A deep shale gas production capacity evaluation method and system

Abstract

The present invention provides a method and system for evaluating the production capacity of deep shale gas, which relates to the field of deep shale technology, including obtaining multiple first information and second information corresponding to each first information; training a preset neural network model based on multiple first information, and stopping the training when the preset first loss function meets the first set condition to obtain an original production capacity prediction sub-model; obtaining an actual production capacity prediction sub-model; obtaining target information of the target deep shale; inputting the target information into the production capacity prediction model to obtain the original production capacity and actual production capacity of the target deep shale. The present invention comprehensively considers the relevant factors affecting the original production capacity and actual production capacity of deep shale gas, constructs two prediction sub-models, can better capture the complex relationship between data, accurately predict the original production capacity and actual production capacity, and help staff accurately understand the current exploitation potential of deep shale gas, so as to make more accurate exploitation decisions and maximize development benefits.

Description

Technical Field

[0001] The present invention relates to the field of deep shale technology, and in particular to a method and system for evaluating deep shale gas production capacity. Background Art

[0002] At present, shale gas development continues to advance, from shallow shale above 3500m to deep shale below 3500m, and has made great strides in implementing the goal of efficient development of unconventional natural gas resources. Conventional gas well productivity evaluation is mainly based on productivity well testing theory. Through production logging, the gas well productivity equation and open flow rate are determined, and then proportional production is carried out.

[0003] The methods for calculating open-flow rate generally use the "one-point method" and "binomial method". These two methods require the existence of multiple stable test points in the production test, and the stability time is long. However, deep shale gas production test has the following three characteristics: first, the test time period is not sufficient; second, the work system changes before the test curve stabilizes; third, stable production is difficult to achieve during the test. Obviously, the conventional gas well capacity evaluation ideas are inconsistent with the characteristics of deep shale gas production test. The above contradictions will make the current capacity evaluation method unsuitable for the development of deep shale gas. Summary of the invention

[0004] The purpose of the present invention is to provide a deep shale gas production capacity evaluation method and system to improve the above problems. In order to achieve the above purpose, the technical solution adopted by the present invention is as follows:

[0005] In a first aspect, the present application provides a method for evaluating deep shale gas production capacity, comprising:

[0006] Acquire multiple first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock formation thickness, organic matter content and thermal maturity of deep shale, the original production capacity is the shale gas production stored in the deep shale gas, and the second information includes rock formation porosity, rock formation permeability, mineral composition, formation pressure and actual production capacity of the deep shale, and the actual production capacity is the shale gas production actually mined;

[0007] Training a preset neural network model based on the plurality of first information, and stopping the training when the preset first loss function satisfies a first set condition, to obtain an original capacity prediction sub-model;

[0008] The neural network model is trained based on the multiple original capacity information and the multiple second information, and when the preset second loss function satisfies the second set condition, the training is stopped to obtain the actual capacity prediction sub-model;

[0009] Acquiring target information of target deep shale, the target information including rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure;

[0010] The target information is input into a production capacity prediction model to obtain the original production capacity and actual production capacity of the target deep shale. The production capacity prediction model is composed of the original production capacity prediction sub-model and the actual production capacity prediction sub-model.

[0011] In the second aspect, the present application also provides a deep shale gas production capacity evaluation system, including:

[0012] a first acquisition unit, configured to acquire a plurality of first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock formation thickness, organic matter content and thermal maturity of the deep shale, the original production capacity is the production of shale gas stored inside the deep shale gas, and the second information includes rock formation porosity, rock formation permeability, mineral composition, formation pressure and actual production capacity of the deep shale, the actual production capacity is the production of shale gas actually mined;

[0013] A first training unit is used to train a preset neural network model based on multiple first information, and when a preset first loss function satisfies a first set condition, stop the training to obtain an original capacity prediction sub-model;

[0014] A second training unit is used to train the neural network model based on the multiple original capacity information and the multiple second information, and when the preset second loss function meets the second set condition, stop the training to obtain the actual capacity prediction sub-model;

[0015] A second acquisition unit is used to acquire target information of target deep shale, wherein the target information includes rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure;

[0016] The first input unit is used to input the target information into a capacity prediction model to obtain the original capacity and actual capacity of the target deep shale. The capacity prediction model is composed of the original capacity prediction sub-model and the actual capacity prediction sub-model.

[0017] The beneficial effects of the present invention are:

[0018] The present invention comprehensively considers the relevant factors affecting the original production capacity and actual production capacity of deep shale gas, and constructs two prediction sub-models, which can better capture the complex relationship between the data, accurately predict the original production capacity and actual production capacity, and help staff accurately understand the current deep shale gas extraction potential, so as to make more accurate extraction decisions and maximize development benefits.

[0019] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or be understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0021] Figure 1 A schematic diagram of a process for evaluating deep shale gas production capacity according to an embodiment of the present invention;

[0022] Figure 2 Schematic diagram of the structure of the deep shale gas production capacity evaluation system described in an embodiment of the present invention.

