Conglomerate fluid property identification method and device and electronic equipment

Abstract

The invention discloses a conglomerate fluid property identification method, a conglomerate fluid property identification device and electronic equipment. The conglomerate fluid property identification method comprises the following steps: acquiring logging curve data of a to-be-identified well; the logging curve data are input into a preset text convolutional neural network model, reservoir fluid properties of the to-be-recognized well are recognized, a fluid property result map of the to-be-recognized well is obtained, and the fluid property result map comprises probability vectors of different types of lithology in the to-be-recognized well. According to the method, the to-be-recognized well is interpreted and analyzed through the text convolutional neural network model, feature information with richer levels is obtained, and the efficiency and accuracy of recognizing the fluid properties of the complex conglomerate reservoir by the interpretation model are improved.

Description

Technical Field

[0001] The invention relates to the field of oil and gas field exploration, and in particular to a method, a device and an electronic device for identifying conglomerate fluid properties. Background Art

[0002] Reservoir fluid property identification is an important research topic in reservoir evaluation and oil reservoir evaluation. Accurate reservoir fluid property identification results are an effective basis for geological exploration and development. At present, reservoir fluid property identification is completely completed manually by operators with geological knowledge background, mainly using technical methods such as cross plot method, statistical method and nuclear magnetic logging. With the continuous improvement of exploration level, reservoir types are becoming more and more complex, and the amount of data is increasing. Traditional identification methods have low accuracy, low efficiency and too much influence from human factors, and are no longer suitable for widespread application.

[0003] In the past two years, artificial intelligence technology has been applied in the field of well logging interpretation, but most of them are the application of fully connected neural networks or conventional convolutional neural networks to identify lithology. The accuracy and efficiency of identification need to be improved, and there are major shortcomings and limitations. In terms of identification accuracy: the above method still follows the traditional well logging interpretation operation mode, identifying characteristic change points based on the original logging curve data, and the error is relatively large. In terms of interpretation efficiency: although the above method enables the well logging interpretation model to have the ability of deep learning, its interpretation efficiency and accuracy need to be improved when dealing with complex lithology logging interpretation. Summary of the invention

[0004] The embodiments of the present application provide a method, device and electronic device for identifying the properties of conglomerate fluids. The method interprets and analyzes the wells to be identified through a text convolutional neural network model to obtain feature information with richer levels, thereby improving the efficiency and accuracy of the interpretation model in identifying the fluid properties of complex conglomerate reservoirs.

[0005] In a first aspect, the present invention provides the following technical solution through an embodiment of the present invention:

[0006] A method for identifying conglomerate fluid properties, comprising:

[0007] Acquire logging curve data of the well to be identified; input the logging curve data into a preset text convolutional neural network model to identify the reservoir fluid properties of the well to be identified, and obtain a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different categories of lithology in the well to be identified.

[0008] Preferably, the text convolutional neural network model is trained according to the following steps: obtaining sample data, the sample data including: sample logging curve data of multiple key wells and fluid property result graphs corresponding to each sample logging curve data; based on the sample data, training a pre-constructed initial network model to obtain the text convolutional neural network model.

[0009] Preferably, the sample logging curve data of the multiple key wells are obtained through the following steps: acquiring the logging curve data of the multiple key wells; performing secondary interpretation of the fluid properties of the multiple key wells based on quantitative identification criteria, and using the results of the secondary interpretation as input feature labels, wherein the quantitative identification criteria are formulated based on the logging response characteristics of different lithological fluid properties; based on the input feature labels, determining the logging curve data of the key wells as sample logging curve data.

[0010] Preferably, the fluid property result graph corresponding to each sample logging curve data is obtained by the following steps: for each sample logging curve data, one-dimensional discrete convolution is performed on the sample logging curve data to extract the extreme values ​​and inflection points of the sample logging curve data as potential stratification points of the sample logging curve data; based on the potential stratification points and preset significance indicators, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, wherein each curve segment includes corresponding curve segment data; word vectors corresponding to the plurality of curve segment data are determined; the plurality of curve segment data are mapped to the corresponding word vectors to obtain the word vector matrix of the plurality of curve segment data; the word vector matrix is ​​convolved and pooled to obtain the fluid property result graph corresponding to the sample logging curve data.

