A method, device and medium for constructing logging quality parameters of low-permeability reservoirs

Abstract

The present invention relates to a method, device and medium for constructing logging quality parameters of low-permeability reservoirs, including the steps of: preprocessing and standardizing logging data; selecting sensitive logging curves of low-permeability reservoir characteristic parameters from logging curves; using the selected sensitive logging curves of low-permeability reservoir characteristic parameters as neural network inputs and reconstructing sensitive curves based on artificial intelligence algorithms; constructing an empirical formula for low-permeability reservoir quality parameters based on the reconstructed sensitive curves; and classifying low-permeability reservoirs based on the low-permeability reservoir quality parameters. The method for constructing logging quality parameters of low-permeability reservoirs considers using reservoir sedimentation parameters as the basis for classifying logging low-permeability reservoirs. By reconstructing sensitive curves such as grain size and soluble mineral content through artificial intelligence and substituting them into an improved empirical formula for low-permeability reservoir quality parameters, the construction of logging quality parameters of low-permeability reservoirs is realized, and then the reservoir quality is discriminated, which has higher accuracy and geological significance compared with conventional methods.

Description

Technical Field

[0001] The present invention relates to a method, device and medium for constructing low-permeability reservoir logging quality parameters. The distribution of low-permeability reservoir sweet spots is mainly controlled by geological sedimentation laws. Even if the dominant sedimentary phase belt suffers destructive diagenesis in the later stage, dissolution transformation is preferentially carried out under the same conditions, thereby forming a low-permeability reservoir logging quality parameter construction technology. Background Art

[0002] At present, there is great uncertainty in the interpretation of permeability logging of low-permeability reservoirs. Low-permeability reservoirs in medium and deep layers are often accompanied by multiple rock types, high matrix content, complex pore structure and other characteristics, resulting in great uncertainty in the determination of sweet spots in low-permeability reservoirs based on logging permeability. It is urgent to improve the logging identification and classification methods of low-permeability sweet spots. The analysis of the genesis of low-permeability reservoirs in typical low-permeability oil fields in Bohai Sea shows that the distribution of sweet spots is mainly controlled by sedimentary laws. Even if a good sedimentary facies belt suffers destructive diagenesis in the later stage, it will be preferentially dissolved and transformed under the same conditions to form a relatively high-permeability favorable area. Based on this understanding, reservoir sedimentary parameters such as grain size, soluble mineral content and mud content are considered as the basis for the classification of low-permeability reservoirs by logging. The sensitive curves such as grain size and soluble mineral content are reconstructed by artificial intelligence, and substituted into the improved empirical formula of low-permeability reservoir quality parameters to realize the construction of low-permeability reservoir logging quality parameters, and then judge the reservoir quality.

[0003] The quality of low-permeability reservoirs is closely related to key parameters such as rock particle size, soluble mineral content, and mud content. However, the conventional logging fitting formula method cannot predict the key parameters of particle size and soluble mineral content, and nonlinear network model prediction is required. The network model for logging curve prediction is complex and diverse, and its mechanism is mostly to use the data with the best correlation with the predicted logging curve to perform nonlinear fitting through neural networks. With the gradual and in-depth application of artificial intelligence technology in the field of logging, compared with shallow neural networks, artificial intelligence can obtain sensitive functions in deeper network layers in multi-layer neural networks, and its advantages in logging lithology and physical property interpretation are gradually reflected. The use of artificial intelligence technology can more effectively obtain sensitive logging curves of low-permeability reservoirs under the influence of multiple parameters. Summary of the invention

[0004] In view of the above problems, the purpose of the present invention is to provide a method, device and medium for constructing low permeability reservoir logging quality parameters. The method for constructing low permeability reservoir logging quality parameters is based on artificial intelligence technology and has high prediction accuracy.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] The present invention provides a method for constructing low permeability reservoir logging quality parameters, comprising the steps of:

[0007] Preprocess and standardize the logging data;

[0008] Optimize the sensitive logging curves of low-permeability reservoir characteristic parameters from the standardized logging curves;

[0009] Reconstruct the sensitive curves of the optimized sensitive logging curves of low-permeability reservoir characteristic parameters based on artificial intelligence algorithms;

[0010] Obtain the low-permeability reservoir quality parameters based on the reconstructed sensitive curves through a pre-established empirical formula for low-permeability reservoir quality parameters:

[0011]

[0012] In the formula: K is the permeability, is the porosity, md is the grain size, corro is the content of soluble minerals, K1 and K2 are constant coefficients, and vsh is the shale content;

[0013] Classify the low-permeability reservoirs based on the low-permeability reservoir quality parameters.