[0023] Markings in the figure:

[0024] 10. First acquisition unit; 20. First training unit; 30. Second training unit; 40. Second acquisition unit; 50. First input unit. DETAILED DESCRIPTION

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0026] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance. Example

[0027] This embodiment provides a method for evaluating deep shale gas production capacity.

[0028] See also Figure 1 , the figure shows that the method includes step S10, step S20, step S30, step S40 and step S50.

[0029] Step S10. Acquire multiple first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock formation thickness, organic matter content and thermal maturity of deep shale, the original production capacity is the shale gas production stored in the deep shale gas, and the second information includes rock formation porosity, rock formation permeability, mineral composition, formation pressure and actual production capacity of the deep shale, and the actual production capacity is the shale gas production actually mined;

[0030] Specifically, this application considers that the initial production of deep shale gas is related to the rock thickness, organic matter content and thermal maturity of deep shale, and the actual production of deep shale gas is related to rock porosity, rock permeability, mineral composition, formation pressure and initial production capacity, so it is possible to establish prediction models between the above-mentioned related parameter information and initial production capacity, and between the above-mentioned related parameter information and actual production capacity. Among them, the thickness of the rock layer can be collected and determined by seismic exploration methods; the organic matter content can be measured by core analysis and geochemical methods; thermal maturity can be evaluated by rock pyrolysis analysis, vitrinite reflectivity measurement and other methods; rock formation porosity can be measured by core analysis, nuclear magnetic resonance logging and other methods; rock formation permeability can be evaluated by laboratory permeability test or pressure pulse attenuation measurement; mineral composition can be determined by X-ray diffraction analysis or electron microscope analysis; formation pressure can be estimated by downhole pressure test.

[0031] Step S20. Training a preset neural network model based on the plurality of first information, and when the preset first loss function satisfies the first set condition, stopping the training to obtain an original capacity prediction sub-model;

[0032] Specifically, the constructed neural network model includes a processing layer, a hidden layer and an output layer. The processing layer is composed of an input layer and an optimization layer, wherein the processing layer is used to process various relevant parameter information and convert it into a data type that can be processed by the neural network model to generate input data for the hidden layer. At the same time, the optimization layer nested in the processing layer optimizes the initial weights and thresholds generated by the neural network model to obtain the initial optimal weights and thresholds of the model; the hidden layer is used to process the output data of the processing layer so as to abstract the characteristics of the input data at multiple levels and linearly divide the input nonlinear data so as to perform weighted correction calculation on the above input data according to the weights and thresholds of this layer, thereby obtaining the input data of the output layer; the output layer is used to perform weighted correction calculation on the output data of the above hidden layer to obtain the original capacity prediction value of the neural network model, use the objective function to evaluate the loss value of the prediction value, calculate the descending gradient of the loss value layer by layer according to each hierarchical structure, perform reverse transmission of the neural network, optimize the neural network prediction model, and finally achieve accurate prediction of the original capacity.

[0033] Specifically, step S20 specifically includes step S21, step S22, step S23, step S24 and step S25:

[0034] Step S21. Input relevant parameter information in the plurality of first information into the processing layer for classification and fusion to obtain a plurality of overall feature data;

[0035] Specifically, considering that the amount of relevant parameter information used to train the model is large, and the relevant parameter information also carries a lot of noise and redundant information, the trained model may be overfitted, that is, the model is too complex and has too strong reliance on the details of the relevant parameter information used for training, which reduces the generalization ability of new information; therefore, in this application, a large amount of relevant parameter information is grouped and processed, and multiple relevant parameter information features in the same group are merged into an overall feature, and then the model is trained for learning, which can reduce the adverse effects of too much relevant parameter information while maintaining model performance.

[0036] Specifically, step S21 specifically includes step S211, step S212, step S213, step S214 and step S215:

[0037] Step S211. Classify the relevant parameter information in the plurality of first information to obtain a plurality of target classification clusters, wherein the number of relevant parameter information in each target classification cluster does not exceed a first set threshold;

[0038] Specifically, by grouping multiple related parameter information, complex related parameter information can be decomposed into multiple simpler and structured subsets to reduce the complexity of information. Usually, the first set threshold is set to 10 to ensure that the correlation of related parameters in each target classification cluster is high, and the first set threshold can be flexibly adjusted to better adapt to different data characteristics, ensure the optimization of grouping processing, and help improve the effectiveness and practicality of grouping.

[0039] Specifically, step S211 specifically includes step S2111, step S2112, step S2113, step S2114, step S2115, step S2116, step S2117, step S2118, step S2119, step S21110, step S21111 and step S21112:

[0040] Step S2111. Normalize multiple related parameter information to obtain multiple parameter information;

[0041] Specifically, considering the dimensional differences between multiple related parameter information, the normalization operation can ensure that each parameter is compared and analyzed on the same scale. The normalization calculation formula is:

[0042] ;

[0043] in, is the normalized parameter information; For relevant parameter information; is the mean of the relevant parameter information; is the standard deviation of the relevant parameter information.

[0044] Step S2112. Randomly determine a plurality of initial center information from a plurality of parameter information;

[0045] Step S2113. First determination operation: determining a plurality of related information from a plurality of parameter information, wherein the related information is the remaining information in the plurality of parameter information except the initial central information;

[0046] Specifically, the initial center information is randomly selected and subsequent iterations are performed, which effectively avoids the local optimal solution problem caused by poor initial selection.