[0011] Preferably, based on the potential stratification points and preset significance indicators, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, including: based on the potential stratification points, the sample logging curve data is divided into a plurality of monotonic interval segments; based on the preset significance indicators and the plurality of monotonic interval segments, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined, and the potential reservoir segment division result of the sample logging curve data is obtained, so that the sample logging curve data is divided into a plurality of curve segments arranged in sequence.

[0012] Preferably, after obtaining the text convolutional neural network model, it also includes: acquiring test data, the test data including: multiple test curve data and a fluid property result graph corresponding to each test curve data; inputting the test curve data into the text convolutional neural network model to obtain a fluid property estimation graph; determining the error between the fluid property result graph and the fluid property estimation graph based on the test curve data, the fluid property estimation graph and a preset loss function; and optimizing the network weight parameters in the text convolutional neural network model based on the error.

[0013] Preferably, after acquiring the logging curve data of a plurality of key wells, the method further comprises: classifying the logging curve data based on the acquired external influencing factors of the logging curve data; and performing curve correction and normalization processing on the classified logging curve data.

[0014] Preferably, the external influencing factors include age and type of collection instrument.

[0015] In a second aspect, the present invention provides the following technical solution through an embodiment of the present invention:

[0016] A device for identifying conglomerate fluid properties, comprising:

[0017] An acquisition module, used for acquiring the logging curve data of the well to be identified;

[0018] The identification module is used to input the logging curve data into a preset text convolutional neural network model to identify the reservoir fluid properties of the well to be identified, and obtain a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

[0019] In a third aspect, the present invention provides the following technical solution through an embodiment of the present invention:

[0020] An electronic device comprises: a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of any one of the methods described in the first aspect are implemented.

[0021] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0022] The conglomerate fluid property identification method provided by the embodiment of the present invention collects well logging data, inputs the well logging data into a preset text convolutional neural network model, and automatically obtains a fluid property result map corresponding to the well logging data, and the fluid property result map includes probability vectors of different types of lithology in the well to be identified. Since the one-dimensional sequential arrangement of the well logging data is similar to text, and the process of obtaining fluid properties according to the curve data is similar to the text semantic analysis process, on the basis of the convolutional neural network, the network structure used for text processing is used for well logging interpretation and analysis, and more hierarchical feature information can be obtained. In terms of recognition accuracy, the text convolutional neural network model can identify feature change points based on the well logging data, and improves the recognition accuracy of the feature points of the well logging curve change compared to the traditional well logging interpretation operation mode; in terms of well logging data interpretation, the interpretation efficiency is compared with the general deep learning ability. When processing complex lithology well logging interpretation, the text convolutional neural network model can effectively extract the feature attributes of the well logging data, and has high interpretation efficiency and accuracy. Therefore, this method improves the efficiency and accuracy of the interpretation model in identifying the fluid properties of complex conglomerate reservoirs, and has the characteristics of high accuracy, strong applicability and simple process. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0024] Figure 1 A flow chart of a method for identifying conglomerate fluid properties provided by an embodiment of the present invention;

[0025] Figure 2 A schematic diagram of a well logging identification standard for complex conglomerate fluid properties provided by an embodiment of the present invention;

[0026] Figure 3 A schematic diagram of discrete convolution calculation curve feature points and division intervals provided in an embodiment of the present invention;

[0027] Figure 4 A schematic diagram of the basic structure of a text convolutional neural network provided by an embodiment of the present invention;

[0028] Figure 5 A schematic diagram of the artificial intelligence logging lithology identification results of Well-A provided in an embodiment of the present invention;

[0029] Figure 6 A schematic diagram of the structure of a device for identifying conglomerate fluid properties provided by an embodiment of the present invention;

[0030] Figure 7 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0031] After research, the inventor found that in the past two years, artificial intelligence technology has been applied in the field of well logging interpretation, but most of them use fully connected neural networks or conventional convolutional neural networks to identify lithology, and the accuracy and efficiency of identification need to be improved. In addition, the above methods all have their shortcomings and limitations. In terms of recognition accuracy, if you want to improve the accuracy of well logging interpretation, the key is to improve the recognition accuracy of the characteristic points of logging curve changes. However, the above methods still follow the traditional well logging interpretation operation mode and identify characteristic change points based on the original logging curve data, and the error is relatively large. In terms of interpretation efficiency, although the above method enables the well logging interpretation model to have the ability of deep learning, its ability to extract characteristic attributes is limited when dealing with complex lithology logging interpretation, and its interpretation efficiency and accuracy need to be improved.