[0014] The optimization of the preprocessing and standardization of the logging data includes:

[0015] Establish the standard mode of the data histogram of the standard layer section in the key wells of the oilfield;

[0016] Make a histogram of the same type of the corresponding layer section for the processing well and compare it with the standard mode. If the coincidence is poor, it indicates that there is a systematic error, and the error curve is systematically corrected by histogram translation.

[0017] It is preferred that before standardization, normalization and the calculation of the red-green mode are first performed on the three porosity-related logging data of density, acoustic wave and neutron;

[0018] Check whether the measured values of the logging density, acoustic wave and neutron curves are generally too large or too small based on the red-green model data and perform standardization correction.

[0019] The optimization of the sensitive logging curves of low-permeability reservoir characteristic parameters includes the steps of:

[0020] Using the XGBoost data mining technology, first transform the original logging data into a set of linearly independent data in each dimension through linear transformation, and extract the main feature components of the data;

[0021] Rotate the data coordinate axis to the direction where the data angle is important, determine the number of characteristic attributes to be retained through eigenvalue analysis, discard other attributes, and realize data dimensionality reduction;

[0022] Optimize the sensitive logging curves related to grain size and soluble mineral content by measuring the importance of features from the perspective of importance measurement.

[0023] The preferred grain size ranks the logging curves optimized by XGBoost as density, neutron porosity, difference ratio of dual laterolog resistivity, natural gamma, resistivity of saturated water, and acoustic travel time;

[0024] The content of soluble minerals ranks the logging curves optimized by XGBoost as natural gamma, density, acoustic travel time, neutron porosity, resistivity of saturated water, and difference ratio of dual induction resistivity.

[0025] The preferred reconstruction of sensitive curves based on the artificial intelligence algorithm includes the steps:

[0026] Normalize the logging curves preferably obtained through sensitivity analysis of grain size and soluble mineral content, use them as the input of the neural network, and use the reservoir grain size value and soluble mineral content value obtained through core analysis as the output to train the prediction model.

[0027] The preferred training process of the neural network model includes the steps:

[0028] Generate a DNN neural network and initialize this neural network;

[0029] Initialize the relevant parameters of the artificial bee colony algorithm and create the connection weights and thresholds of the DNN;

[0030] Load the relevant data set, determine the input data, output data, and perform normalization processing;

[0031] Train the neural network and obtain the optimal training parameters;

[0032] The training is completed.

[0033] The preferred includes:

[0034] Automatically cluster the logging gamma curve with the reservoir quality parameter RQI obtained in the above steps. According to different RQI thresholds, the reservoir is divided into three categories, where one category is the highest quality reservoir, the second category is the sub - highest quality reservoir, and the third category is the poor reservoir. The RQI of the first category > 10, the RQI of the second category is 1 - 10, and the RQI of the third category is 0.1 - 1.

[0035] In the second aspect, the present invention also provides a device for constructing logging quality parameters of low - permeability reservoirs, including:

[0036] The first processing unit is used for pre - processing and standardizing logging data;

[0037] The second processing unit is used for preferably selecting sensitive logging curves of low - permeability reservoir characteristic parameters from the standardized logging curves;

[0038] The third processing unit reconstructs the sensitive log curves of the low-permeability reservoir characteristic parameters obtained by optimization based on an artificial intelligence algorithm;

[0039] The fourth processing unit is used to obtain the low-permeability reservoir quality parameters based on the reconstructed sensitive curves through a pre-established empirical formula for low-permeability reservoir quality parameters:

[0040]

[0041] In the formula: K is the permeability, is the porosity, md is the grain size, corro is the content of soluble minerals, K1 and K2 are constant coefficients, and vsh is the shale content;

[0042] The fifth processing unit is used to classify the low-permeability reservoir based on the low-permeability reservoir quality parameters.

[0043] In a third aspect, the present invention also provides a computer-readable storage medium storing computer instructions, and the computer instructions are used to implement the method for constructing the logging quality parameters of the low-permeability reservoir when executed by a processor.