[0047] Step S2114. First calculation operation: Calculate the difference between the relevant information and the initial center information to obtain multiple differences, including the difference in rock layer thickness, the difference in organic matter content and the difference in thermal maturity;

[0048] Step S2115. Second calculation operation: calculate the Pearson correlation coefficient between the relevant information and the initial center information to obtain the correlation coefficient value;

[0049] Specifically, by calculating the Pearson correlation coefficient, the linear correlation between the relevant information and the initial central information can be analyzed, so as to better understand and utilize the inherent relationship in the information, which helps to improve the accuracy of classification.

[0050] Step S2116. The third calculation operation: Calculate the ratios of the plurality of difference values ​​to the second set threshold value to obtain a plurality of ratios;

[0051] Step S2117. A fourth calculation operation: calculating the square of the sine function value of each ratio, and performing a sum calculation to obtain a first value;

[0052] Step S2118. Fifth calculation operation: Calculate the square root of the first value to obtain a second value;

[0053] Step S2119. Sixth calculation operation: calculate the product of the second value and the correlation coefficient value to obtain a distance value;

[0054] Step S21110. Second determination operation: determining the initial center information corresponding to the relevant information based on the distance value, and obtaining a plurality of initial classification clusters;

[0055] Step S21111. The third determination operation: determining a plurality of updated initial center information from a plurality of initial classification clusters;

[0056] Step S21112. Repeat the first determination operation, the first calculation operation, the second calculation operation, the third calculation operation, the fourth calculation operation, the fifth calculation operation, the sixth calculation operation, the second determination operation and the third determination operation until the Pearson correlation coefficient between the updated initial center information and the initial center information before the update is greater than the third set threshold, and take the corresponding initial classification cluster as the target classification cluster;

[0057] Specifically, the distance value calculation formula is:

[0058] ;

[0059] in, is the distance value; The initial center information and Pearson correlation coefficient of relevant information; The rock layer thickness is the initial center information; is the organic matter content of the initial central information; is the thermal maturity of the initial central information; For the The thickness of the rock layer with relevant information; For the The organic matter content of each relevant information; For the The thermal maturity of relevant information;

[0060] The second threshold is set to 2. Through the above iterative operation, the initial center information is gradually optimized to ensure that the classification clusters after each iteration are moving closer to the optimal result, until the Pearson correlation coefficient of the new and old center information is greater than the set threshold, ensuring that the final classification cluster is the most stable and reasonable. Usually, the third threshold is set to 0.85, which can ensure that there is a strong correlation between the updated initial center information and the initial center information before the update.

[0061] Step S212: Encode the relevant parameter information in each target classification cluster to obtain target feature data and multiple feature data, where the target feature data represents the central data in the target classification cluster;

[0062] Specifically, through encoding operations, complex related parameter information can be compressed into concise target feature data and feature data, which not only reduces the burden of data storage, but also simplifies the subsequent calculation process, which helps to improve the efficiency of model training.

[0063] Step S213. Calculate the correlation degree value between the target feature data and the multiple feature data in each target classification cluster based on the grey correlation analysis algorithm, and use the target feature data and the multiple feature data as a node, and use the correlation degree values ​​between the target feature data and the multiple feature data as edges between the nodes, and the length of the edge is related to the correlation degree value;

[0064] Specifically, the grey correlation analysis algorithm can accurately calculate the degree of correlation between target feature data and feature data, improve the accuracy of data analysis, and enhance the reliability of classification and clustering results.

[0065] Step S214: updating the plurality of feature data based on the association degree value and the target feature data to obtain the plurality of updated feature data;

[0066] Step S215. Fusing the target feature data in each target classification cluster with the updated plurality of feature data to obtain a plurality of overall feature data;

[0067] Specifically, update the feature data:

[0068] ;

[0069] in, For the Updated feature data; For the Characteristic data; For the The correlation degree value between the feature data and the target feature data; is the target feature data; is the activation function.

[0070] The updated multiple feature data can better emphasize similar features, that is, the updated data will highlight the common features between multiple feature data in the classification cluster, making the similarity more significant. Therefore, in the feature fusion process, the model can better capture the essential characteristics of the data and reduce dependence on specific samples.

[0071] Step S22. Inputting a plurality of overall feature data into a hidden layer to obtain a plurality of feature vector data;

[0072] Specifically, multiple overall feature data are input into the hidden layer, and through multiple layers of nonlinear transformation, the hidden layer can automatically extract more useful feature vectors.

[0073] Step S23. Input multiple feature vector data into the output layer to obtain multiple original capacity prediction values;

[0074] Specifically, the feature vector extracted by the hidden layer usually has a strong recognition. After inputting it into the output layer, it can more accurately predict the original capacity value, which can improve the accuracy and robustness of the prediction model.

[0075] Step S24. Calculate the loss function based on the original capacity prediction value and the corresponding original capacity. When the loss function does not meet the first set condition, adjust the weight threshold vector combination to obtain an adjusted neural network model. The weight threshold vector combination includes the initial weight vectors between neurons in each layer of the neural network model and the initial threshold matrix vector of each neuron.