[0032] In view of this, the embodiments of the present application provide a method, device and electronic device for identifying the properties of conglomerate fluids. The method interprets and analyzes the wells to be identified through a text convolutional neural network model to obtain feature information with richer levels, thereby improving the efficiency and accuracy of the interpretation model in identifying the fluid properties of complex conglomerate reservoirs.

[0033] The overall idea of ​​the technical solution of the embodiment of the present application is as follows:

[0034] A method for identifying conglomerate fluid properties comprises: obtaining logging curve data of a well to be identified; inputting the logging curve data into a preset text convolutional neural network model, identifying the reservoir fluid properties of the well to be identified, and obtaining a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

[0035] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0036] There are thousands of wells in a certain oil field, and the exploration degree is relatively high. Since the research target has not been studied as the main oil and gas strata in the past, the accuracy of logging data and logging interpretation is poor, and secondary interpretation of reservoir fluid properties is needed. Due to the large number of wells in the study area, the huge amount of logging data, and the complex rock types, it is difficult to ensure the efficiency and accuracy of manual interpretation. Therefore, the identification method provided by this application can be used for logging interpretation.

[0037] In a first aspect, an embodiment of the present invention provides a method for identifying conglomerate fluid properties. Specifically, Figure 1 As shown, the method includes the following steps S101 to S102:

[0038] Step S101, obtaining the logging curve data of the well to be identified;

[0039] Step S102, input the logging curve data into a preset text convolutional neural network model, identify the reservoir fluid properties of the well to be identified, and obtain a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

[0040] The logging curve data may include: natural potential, natural gamma (GR), well diameter, deep lateral, shallow lateral, acoustic time difference, compensated neutron and density.

[0041] In the specific implementation process, the text convolutional neural network model is trained according to the following steps: obtaining sample data, wherein the sample data includes: sample logging curve data of multiple key wells and fluid property result graphs corresponding to each sample logging curve data; based on the sample data, training the pre-constructed initial network model to obtain the text convolutional neural network model.

[0042] Specifically, the sample logging curve data of multiple key wells are obtained through the following steps: acquiring the logging curve data of multiple key wells; performing secondary interpretation of fluid properties of multiple key wells based on quantitative identification standards, and using the results of the secondary interpretation as input feature labels, wherein the quantitative identification standards are formulated based on the logging response characteristics of fluid properties of different lithologies; based on the input feature labels, determining the logging curve data of the key wells as sample logging curve data.

[0043] Among them, the key wells may be wells with relatively complete measured data.

[0044] In one embodiment, obtaining logging curve data of multiple key wells may include: obtaining logging data of all wells in the study area and organizing them in a unified manner, wherein the logging curve data may include: natural potential, natural gamma, well diameter, deep lateral, shallow lateral, acoustic wave time difference, compensated neutron and density.

[0045] In a specific embodiment, according to the known rock and mineral database and rock physics analysis conclusions, conglomerate is divided into multiple types of conglomerate, such as granite conglomerate, sandy conglomerate, and so on. Figure 2 The following are the identification plates of granite conglomerate oil layers and sandy conglomerate oil layers. It can be seen that the lithology of the two types of conglomerates is different, and the granite conglomerate has relatively good oil content as a whole.

[0046] Based on the test oil and production data, the logging response characteristics of different lithology fluid properties can be clarified, so as to formulate quantitative identification standards for different conglomerate fluid properties according to lithology. Based on the quantitative identification standards, secondary interpretation of fluid properties of key wells is carried out, and the interpretation results are used as input feature labels. Based on the input feature labels, the logging data samples of key wells are used as sample logging curve data.

[0047] Furthermore, in order to obtain the accuracy of the data, after obtaining the logging curve data of multiple key wells, it can also include: classifying the logging curve data based on the external influencing factors of the acquired logging curve data; and performing curve correction and normalization processing on the classified logging curve data.