[0044] Due to the above technical solutions adopted by the present invention, it has the following advantages:

[0045] In the construction of the logging quality parameters of the low-permeability reservoir, sediment parameters of the reservoir such as grain size, content of soluble minerals, and shale content are considered as the basis for logging classification of the low-permeability reservoir. By reconstructing sensitive curves such as grain size and content of soluble minerals through artificial intelligence and substituting them into the improved empirical formula for low-permeability reservoir quality parameters, the construction of the logging quality parameters of the low-permeability reservoir is realized, and then the reservoir quality is discriminated, which has higher accuracy and geological significance compared with the conventional method. Description of the Drawings

[0046] By reading the detailed description of the preferred embodiments below, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference numerals are used to represent the same components.

[0047] In the drawings:

[0048] Figure 1 is a technical flow chart of a method for constructing the logging quality parameters of a low-permeability reservoir;

[0049] Figure 2 is a comparison chart of logging standardization processing based on the red-green mode;

[0050] Figure 3 is a flow chart of the bee colony algorithm-based deep neural network algorithm (BC-DNN);

[0051] Figure 4 It is the flow chart of the BC-DNN cyclic update process;

[0052] Figure 5 (a) is the crossplot of grain size prediction data and core analysis data, and (b) is the crossplot of soluble mineral content and core analysis data;

[0053] Figure 6 (a), (b), (c), (d) and (e) are the correlation diagrams of Class I, II, and III reservoirs discriminated based on the RQI value with porosity, permeability, shale content, grain size, and soluble mineral content respectively. Detailed implementation manners

[0054] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present invention can be more thoroughly understood and the scope of the present invention can be fully conveyed to those skilled in the art.

[0055] An embodiment of the present invention provides a method for constructing logging quality parameters of low-permeability reservoirs, including the steps of: preprocessing and standardizing logging data; preferably sensitive logging curves of low-permeability reservoir characteristic parameters; using the preferably obtained sensitive logging curves of low-permeability reservoir characteristic parameters as neural network inputs, and reconstructing the sensitive curves based on artificial intelligence algorithms; constructing an empirical formula for low-permeability reservoir quality parameters; and classifying low-permeability reservoirs based on the low-permeability reservoir quality parameters. The method for constructing logging quality parameters of low-permeability reservoirs considers using reservoir deposition parameters such as grain size, soluble mineral content, and shale content as the basis for classifying low-permeability logging reservoirs. By artificially reconstructing sensitive curves such as grain size and soluble mineral content and substituting them into an improved empirical formula for low-permeability reservoir quality parameters, the construction of logging quality parameters of low-permeability reservoirs is realized, and then the reservoir quality is discriminated, which has higher accuracy and geological significance compared with conventional methods.

[0056] Example 1

[0057] As Figure 1 shown, Figure 1 is the technical flow chart of the method for constructing logging quality parameters of low-permeability reservoirs. The method for constructing logging quality parameters of low-permeability reservoirs is characterized by including the steps of:

[0058] S1. Preprocess and standardize the logging data;

[0059] The conventional natural gamma logging curve is standardized by using the multi-well histogram comparison method, which specifically includes:

[0060] S1-1. First, establish the standard mode of the data histogram of the standard layer section in the key wells of the oilfield;

[0061] S1-2. Secondly, create histograms for corresponding intervals of the treatment well and compare them with the standard pattern. If the overlap is poor, it indicates the existence of systematic errors, and the error curve is systematically corrected by translating the histogram.

[0062] It should be noted that for the three porosity-related logging data of density (DEN), acoustic wave (AC), and neutron (CNL), the data is first normalized before standardization, and the calculation of the red-green mode is carried out. The formula is as follows.

[0063] Red = CNLgyh - DENgyh; (1)

[0064] Green = DENgyh - ACgyh; (2)

[0065] In the formula: CNLgyh is the data after neutron normalization; DENgyh is the data after density normalization; ACgyh is the data after acoustic wave normalization.

[0066] Well sections with more Red values in the red mode have good physical properties and high porosity and permeability values, while well sections with more Green values in the green mode have poor physical properties and low porosity and permeability values. Based on the red-green model data, check whether the measured values of the logging AC, CNL, and DEN curves are generally too large or too small, and perform standardization correction. It can be seen that before standardization, the red mode of the target interval of this well is too large and the green mode is too low, so the CNL curve of the entire target interval needs to be standardized. Figure 2 It can be seen that before standardization, the red mode of the target interval of this well is too large and the green mode is too low, and the CNL curve of the entire target interval needs to be standardized.