[0076] Specifically, step S24 specifically includes step S241, step S242, step S243, step S244, step S245, step S246, step S247, step S248 and step S249:

[0077] Step S241. Input the weight threshold vector combination into the optimization layer for encoding processing to obtain an encoding set, which includes multiple encoding individual arrays;

[0078] Step S242: input relevant parameter information into the processing layer for training, and calculate the fitness function corresponding to different encoding individual arrays in the encoding set to obtain multiple fitness function values;

[0079] Specifically, the network structure of the optimization layer includes a coding layer, a fitness layer, an exchange layer and a mutation layer connected in sequence. The weight threshold vector combination is input into the neural network model optimization layer for encoding to generate initial data. At the same time, the training set is input into the neural network model optimization layer, and the neural network model error is trained by calculation and operation with the weight threshold vector combination. The neural network model error is set as a fitness value function to determine whether the optimization needs to be ended.

[0080] Step S243. Sorting operation: sorting the plurality of encoding individual arrays based on the size relationship of the fitness function values, and selecting the plurality of encoding individual arrays located in the front part from the sorting queue as the first encoding individual array;

[0081] Step S244. Interaction operation: performing segment interaction between the plurality of first encoding individual arrays to generate a plurality of second encoding individual arrays, wherein the second encoding individual arrays are different from the first encoding individual arrays;

[0082] Step S245. Combination operation: combining the first coding individual and the second coding individual to form an updated coding set;

[0083] Step S246. Training operation: inputting relevant parameter information into the processing layer for training, and calculating the fitness function corresponding to different encoding individual arrays in the updated encoding set to obtain multiple fitness function values;

[0084] Step S247. Repeat the sorting operation layer, the interactive operation, the combination operation and the training operation until all fitness function values ​​are greater than the fourth set threshold, and take the encoding individual array corresponding to the minimum fitness function value as the target encoding individual array;

[0085] Specifically, according to the fitness function value of the coding individual array, the coding set is subjected to fragment interaction to generate an updated coding set, and the fitness calculation is performed again on the updated coding set. Each time, the coding individual arrays that meet the fitness function value are retained in the first 15% to ensure the stability of the optimized structure. The remaining data continue to be iteratively processed by fragment interaction operations to obtain the final optimized coding set, and the coding individual array with the smallest fitness value is selected from the optimized coding set. It is considered that the coding individual array corresponding to the minimum fitness is the current optimal coding individual array, which can ensure that the constructed model has good robustness. The fourth set threshold is usually set to 0.65, and the fourth set threshold can be adjusted according to the performance required by the constructed model, and there is no special restriction here.

[0086] Step S248. Perform inverse encoding processing on the target encoding individual array to obtain a target weight threshold vector combination;

[0087] Step S249. Adjust the neural network model based on the target weight threshold vector combination to obtain an adjusted neural network model;

[0088] Specifically, the encoded individual array corresponding to the minimum fitness function value is decoded and mapped from binary to decimal, so as to obtain the initial weight vectors between neurons in each layer and the initial threshold matrix vector of each neuron, and obtain the adjusted neural network model.

[0089] Step S25. When the loss function satisfies the first set condition, the corresponding neural network model is used as the original capacity prediction sub-model;

[0090] Specifically, by using a neural network model that meets the set conditions as the original capacity prediction sub-model, it can be ensured that the model reaches the optimal state in terms of performance, stability, efficiency, etc., and can provide more reliable original capacity prediction results.

[0091] Step S30. Training the neural network model based on the multiple original capacity information and the multiple second information, and when the preset second loss function meets the second set condition, stopping the training to obtain the actual capacity prediction sub-model;

[0092] Specifically, the training process of the actual capacity prediction sub-model is roughly the same as the training process of the original capacity prediction sub-model, and will not be repeated here.

[0093] Step S40. Obtain target information of target deep shale, the target information including rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure;

[0094] Step S50. Input the target information into the capacity prediction model to obtain the original capacity and actual capacity of the target deep shale. The capacity prediction model consists of an original capacity prediction sub-model and an actual capacity prediction sub-model.

[0095] Specifically, by inputting the target information of the target deep shale into the trained capacity prediction model, relatively accurate original capacity prediction values ​​and actual capacity prediction values ​​can be obtained.

[0096] Considering that the capacity prediction model only gives two capacity values, the original capacity and the actual capacity, it cannot well characterize whether the target deep shale is worth mining. The important factor of mining cost also needs to be considered.

[0097] Specifically, the present application further includes step S60, step S70, step S80, step S90, step S100 and step S110:

[0098] Step S60. Calculate the mining cost based on the target information of the target deep shale;

[0099] Step S70. Calculate a first weight based on the original capacity and the actual capacity;

[0100] Specifically, the first weight is used to characterize the impact of initial production capacity on the evaluation index, reflecting the importance of initial production capacity in the overall mining evaluation. The first weight calculation formula is:

[0101] ;

[0102] in, is the original capacity; is the actual production capacity; is the first weight; is the first adjustment factor.