[0048] Optionally, the external influencing factors may include the type of collection instrument corresponding to the era.

[0049] For example, for the same type of logging curve data, assuming that the types of instruments used to acquire the logging curve data include equipment A and equipment B, and the eras include era x and era y, and equipment A corresponds to era x, and equipment B corresponds to era y, then the classified logging curve data include the logging curve data acquired by equipment A in era x, and the logging curve data acquired by equipment B in era y. Curve correction and normalization processing are performed on the classified logging curve data to obtain the processed logging curve data.

[0050] It should be noted that the types of acquisition instruments here may be classified according to instrument performance, that is, the aforementioned device A and device B may be instruments of the same type but with different performances, or instruments of different types.

[0051] Of course, external influencing factors may also include other factors, such as isotope dosage, isotope interpretation depth, etc. This application does not limit the external influencing factors.

[0052] In a specific embodiment, in order to improve the recognition accuracy of the characteristic points of the logging curve change, the fluid property result graph corresponding to each sample logging curve data is obtained by the following steps:

[0053] For each sample logging curve data, one-dimensional discrete convolution is performed on the sample logging curve data to extract the extreme values ​​and inflection points of the sample logging curve data as potential stratification points of the sample logging curve data; based on the potential stratification points and the preset significance index, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, wherein each curve segment includes corresponding curve segment data; the word vectors corresponding to the plurality of curve segment data are determined, and the plurality of curve segment data are mapped to the corresponding word vectors to obtain the word vector matrix of the plurality of curve segment data; the word vector matrix is ​​convolved and pooled to obtain the fluid property result map corresponding to the sample logging curve data.

[0054] The significance index refers to the stratified values ​​and amplitude thresholds corresponding to different curve segments determined according to the reconstructed fluid property quantitative identification standard. Of course, the significance index can also be the stratified values ​​and amplitude thresholds corresponding to different curve segments divided based on empirical data.

[0055] Specifically, for each sample logging curve data (such as GR), first, a one-dimensional signal discrete convolution is performed on the sample logging curve data to obtain the first-order and second-order derivatives of the data, and then the extreme values ​​and inflection points of the curve are extracted as potential stratification points of the curve, and the curve is divided into several monotonic interval segments. Then, based on the potential stratification points and the preset significance index, the sample logging curve data is divided into multiple curve segments arranged in sequence.

[0056] Specifically, based on potential stratification points and preset significance indicators, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, which may include: based on potential stratification points, the sample logging curve data is divided into a plurality of monotonic interval segments; based on preset significance indicators and a plurality of monotonic interval segments, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined, and the potential reservoir segment division result of the sample logging curve data is obtained, so that the sample logging curve data is divided into a plurality of curve segments arranged in sequence. Figure 3 As shown, it is a schematic diagram of discrete convolution calculation curve feature points and division intervals.

[0057] Specifically, based on a preset significance index and several monotonic interval segments, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined, and the potential reservoir segment division result of the sample logging curve data is obtained, which may include: according to the stratification value and amplitude threshold of the logging curve data corresponding to the preset significance index, the half-amplitude point positions corresponding to several monotonic interval segments are screened to obtain the depth position of the stratification points of each potential stratification segment in the curve, thereby obtaining the potential reservoir segment division result of the sample logging curve data.

[0058] Specifically, according to the stratification value and amplitude threshold of the logging curve data corresponding to the preset significance index, the half-amplitude point positions corresponding to several monotonic interval segments are screened to obtain the depth positions of the stratification points of each potential stratification segment in the curve data, which may include: screening out the points whose curve value at the half-amplitude point position is greater than the stratification value, and the difference between the maximum and minimum values ​​in the curve segment is greater than the amplitude threshold, and using the screened points as the depth positions of the stratification points of each potential segment in the logging data.

[0059] Furthermore, in order to improve the accuracy of the measured depth position and obtain more accurate identification points, based on potential stratification points and preset significance indicators, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, which may include: based on potential stratification points, the sample logging curve data is divided into a plurality of monotonic interval segments; based on preset weights, preset significance indicators and a plurality of monotonic interval segments, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined, and the potential reservoir segment division result of the sample logging curve data is obtained, and the sample logging curve data is divided into a plurality of curve segments arranged in sequence. Among them, the preset weight is used to calculate the weight of each logging curve feature point in different depth ranges.