[0067] S2. Optimize the logging curves sensitive to the characteristic parameters of low-permeability reservoirs;

[0068] S2-1. Using the XGBoost data mining technology, first transform the original logging data into a set of linearly independent data in each dimension through linear transformation, extract the main characteristic components of the data, and on this basis, rotate the data coordinate axes to the direction where the data angle is important. Through eigenvalue analysis, determine the number of characteristic attributes to be retained, discard other attributes, and achieve data dimensionality reduction.

[0069] S2-2. Optimize the logging curves sensitive to grain size and soluble mineral content by measuring the importance of features from the perspective of importance measurement.

[0070] The sorting of the logging curves optimized by XGBoost for grain size: density (DEN), neutron porosity (CNL), ratio of dual laterolog resistivity difference (RDS), natural gamma ray (GR), resistivity of saturated water (R0), acoustic wave transit time (AC);

[0071] Sorting of well logging curves optimized by XGBoost for soluble mineral content: natural gamma ray (GR), density (DEN), acoustic travel time (AC), neutron porosity (CNL), resistivity of water-saturated formation (R0), ratio of dual induction resistivity difference (RDM);

[0072] S3. Use the sensitive well logging curves of the low-permeability reservoir characteristic parameters obtained by the above optimization as the input of the neural network, and reconstruct the sensitive curves based on the artificial intelligence algorithm;

[0073] Normalize the well logging curves optimized by the sensitivity analysis of grain size and soluble mineral content as the input of the neural network, and use the reservoir grain size value and soluble mineral content value obtained by core analysis as the output to train the prediction model.

[0074] Select the deep neural network based on the bee colony algorithm (BC-DNN) algorithm in the artificial intelligence algorithm. It uses the bee colony (BC) algorithm to optimize the weight parameters (W ij ) and thresholds (b j ) in the deep neural network (DNN). Realize the double-loop mutual adjustment of the dynamic adjustment of the weight parameters and thresholds in the deep neural network and the optimization of the bee colony algorithm, so as to reduce the number of iterations of the deep neural network. The technical process is as Figure 3 shown, specifically including the following steps:

[0075] S3-1. Generate a DNN neural network and initialize this neural network;

[0076] S3-2. Initialize the relevant parameters of the artificial bee colony algorithm. The main parameters include the number of artificial bee colonies, the number of leading bees, the number of follower bees, the number of solutions N of the bee colony algorithm, the threshold of the number of times of failure to update the optimal solution, the maximum number of iterations, etc., and create the connection weights and thresholds of the DNN;

[0077] S3-3. Load the relevant data set, determine the input data, output data, and perform normalization processing;

[0078] S3-4. Train the neural network and obtain the optimal training parameters;

[0079] S3-5. End of training.

[0080] The DNN neural network layer can be divided into three categories: input layer, hidden layer and output layer. The first layer is the input layer, the last layer is the output layer, and the middle layers are all hidden layers. There are full connections between layers, that is, any neuron in the i-th layer must be connected to any neuron in the i+1-th layer. In the DNN, the process of optimizing the extreme value of the loss function is completed by iterating through the gradient descent method. For each training sample, it is expected to minimize the following formula:

[0081]

[0082] where a L and y are vectors with a feature dimension of n_out, and ||S||2 is the L2 norm of S.

[0083] When solving the optimization problem, the bee colony algorithm can be divided into four stages: initialization stage, leading bee update, following bee update, and scout bee update.

[0084] Initialization stage: First, initialize the population of the artificial bee colony algorithm. The number of evolutionary generations is 200, and the population size N is 600. Generate the food source X according to formula (4) i , where rand() is a random matrix equivalent to the scale of the weight matrix, and its value range is [-1, 1].