[0103] Step S80. Calculate the second weight based on the actual production capacity and mining cost;

[0104] Specifically, the second weight is used to characterize the relationship between actual production capacity and mining cost, reflecting the importance of mining cost in the overall mining assessment. The second weight calculation formula is:

[0105] ;

[0106] in, For mining costs; is the actual production capacity; is the second weight; is the second adjustment factor.

[0107] The unit of mining costs is millions.

[0108] Step S90. Calculate the product of the original capacity and the first weight to obtain a first product;

[0109] Step S100. Calculate the product of the mining cost and the second weight to obtain a second product;

[0110] Step S110. Calculate the sum of the first product and the second product to obtain an evaluation index, which is used to characterize the necessity of mining the target deep shale;

[0111] Specifically, the evaluation index calculation formula is:

[0112] ;

[0113] in, is the evaluation index; is the original capacity; is the first weight; is the second weight; For mining costs.

[0114] The numerical value of the evaluation index can be used as an important indicator for determining whether to mine the target deep shale. Normally, when the numerical value is greater than a set threshold, the target deep shale can be mined. The threshold can be modified for different scenarios and mining requirements, and no special restrictions are imposed here.

[0115] Embodiment 2:

[0116] like Figure 2 As shown, this embodiment provides a deep shale gas production capacity evaluation system, the system comprising:

[0117] A first acquisition unit 10 is used to acquire a plurality of first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock layer thickness, organic matter content and thermal maturity of deep shale, the original production capacity is the shale gas production stored inside the deep shale gas, and the second information includes rock layer porosity, rock layer permeability, mineral composition, formation pressure and actual production capacity of the deep shale, and the actual production capacity is the shale gas production actually mined;

[0118] A first training unit 20 is used to train a preset neural network model based on a plurality of first information, and when a preset first loss function satisfies a first set condition, stop the training to obtain an original capacity prediction sub-model;

[0119] A second training unit 30 is used to train the neural network model based on the multiple original capacity information and the multiple second information, and when the preset second loss function meets the second set condition, stop the training to obtain the actual capacity prediction sub-model;

[0120] A second acquisition unit 40 is used to acquire target information of target deep shale, the target information including rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure;

[0121] The first input unit 50 is used to input the target information into the capacity prediction model to obtain the original capacity and actual capacity of the target deep shale. The capacity prediction model consists of an original capacity prediction sub-model and an actual capacity prediction sub-model.

[0122] In a specific implementation disclosed in the present application, the first training unit 20 includes:

[0123] A second input unit is used to input relevant parameter information in the plurality of first information into the processing layer for classification and fusion to obtain a plurality of overall feature data;

[0124] A third input unit is used to input a plurality of overall feature data into a hidden layer to obtain a plurality of feature vector data;

[0125] A fourth input unit is used to input a plurality of feature vector data into the output layer to obtain a plurality of original capacity prediction values;

[0126] A first calculation unit is used to calculate a loss function based on the original capacity prediction value and the corresponding original capacity. When the loss function does not meet the first set condition, the weight threshold vector combination is adjusted to obtain an adjusted neural network model, where the weight threshold vector combination includes an initial weight vector between neurons in each layer of the neural network model and an initial threshold matrix vector of each neuron;

[0127] The first is used as a unit, and when the loss function meets the first set condition, the corresponding neural network model is used as the original capacity prediction sub-model.

[0128] In a specific implementation disclosed in the present application, the second input unit includes:

[0129] A classification unit, used for classifying the relevant parameter information in the plurality of first information to obtain a plurality of target classification clusters, wherein the number of relevant parameter information in each target classification cluster does not exceed a first set threshold, wherein the first set threshold is set to 10;

[0130] An encoding unit, used for encoding relevant parameter information in each target classification cluster to obtain target feature data and multiple feature data, wherein the target feature data represents the central data in the target classification cluster;

[0131] The second calculation unit is used to calculate the correlation degree value between the target feature data and the multiple feature data in each target classification cluster based on the grey correlation analysis algorithm, and take the target feature data and the multiple feature data as a node, and take the correlation degree values ​​between the target feature data and the multiple feature data as edges between the nodes, and the length of the edge is related to the correlation degree value;

[0132] An updating unit, used for updating the plurality of feature data based on the association degree value and the target feature data to obtain the plurality of updated feature data;

[0133] The fusion unit is used to fuse the target feature data in each target classification cluster with the updated multiple feature data to obtain multiple overall feature data.