[0060] It should be noted that the method for obtaining the preset weights includes: performing sensitivity analysis on different curves in combination with different fluid properties according to the reconstructed fluid property quantitative identification standard, and calculating the weights corresponding to different curves when different fluid properties are divided.

[0061] Specifically, based on preset weights, preset significance indicators and several monotonic intervals, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined, which may include: for the stratification value and amplitude threshold of the logging curve data corresponding to the preset significance indicator, the half-amplitude point positions corresponding to several monotonic intervals are screened to obtain the depth position of the stratification points of each potential stratification segment in the curve; the depth position of the stratification point of each potential stratification segment is weighted with the corresponding weight under the depth range to obtain the corrected depth position of the stratification point of each potential stratification segment (i.e., each curve segment data), and the corrected depth position more accurately reflects the depth data of the stratification point of each potential stratification segment, so as to obtain a more accurate potential reservoir segment division result.

[0062] Specifically, according to the stratification value and amplitude threshold of the logging curve data corresponding to the preset significance index, the half-amplitude point positions corresponding to several monotonic interval segments are screened to obtain the depth positions of the stratification points of each potential stratification segment in the curve, which may include: screening out the points whose curve value at the half-amplitude point position is greater than the stratification value, and the difference between the maximum and minimum values ​​in the curve segment is greater than the amplitude threshold, and using the screened points as the depth positions of the stratification points of each potential segment in the logging data.

[0063] Then, if Figure 4 As shown in the figure, the obtained curve segment data is mapped to the corresponding word vector, and multi-layer convolution and pooling operations are performed on the word vector matrix of each segment of the curve data, and finally mapped to the feature vector of the curve segment, and then the probability vector of the curve segment belonging to each lithology category is calculated through the fully connected layer.

[0064] In a specific embodiment, training a pre-built initial network model based on sample data may include: initializing network weight parameters in the initial network model, wherein the network weight parameters may be preset arbitrary values. Inputting the obtained sample data, i.e., sample logging curve data of multiple key wells and fluid property result graphs corresponding to each sample logging curve data, into the initial network model to obtain a learned initial network model, wherein the learned initial network model includes a mapping relationship between the sample logging curve data and the fluid property result graph.

[0065] It should be noted that since the one-dimensional sequential arrangement of well logging curve data is similar to text, and the process of obtaining fluid properties based on curve data is similar to the text semantic analysis process, the TextCNN network structure used for text processing is adopted for well logging interpretation and analysis based on the convolutional neural network. The unique N-gram feature analysis function of the text convolutional neural network can obtain feature information with richer abstract levels, and greatly improve the recognition accuracy of the characteristic points of the logging curve changes. When dealing with complex lithology logging interpretation, the recognition accuracy of the learned text convolutional neural network model is higher.

[0066] Furthermore, in order to improve the accuracy of the text convolutional neural network model, after obtaining the text convolutional neural network model, it can also include: obtaining test data, the test data including: multiple test curve data and a fluid property result graph corresponding to each test curve data; inputting the test curve data into the text convolutional neural network model to obtain a fluid property estimation graph; determining the error between the fluid property result graph and the fluid property estimation graph based on the test curve data, the fluid property estimation graph and a preset loss function; and optimizing the network weight parameters in the text convolutional neural network model based on the error.

[0067] In a specific embodiment, the text convolutional neural network model is optimized using test data until the accuracy of the text convolutional neural network model meets the accuracy requirement.

[0068] Specifically, the method of obtaining the test data is the same as the method of obtaining the sample data described above, that is, obtaining the logging curve data of multiple key wells, performing secondary interpretation of the fluid properties of the multiple key wells based on the quantitative identification standard, and using the results of the secondary interpretation as the input feature labels; based on the input feature labels, the logging curve data of the key wells are determined as the test curve data. For each test curve data, a one-dimensional discrete convolution is performed on the test curve data to extract the extreme values ​​and inflection points of the test curve data as potential stratification points of the test curve data; based on the potential stratification points and the preset significance index, the test curve data is divided into a plurality of curve segments arranged in sequence; the word vectors corresponding to the plurality of curve segment data are determined, and the plurality of curve segment data are mapped to the corresponding word vectors to obtain the word vector matrix of the plurality of curve segment data; the word vector matrix is ​​convolved and pooled to obtain the fluid property result graph corresponding to the test curve data.

