, and production performance data; Based on the basic characteristics of the reservoir in the study area, determine the main geological factors controlling the gas absorption capacity of the reservoir; A typical geological model is constructed based on the main controlling geological factors of the reservoir's gas absorption capacity and the actual geological parameters of the study area; By adopting machine learning algorithm, based on the simulated gas suction profile data of typical geological models and the actual measured gas suction profile test data, an intelligent prediction model of the gas suction profile is established to realize the prediction of the gas suction profile of gas injection wells.

2. The method for predicting the air suction profile in gas drive development according to claim 1, characterized in that: The basic characteristics of the reservoir in the study area are determined based on drilling coring and analysis of test data, logging data, and production performance data, including: The logging response anomalies of production wells are used to interpret the extent and type of reservoirs, including natural gamma and acoustic transit time.

3. The method for predicting the air suction profile in gas drive development according to claim 1, characterized in that: The basic characteristics of the reservoir in the study area include rock type, pore structure and physical properties.

4. The method for predicting the air suction profile in gas drive development according to claim 1, characterized in that: The multiple main geological factors that determine the gas absorption capacity of the reservoir based on the basic characteristics of the reservoir in the study area include: The Pearson correlation coefficient method, univariate feature selection, recursive feature elimination method and random forest method were used to analyze the correlation between the basic characteristics of the reservoir and the relative gas absorption capacity of the reservoir.

5. The method for predicting the air suction profile in gas drive development according to claim 1, characterized in that: The typical geological model constructed based on the main controlling geological factors of the reservoir gas absorption capacity and the actual geological parameters of the study area includes: The single controlling geological factor is changed to perform simulation and obtain model training samples.

6. The method for predicting the air suction profile in gas drive development according to claim 1, characterized in that: The intelligent prediction model for establishing the inhalation profile includes: A multi-layer model suitable for the main production layer is established based on the LSTM neural network.

7. The method for predicting the air suction profile in gas drive development according to claim 6, characterized in that: The intelligent prediction model for establishing the inhalation profile also includes: The output layer of the intelligent prediction model is set to a fully connected layer with 5 neurons.

8. The method for predicting the air suction profile in gas drive development according to claim 7, characterized in that: The intelligent prediction model for establishing the inhalation profile also includes: Multiple layers of invalid data are added to the typical geological model to make the data shape of the typical geological model consistent with the intelligent prediction model.

9. A method for predicting a gas suction profile in gas drive development according to any one of claims 6 to 8, characterized in that: The intelligent prediction model for establishing the inhalation profile also includes: The key controlling factor of the intelligent prediction model adopts the feature selection result.

10. A device for predicting the air suction profile in gas drive development, characterized in that: include: The basic reservoir characteristics determination unit is used to determine the basic reservoir characteristics of the study area based on drilling coring and analysis of test data, logging data, and production dynamic data; A main controlling geological factor determination unit is used to determine multiple main controlling geological factors of the reservoir's gas absorption capacity based on the basic characteristics of the reservoir in the study area; Typical geological model construction unit, used to construct a typical geological model based on the main controlling geological factors of reservoir gas absorption capacity and the actual geological parameters of the study area; The gas suction profile prediction unit is used to establish an intelligent prediction model of the gas suction profile based on the simulated gas suction profile data of the typical geological model and the actual measured gas suction profile test data by adopting a machine learning algorithm, so as to realize the prediction of the gas suction profile of the gas injection well.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the steps of a method for predicting an air intake profile in gas drive development as described in any one of claims 1 to 9 are implemented.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of a method for predicting a gas suction profile in gas drive development as described in any one of claims 1 to 9 are implemented.

Images