[0134] In a specific implementation disclosed in the present application, the classification process includes:

[0135] A normalization unit, used for normalizing a plurality of related parameter information to obtain a plurality of parameter information;

[0136] A first determining unit, used for randomly determining a plurality of initial center information from a plurality of parameter information;

[0137] A second determining unit is used for a first determining operation: determining a plurality of related information from the plurality of parameter information, wherein the related information is the remaining information in the plurality of parameter information except the initial central information;

[0138] A third calculation unit is used for a first calculation operation: calculating the difference between the relevant information and the initial central information to obtain a plurality of differences, the differences including the difference in rock layer thickness, the difference in organic matter content and the difference in thermal maturity;

[0139] A fourth calculation unit is used for a second calculation operation: calculating a Pearson correlation coefficient between the relevant information and the initial center information to obtain a correlation coefficient value;

[0140] a fifth calculation unit, configured for a third calculation operation: calculating ratios of a plurality of difference values ​​to a second set threshold value to obtain a plurality of ratios, wherein the second set threshold value is 2;

[0141] a sixth calculation unit, configured for a fourth calculation operation: calculating the square of the sine function value of each ratio, and performing a sum calculation to obtain a first value;

[0142] a seventh calculation unit, configured for a fifth calculation operation: calculating a square root of the first value to obtain a second value;

[0143] an eighth calculation unit, configured for a sixth calculation operation: calculating a product of the second value and the correlation coefficient value to obtain a distance value;

[0144] A third determining unit is used for a second determining operation: determining initial center information corresponding to the relevant information based on the distance value to obtain a plurality of initial classification clusters;

[0145] A fourth determining unit, configured for a third determining operation: determining a plurality of updated initial center information from the plurality of initial classification clusters;

[0146] The first repetition unit is used to repeat the first determination operation, the first calculation operation, the second calculation operation, the third calculation operation, the fourth calculation operation, the fifth calculation operation, the sixth calculation operation, the second determination operation and the third determination operation until the Pearson correlation coefficient between the updated initial center information and the initial center information before the update is greater than a third set threshold, and the corresponding initial classification cluster is used as the target classification cluster, wherein the third set threshold is set to 0.85.

[0147] In a specific implementation disclosed in the present application, the first computing unit includes:

[0148] A fifth input unit is used to input the weight threshold vector combination into the optimization layer for encoding processing to obtain an encoding set, which includes multiple encoding individual arrays;

[0149] The sixth input unit is used to input relevant parameter information into the processing layer for training, and calculate the fitness function corresponding to different encoding individual arrays in the encoding set to obtain multiple fitness function values;

[0150] A sorting unit, used for sorting operation: sorting a plurality of encoding individual arrays based on the size relationship of fitness function values, and selecting a plurality of encoding individual arrays located in the front part from the sorting queue as the first encoding individual array;

[0151] An interaction unit, used for interaction operation: performing segment interaction between the plurality of first encoding individual arrays to generate a plurality of second encoding individual arrays, wherein the second encoding individual arrays are different from the first encoding individual arrays;

[0152] A combining unit, used for combining operation: combining the first coding individual with the second coding individual to form an updated coding set;

[0153] The third training unit is used for training operation: inputting relevant parameter information into the processing layer for training, and calculating the fitness function corresponding to different encoding individual arrays in the updated encoding set to obtain multiple fitness function values;

[0154] A second repetition unit is used to repeat the sorting operation layer, the interaction operation, the combination operation and the training operation until all fitness function values ​​are greater than a fourth set threshold, and the encoding individual array corresponding to the minimum fitness function value is used as the target encoding individual array, wherein the fourth set threshold is set to 0.65;

[0155] A reverse coding unit, used for reverse coding the target coding individual array to obtain a target weight threshold vector combination;

[0156] The adjustment unit is used to adjust the neural network model based on the target weight threshold vector combination to obtain an adjusted neural network model.

[0157] In a specific implementation disclosed in the present application, the system further includes:

[0158] A ninth calculation unit, for calculating the mining cost based on the target information of the target deep shale;

[0159] a tenth calculation unit, configured to calculate a first weight based on the original capacity and the actual capacity;

[0160] an eleventh calculation unit, configured to calculate a second weight based on actual production capacity and mining cost;

[0161] A twelfth calculation unit, used to calculate the product of the original capacity and the first weight to obtain a first product;

[0162] A thirteenth calculation unit, used for calculating the product of the mining cost and the second weight to obtain a second product;

[0163] A fourteenth calculation unit is used to calculate the sum of the first product and the second product to obtain an evaluation index, where the evaluation index is used to characterize the necessity of mining the target deep shale.

[0164] In a specific implementation disclosed in the present application, the updating unit includes:

[0165] ;

[0166] in, For the Updated feature data; For the Characteristic data; For the The correlation degree value between the feature data and the target feature data; is the target feature data; is the activation function.

[0167] In a specific implementation disclosed in the present application, the normalization unit includes:

[0168] ;

[0169] in, is the normalized parameter information; For relevant parameter information; is the mean of the relevant parameter information; is the standard deviation of the relevant parameter information.

[0170] In a specific implementation manner disclosed in the present application, the tenth computing unit includes:

[0171] ;

[0172] in, is the original capacity; is the actual production capacity; is the first weight; is the first adjustment factor.