[0069] Optionally, the determination of the loss function may include: using the Euclidean distance as a preset loss function to calculate the error between the fluid property result map and the fluid property estimation map.

[0070] The trained text convolutional neural network model is used to identify the reservoir fluid properties of the well to be identified, and the text convolutional neural network model is iteratively updated according to the identification results, so that the text convolutional neural network model after each identification is the optimal model after the updated iteration.

[0071] The iterative update process may specifically include: revising the model interpretation conclusion according to the actual test results, and using the revised recognition results as new sample data to train the model. In one application scenario, after 737 continuous iterative trainings, the model prediction accuracy was 93%.

[0072] like Figure 5 As shown in the figure, according to the interpretation results of this method, Well A in an oil field conducted an old well test on the well section of 3710.77m-3761.25m. After pressure treatment, the daily oil production was 17.2 tons, and a high-yield industrial oil flow was obtained.

[0073] The method provided in this application can be applied in the exploration of deep layers in steep slope belts. There are many types of conglomerate rocks in the deep layers of this area. On the basis of clarifying the logging response characteristics of various lithologic reservoirs, this method is used to conduct secondary interpretation of fluid properties for key wells in the area, and the distribution range of various lithologic reservoirs on the plane is determined. The efficiency is significantly improved, the accuracy rate reaches 90%, and exploration deployment is carried out according to the interpretation results, and the newly added reserves exceed 100 million tons. Therefore, the conglomerate fluid property identification method provided in this application has a high identification accuracy and a good application effect, and is particularly suitable for areas with a high degree of exploration and development.

[0074] For ease of understanding, the complex conglomerate reservoir fluid property identification method based on text convolutional neural network and discrete convolution algorithm mentioned in this application is introduced in detail below:

[0075] Step 1: Unify the logging curve data of key wells, including natural potential, natural gamma, well diameter, deep lateral, shallow lateral, acoustic time difference, compensated neutron and density; perform curve correction and normalization on the logging curve data obtained from different eras and different acquisition instrument types. Establish learning sample data based on the processed logging curve data.

[0076] Step 2: Based on the test oil and production data, combined with rock and mineral and rock physics analysis, clarify the logging response characteristics of different lithology fluid properties, and formulate quantitative identification standards for fluid properties by lithology; conduct secondary interpretation of fluid properties for key wells based on quantitative identification standards, use the interpretation results as input feature labels, and divide the logging data samples of key wells into sample logging curve data and test curve data.

[0077] Step 3: Perform one-dimensional signal discrete convolution on the logging curve data (sample logging curve data and test curve data) in the sample data to obtain the first-order and second-order derivatives of the data, extract the extreme values ​​and inflection points of the curve as the potential stratification points of the curve, and divide the curve into several monotonic interval segments. According to parameters such as significance indicators, calculate the depth position of the stratification points of each potential stratification segment in the curve, calculate the potential reservoir segment division results of each curve, and thus divide the logging curve data into each curve segment arranged in sequence.

[0078] Step 4: Establish a text convolutional neural network model, including the N-gram model, convolution layer size, pooling layer calculation method, fully connected layer calculation, activation function, and loss function.

[0079] Step 5: Train the initial network model based on the sample data to obtain the trained text convolutional neural network model. Use the test data curve to test the accuracy of the model. If the required accuracy is met, the model enters the next step. If it does not meet the required accuracy, additional training is performed.

[0080] Step 6: Use the trained text convolutional neural network model to identify the reservoir fluid properties of the new well; and iteratively update the text convolutional neural network model based on the results.

[0081] In summary, a method for identifying conglomerate fluid properties is provided through an embodiment of the present invention. The method can be specially designed for predicting the properties of different reservoir fluids based on logging curves during oil and gas field exploration and development. The text convolutional neural network model is used to interpret and analyze the wells to be identified, and feature information with richer levels is obtained, thereby improving the efficiency and accuracy of the interpretation model in identifying the fluid properties of complex conglomerate reservoirs.