[0085] X i = X min + rand() × (X max - X min ) (4)

[0086] Leading bee update: According to the neural network parameters, the update formula is shown in formula (5)

[0087] V(P) = X(k, P) + (X(k, P) - X(nei, P)) * rand i + (Gbest(1, P) - X(nei, P)) * rand j (5)

[0088] P is the number of mutation dimensions, which can be determined as an integer within the weight dimension range according to the neural network nodes. nei is the selected corresponding mutation neighbor, and the value of nei is prevented from being the same as k. Calculate the fitness value, limit the position range, and retain the better fitness value and the corresponding food source position. Update the optimal food source position and the fitness value. If the fitness value is not updated, the stagnation parameter m of the scout bee is incremented by 1, laying the groundwork for the scout bee update. rand j and rand i are random numbers between -1 and 1.

[0089] Following bee update: Determine the update probability P of each individual according to the roulette wheel rule i , generate a random number between 0 and 1, and when the random number is less than the probability P i , update according to formula (5). Similarly, calculate the fitness value, limit the position range, and retain the better fitness value and the corresponding food source position. Update the optimal food source position and the fitness value. If the fitness value is not updated, the stagnation parameter m of the scout bee is incremented by 1.

[0090] Scout bee update: When the stagnation condition m reaches 100, scout bee update is performed. The update formula is as follows to update the optimal food source position and fitness value.

[0091] V i = X min + rand i () × (X max - X min ) (6)

[0092] The BC-DNN cyclic update process is as Figure 4 shown. Figure 5 (a) is the crossplot analysis of the grain size data predicted by this method and the actual core analysis grain size data. Figure 5 (b) is the crossplot analysis of the soluble mineral content data predicted by this method and the actual core analysis soluble mineral content data. It can be seen that the predicted data and the measured data have a high degree of agreement.

[0093] S4. Construct an empirical formula for low-permeability reservoir quality parameters;

[0094] In low-porosity and low-permeability reservoirs, the complexity of the pore structure results in a large difference in permeability between reservoirs with basically the same porosity. Therefore, the method of classifying reservoir quality using a single parameter such as porosity, permeability, or their combination in low-porosity and low-permeability reservoirs is obviously not applicable. Reservoir physical properties are closely related to rock grain size, soluble mineral content parameters, and shale content. In addition to considering porosity and permeability, an improved method for constructing the low-permeability reservoir quality parameter RQI is proposed by further considering grain size, soluble mineral content, and shale content. The formula is as follows:

[0095]

[0096] In the formula: K is the permeability, is the porosity, md is the grain size, corro is the soluble mineral content, K1 and K2 are constant coefficients, and vsh is the shale content.

[0097] S5. Classify low-permeability reservoirs based on low-permeability reservoir quality parameters.

[0098] By using the logging gamma curve and the reservoir quality parameter RQI obtained in the above steps for automatic clustering, according to different RQI thresholds, the reservoirs are divided into three categories. Among them, the first category is the highest-quality reservoir, the second category is the sub-high-quality reservoir, and the third category is the poor reservoir. The class I, II, and III reservoirs discriminated by the RQI value are crossplotted with porosity (POR), permeability (K), grain size (Md), soluble mineral content (Corro), and shale content (Vsh), as Figure 6As shown, and the numerical range is counted, as shown in Table 1. It can be seen that there is a good correlation, indicating that the low-permeability reservoir quality parameter RQI established by the present invention can better divide the low-permeability reservoir quality.

[0099] Table 1 Statistical Table of Reservoir Classification Parameters

[0100] Classification RQI POR (%) K (md) Md (um) Corro (%) Vsh (%) Class I >10 12-18 2-50 300-800 0.5-1.8 0-16 Class II 1-10 10.5-16 0.8-9 70-600 0.2-1.4 4-25 Class III 0.1-1 9.5-15 0.2-6 30-550 0.1-1.2 6-30