[0173] It should be noted that, regarding the system in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0174] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0175] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for evaluating deep shale gas productivity, characterized in that: include: Acquire multiple first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock formation thickness, organic matter content and thermal maturity of deep shale, the original production capacity is the shale gas production stored in the deep shale gas, and the second information includes rock formation porosity, rock formation permeability, mineral composition, formation pressure and actual production capacity of the deep shale, and the actual production capacity is the shale gas production actually mined; Training a preset neural network model based on the plurality of first information, and stopping the training when the preset first loss function satisfies a first set condition, to obtain an original capacity prediction sub-model; The neural network model is trained based on the multiple original capacity information and the multiple second information, and when the preset second loss function satisfies the second set condition, the training is stopped to obtain the actual capacity prediction sub-model; Acquiring target information of target deep shale, the target information including rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure; Inputting the target information into a capacity prediction model to obtain the original capacity and actual capacity of the target deep shale, wherein the capacity prediction model is composed of the original capacity prediction sub-model and the actual capacity prediction sub-model; The target information is input into the capacity prediction model to obtain the original capacity and actual capacity of the target deep shale, and then includes: Calculating the mining cost based on the target information of the target deep shale; Calculating a first weight based on the original capacity and the actual capacity; Calculating a second weight based on the actual production capacity and the mining cost; Calculating the product of the original capacity and the first weight to obtain a first product; Calculating the product of the mining cost and the second weight to obtain a second product; Calculating the sum of the first product and the second product to obtain an evaluation index, wherein the evaluation index is used to characterize the necessity of mining the target deep shale; Among them, the first weight calculation formula is: Among them, C1 is the original capacity; C2 is the actual capacity; λ1 is the first weight; k1 is the first adjustment coefficient; The second weight calculation formula is: Among them, S is the mining cost; C2 is the actual production capacity; λ2 is the second weight; k2 is the second adjustment coefficient.

2. The method for evaluating deep shale gas productivity according to claim 1, characterized in that , based on multiple first information, the preset neural network model is trained, and when the preset first loss function meets the first set condition, the training is stopped to obtain the original capacity prediction sub-model, the preset neural network model includes a processing layer, a hidden layer and an output layer, the processing layer includes an input layer and an optimization layer, including: Inputting relevant parameter information in the plurality of first information into the processing layer for classification and fusion to obtain a plurality of overall feature data; Inputting a plurality of overall feature data into the hidden layer to obtain a plurality of feature vector data; Inputting a plurality of feature vector data into the output layer to obtain a plurality of original capacity prediction values; Calculating a loss function based on the original capacity prediction value and the corresponding original capacity, and when the loss function does not meet the first set condition, adjusting the weight threshold vector combination to obtain the adjusted neural network model, wherein the weight threshold vector combination includes an initial weight vector between neurons in each layer of the neural network model and an initial threshold matrix vector of each neuron; When the loss function satisfies the first set condition, the corresponding neural network model serves as the original capacity prediction sub-model.

3. The method for evaluating deep shale gas productivity according to claim 2, characterized in that , input relevant parameter information in the multiple first information into the processing layer for classification and fusion, and obtain multiple overall feature data, including: Classify the relevant parameter information in the multiple first information to obtain multiple target classification clusters, where the number of relevant parameter information in each target classification cluster does not exceed a first set threshold, wherein the first set threshold is set to 10; Encoding relevant parameter information in each target classification cluster to obtain target feature data and multiple feature data, wherein the target feature data represents the central data in the target classification cluster; Calculate the association degree value between the target feature data and multiple feature data in each target classification cluster based on the grey correlation analysis algorithm, and take the target feature data and multiple feature data as a node, and take the association degree values ​​between the target feature data and multiple feature data as edges between the nodes, and the length of the edge is related to the association degree value; Based on the association degree value and the target feature data, a plurality of feature data are updated to obtain a plurality of updated feature data; The target feature data in each target classification cluster and the updated multiple feature data are fused to obtain multiple overall feature data.

4. The method for evaluating deep shale gas productivity according to claim 3, characterized in that , input relevant parameter information in the multiple first information into the processing layer for classification processing, and obtain multiple target classification clusters, including: Normalizing multiple related parameter information to obtain multiple parameter information; Randomly determine multiple initial center information from multiple parameter information; A first determining operation: determining a plurality of related information from the plurality of parameter information, wherein the related information is the remaining information in the plurality of parameter information except the initial central information; The first calculation operation is to calculate the difference between the relevant information and the initial central information to obtain a plurality of differences, wherein the differences include a rock layer thickness difference, an organic matter content difference and a thermal maturity difference; Second calculation operation: calculating the Pearson correlation coefficient between the relevant information and the initial center information to obtain a correlation coefficient value; A third calculation operation: calculating ratios of a plurality of difference values ​​to a second set threshold value to obtain a plurality of ratios, wherein the second set threshold value is 2; A fourth calculation operation: calculating the square of the sine function value of each ratio, and performing a sum calculation to obtain a first value; Fifth calculation operation: calculating the square root of the first value to obtain a second value; A sixth calculation operation: calculating the product of the second value and the correlation coefficient value to obtain a distance value; A second determination operation: determining initial center information corresponding to the relevant information based on the distance value to obtain a plurality of initial classification clusters; A third determining operation: determining a plurality of updated initial center information from the plurality of initial classification clusters; Repeat the first determination operation, the first calculation operation, the second calculation operation, the third calculation operation, the fourth calculation operation, the fifth calculation operation, the sixth calculation operation, the second determination operation and the third determination operation until the Pearson correlation coefficient between the updated initial center information and the initial center information before the update is greater than a third set threshold, and take the corresponding initial classification cluster as the target classification cluster, wherein the third set threshold is set to 0.