[0082] In the second aspect, based on the same inventive concept, this embodiment provides a device for identifying conglomerate fluid properties, such as Figure 6 As shown, including:

[0083] An acquisition module 401 is used to acquire the logging curve data of the well to be identified;

[0084] The identification module 402 is used to input the logging curve data into a preset text convolutional neural network model to identify the reservoir fluid properties of the well to be identified, and obtain a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

[0085] As an optional embodiment, the text convolutional neural network model is trained according to the following steps: obtaining sample data, the sample data including: sample logging curve data of multiple key wells and fluid property result graphs corresponding to each sample logging curve data; based on the sample data, training a pre-constructed initial network model to obtain a text convolutional neural network model.

[0086] As an optional embodiment, the sample logging curve data of the multiple key wells are obtained through the following steps: acquiring the logging curve data of the multiple key wells; performing secondary interpretation of the fluid properties of the multiple key wells based on quantitative identification standards, and using the results of the secondary interpretation as input feature labels, wherein the quantitative identification standards are formulated based on the logging response characteristics of the fluid properties of different lithologies; based on the input feature labels, determining the logging curve data of the key wells as sample logging curve data.

[0087] As an optional embodiment, the fluid property result graph corresponding to each sample logging curve data is obtained by the following steps: for each sample logging curve data, one-dimensional discrete convolution is performed on the sample logging curve data to extract the extreme values ​​and inflection points of the sample logging curve data as potential stratification points of the sample logging curve data; based on the potential stratification points and preset significance indicators, the sample logging curve data is divided into a plurality of curve segments arranged in sequence, wherein each curve segment includes corresponding curve segment data; word vectors corresponding to the plurality of curve segment data are determined, and the plurality of curve segment data are mapped to the corresponding word vectors to obtain a word vector matrix of the plurality of curve segment data; the word vector matrix is ​​convolved and pooled to obtain a fluid property result graph corresponding to the sample logging curve data.

[0088] As an optional embodiment, the sample logging curve data is divided into a plurality of curve segments arranged in sequence based on the potential stratification points and preset significance indicators, including: based on the potential stratification points, the sample logging curve data is divided into a plurality of monotonic interval segments; based on the preset significance indicators and a plurality of monotonic interval segments, the depth position of the stratification points of each potential stratification segment in the sample logging curve data is determined to obtain the potential reservoir segment division result of the sample logging curve data, and the sample logging curve data is divided into a plurality of curve segments arranged in sequence.

[0089] As an optional embodiment, after obtaining the text convolutional neural network model, it also includes: obtaining test data, the test data including: multiple test curve data and a fluid property result graph corresponding to each test curve data; inputting the test curve data into the text convolutional neural network model to obtain a fluid property estimation graph; determining the error between the fluid property result graph and the fluid property estimation graph based on the test curve data, the fluid property estimation graph and a preset loss function; and optimizing the network weight parameters in the text convolutional neural network model based on the error.

[0090] As an optional embodiment, after acquiring the logging curve data of multiple key wells, the method further includes: classifying the logging curve data based on external influencing factors of the acquired logging curve data; and performing curve correction and normalization processing on the classified logging curve data.

[0091] As an optional embodiment, the external influencing factors include age and type of collection instrument.

[0092] The above modules can be implemented by software codes, in which case, the above modules can be stored in the memory of the control device. The above modules can also be implemented by hardware such as integrated circuit chips.

[0093] An embodiment of the present invention provides a device for identifying the properties of gravel fluids, and its implementation principle and technical effects are the same as those of the aforementioned method embodiment. For the sake of brief description, for matters not mentioned in the device embodiment, reference may be made to the corresponding contents in the aforementioned method embodiment.

[0094] In a third aspect, based on the same inventive concept, this embodiment provides an electronic device 500, such as Figure 7 As shown, it includes: a memory 501, a processor 502 and a computer program 503 stored in the memory and executable on the processor. When the processor 501 executes the program, the steps of the method for identifying the properties of conglomerate fluids described in the first aspect are implemented.

[0095] Those skilled in the art will readily appreciate other embodiments of the invention after considering the specification and practicing the invention disclosed herein. The present invention is intended to cover any variations, uses or adaptations of the present invention that follow the general principles of the present invention and include common knowledge or customary techniques in the art that are not disclosed in this disclosure. The description and examples are to be considered exemplary only, and the true scope and spirit of the present invention are indicated by the following claims.

[0096] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0097] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A module that specifies functions in one or more boxes.

[0098] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction module, which is implemented in the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0099] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0100] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0101] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for identifying conglomerate fluid properties, characterized in that: include: Obtaining well logging curve data of the well to be identified; The logging curve data is input into a preset text convolutional neural network model to identify the reservoir fluid properties of the well to be identified, and a fluid property result map of the well to be identified is obtained, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

2. The identification method according to claim 1, characterized in that: The text convolutional neural network model is trained according to the following steps: Acquire sample data, the sample data including: sample logging curve data of a plurality of key wells and fluid property result graphs corresponding to each sample logging curve data; Based on the sample data, the pre-built initial network model is trained to obtain the text convolutional neural network model.

3. The identification method according to claim 2, characterized in that: The sample logging curve data of the plurality of key wells are obtained by the following steps: Obtain logging curve data of multiple key wells; Performing secondary interpretation of fluid properties on the multiple key wells based on quantitative identification criteria, and using the results of the secondary interpretation as input feature labels, wherein the quantitative identification criteria are formulated based on logging response characteristics of fluid properties of different lithologies; Based on the input feature label, the logging curve data of the key well is determined as sample logging curve data.

4. The identification method according to claim 2, characterized in that: The fluid property result graph corresponding to each sample logging curve data is obtained by the following steps: For each sample well logging curve data, one-dimensional discrete convolution is performed on the sample well logging curve data to extract the extreme value and inflection point of the sample well logging curve data as potential stratification points of the sample well logging curve data; Based on the potential stratification points and the preset significance index, the sample logging curve data is segmented into a plurality of curve segments arranged in sequence, wherein each curve segment includes corresponding curve segment data; Determine word vectors corresponding to the plurality of curve segment data, and map the plurality of curve segment data to the corresponding word vectors to obtain a word vector matrix of the plurality of curve segment data; The word vector matrix is ​​convolved and pooled to obtain a fluid property result graph corresponding to the sample logging curve data.

5. The identification method according to claim 4, characterized in that: The method of dividing the sample logging curve data into a plurality of curve segments arranged in sequence based on the potential stratification points and the preset significance index comprises: Based on the potential stratification points, the sample logging curve data is divided into a number of monotonic interval segments; Based on the preset significance index and the several monotonic interval segments, the depth position of the stratification point of each potential stratification segment in the sample logging curve data is determined, and the potential reservoir segment division result of the sample logging curve data is obtained, so that the sample logging curve data is divided into a plurality of curve segments arranged in sequence.

6. The identification method according to claim 2, characterized in that: After obtaining the text convolutional neural network model, the method further includes: Acquire test data, the test data comprising: a plurality of test curve data and a fluid property result graph corresponding to each test curve data; Inputting the test curve data into the text convolutional neural network model to obtain a fluid property estimation graph; Determining the error between the fluid property result graph and the fluid property estimation graph according to the test curve data, the fluid property estimation graph and a preset loss function; The network weight parameters in the text convolutional neural network model are optimized based on the error.

7. The identification method according to claim 3, characterized in that: After obtaining the logging curve data of multiple key wells, the method further includes: Classifying the well logging curve data based on the acquired external influencing factors of the well logging curve data; The classified logging curve data are corrected and normalized.

8. The identification method according to claim 7, characterized in that: The external influencing factors include age and type of collection instrument.

9. A device for identifying conglomerate fluid properties, characterized in that: include: An acquisition module, used for acquiring the logging curve data of the well to be identified; The identification module is used to input the logging curve data into a preset text convolutional neural network model to identify the reservoir fluid properties of the well to be identified, and obtain a fluid property result map of the well to be identified, wherein the fluid property result map includes probability vectors of different types of lithology in the well to be identified.

10. An electronic device, characterized in that: include: 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 method according to any one of claims 1 to 8 when executing the program.

Images