85.

5. The method for evaluating deep shale gas productivity according to claim 3, characterized in that , based on the association degree value and the target feature data, multiple feature data are updated to obtain multiple updated feature data, including: s i '=η(s i *W im *s m ) Among them, s i ' is the i-th updated feature data; s i is the i-th feature data; W im is the correlation degree between the ith feature data and the target feature data; s m is the target feature data; η is the activation function.

6. The method for evaluating deep shale gas productivity according to claim 4, characterized in that , multiple related parameter information is normalized to obtain multiple parameter information, and the normalization calculation formula is: Among them, x' is the normalized parameter information; x is the relevant parameter information; μ is the mean of the relevant parameter information; σ is the standard deviation of the relevant parameter information.

7. A deep shale gas production capacity evaluation system, characterized in that: include: a first acquisition unit, configured to acquire a plurality of first information and second information corresponding to each first information, wherein the first information includes relevant parameter information and original production capacity of deep shale, the relevant parameter information includes rock formation thickness, organic matter content and thermal maturity of the deep shale, the original production capacity is the production of shale gas stored inside the deep shale gas, and the second information includes rock formation porosity, rock formation permeability, mineral composition, formation pressure and actual production capacity of the deep shale, the actual production capacity is the production of shale gas actually mined; A first training unit is used to train a preset neural network model based on multiple first information, and when a preset first loss function satisfies a first set condition, stop the training to obtain an original capacity prediction sub-model; A second training unit is used to train the neural network model based on the multiple original capacity information and the multiple second information, and when the preset second loss function meets the second set condition, stop the training to obtain the actual capacity prediction sub-model; A second acquisition unit is used to acquire target information of target deep shale, wherein the target information includes rock formation thickness, organic matter content, thermal maturity, rock formation porosity, rock formation permeability, mineral composition and formation pressure; A first input unit is used to input the target information into a capacity prediction model to obtain the original capacity and actual capacity of the target deep shale, wherein the capacity prediction model is composed of the original capacity prediction sub-model and the actual capacity prediction sub-model; Wherein, the system further comprises: A ninth calculation unit, comprising calculating a mining cost based on target information of the target deep shale; A tenth calculation unit, comprising calculating a first weight based on the original capacity and the actual capacity; an eleventh calculation unit, comprising calculating a second weight based on the actual production capacity and the mining cost; A twelfth calculation unit, comprising calculating the product of the original capacity and the first weight to obtain a first product; A thirteenth calculation unit, comprising calculating the product of the mining cost and the second weight to obtain a second product; A fourteenth calculation unit, comprising calculating the sum of the first product and the second product to obtain an evaluation index, wherein the evaluation index is used to characterize the necessity of exploitation of the target deep shale; Among them, the first weight calculation formula is: Among them, C1 is the original capacity; C2 is the actual capacity; λ1 is the first weight; k1 is the first adjustment coefficient; The second weight calculation formula is: Among them, S is the mining cost; C2 is the actual production capacity; λ2 is the second weight; k2 is the second adjustment coefficient.

8. The deep shale gas productivity evaluation system according to claim 7, characterized in that: The preset neural network model includes a processing layer, a hidden layer and an output layer, the processing layer includes an input layer and an optimization layer, and the first training unit includes: A second input unit is used to input relevant parameter information in the plurality of first information into the processing layer for classification and fusion to obtain a plurality of overall feature data; A third input unit is used to input a plurality of overall feature data into the hidden layer to obtain a plurality of feature vector data; A fourth input unit, used for inputting a plurality of feature vector data into the output layer to obtain a plurality of original capacity prediction values; A first calculation unit is used to calculate a loss function based on the original capacity prediction value and the corresponding original capacity, and when the loss function does not meet the first set condition, adjust the weight threshold vector combination to obtain the adjusted neural network model, wherein the weight threshold vector combination includes an initial weight vector between neurons in each layer of the neural network model and an initial threshold matrix vector of each neuron; The first is a unit, which is used to use the corresponding neural network model as the original capacity prediction sub-model when the loss function meets the first set condition.

9. The deep shale gas productivity evaluation system according to claim 8, characterized in that: The second input unit comprises: A classification unit, configured to classify the relevant parameter information in the plurality of first information to obtain a plurality of target classification clusters, wherein the number of relevant parameter information in each target classification cluster does not exceed a first set threshold, wherein the first set threshold is set to 10; An encoding unit, used for encoding relevant parameter information in each target classification cluster to obtain target feature data and multiple feature data, wherein the target feature data represents the central data in the target classification cluster; A second calculation unit is used to calculate the association degree value between the target feature data and the multiple feature data in each target classification cluster based on the grey correlation analysis algorithm, and take the target feature data and the multiple feature data as a node, and take the association degree values ​​between the target feature data and the multiple feature data as edges between the nodes, and the length of the edge is related to the association degree value; An updating unit, configured to update the plurality of feature data based on the association degree value and the target feature data to obtain the plurality of updated feature data; The fusion unit is used to fuse the target feature data in each target classification cluster with the updated multiple feature data to obtain multiple overall feature data.

Images