[0101] In the construction of logging quality parameters for low-permeability reservoirs, the present invention considers using reservoir sedimentation parameters such as grain size, soluble mineral content, and shale content as the basis for classifying logging low-permeability reservoirs. By using artificial intelligence to reconstruct sensitive curves such as grain size and soluble mineral content, and substituting them into the improved empirical formula for low-permeability reservoir quality parameters, the construction of logging quality parameters for low-permeability reservoirs is realized, and then the reservoir quality is discriminated, which has higher accuracy and geological significance compared with conventional methods.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for constructing logging quality parameters of a low-permeability reservoir, characterized in that, Including the steps: Preprocess and standardize the logging data; Optimize the logging curves sensitive to the characteristic parameters of low-permeability reservoirs from the standardized logging curves; Reconstruct the sensitive curves of the low-permeability reservoir characteristic parameters based on the artificial intelligence algorithm for the optimized sensitive logging curves; Obtain the low-permeability reservoir quality parameters based on the reconstructed sensitive curves through the empirical formula of the low-permeability reservoir quality parameters constructed in advance: where: K is the permeability, φ is porosity, md is the grain size, corro is the content of soluble minerals, K1 and K2 are constant coefficients, and vsh is the shale content; Classify the low-permeability reservoirs based on the low-permeability reservoir quality parameters; The preprocessing and standardization of the logging data include: Establish the standard mode of the data histogram of the standard layer section in the key wells of the oilfield; Make the histogram of the same layer section for the processing wells and compare it with the standard mode. If the coincidence is poor, it indicates that there is a systematic error, and the error curve is systematically corrected by histogram translation; Perform normalization and the calculation of the red-green mode on the logging data related to density, acoustic wave, and neutron porosity before standardization; Check whether the measured values of the logging density, acoustic wave, and neutron curves are too large or too small as a whole based on the red-green model data and perform standardization correction; The steps for optimizing the logging curves sensitive to the characteristic parameters of low-permeability reservoirs include: Using the XGBoost data mining technology, first transform the original logging data into a set of linearly independent data in each dimension through linear transformation, and extract the main feature components of the data; Rotate the data coordinate axis to the direction where the data angle is important, determine the number of characteristic attributes to be retained through eigenvalue analysis, discard other attributes, and realize data dimensionality reduction; Optimize the logging curves sensitive to grain size and soluble mineral content by measuring the importance of features from the perspective of importance measurement; The reconstruction of the sensitive curves based on the artificial intelligence algorithm includes the steps: Normalize the logging curves optimized through the sensitivity analysis of grain size and soluble mineral content, use them as the input of the neural network, and use the reservoir grain size value and soluble mineral content value obtained through core analysis as the output to train the prediction model.

2. The method for constructing logging quality parameters of a low-permeability reservoir according to claim 1, characterized in that, The logging curves optimized for grain size by XGBoost are ranked as density, neutron porosity, difference ratio of dual laterolog resistivity, natural gamma, resistivity of saturated water, acoustic travel time; The logging curves optimized for soluble mineral content by XGBoost are ranked as natural gamma, density, acoustic travel time, neutron porosity, resistivity of saturated water, difference ratio of dual induction resistivity; 3. The method for constructing logging quality parameters of a low-permeability reservoir according to claim 2, wherein The training process of the neural network model includes the steps: Generate a DNN neural network and initialize this neural network; Initialize the relevant parameters of the artificial bee colony algorithm and create the connection weights and thresholds of the DNN; Load the relevant data sets, determine the input data, output data, and perform normalization processing; Train the neural network and obtain the optimal training parameters; The training ends.

4. The method for constructing logging quality parameters of a low-permeability reservoir according to claim 1, wherein The classification of low-permeability reservoirs based on the low-permeability reservoir quality parameters includes: By automatically clustering the logging gamma curve and the reservoir quality parameter RQI obtained in the above steps, according to different RQI thresholds, the reservoir is divided into three categories, where the first category is the highest quality reservoir, the second category is the sub-high quality reservoir, and the third category is the poor reservoir. The RQI of the first category is > 10, the RQI of the second category is 1 - 10, and the RQI of the third category is 0.1 - 1.

5. A device for constructing logging quality parameters of a low-permeability reservoir, characterized in that It includes: The first processing unit is used to preprocess and standardize the logging data; The second processing unit is used to select sensitive logging curves of low-permeability reservoir characteristic parameters from the standardized logging curves; The third processing unit reconstructs the sensitive curves of the selected sensitive logging curves of low-permeability reservoir characteristic parameters based on an artificial intelligence algorithm; The fourth processing unit is used to obtain the low-permeability reservoir quality parameter based on the reconstructed sensitive curve through a pre-constructed empirical formula for low-permeability reservoir quality parameters: where: K is the permeability, φ is porosity, md is the grain size, corro is the content of soluble minerals, K1 and K2 are constant coefficients, and vsh is the shale content; The fifth processing unit is used to classify the low-permeability reservoir based on the low-permeability reservoir quality parameter; The preprocessing and standardization of the logging data include: Establishing the standard mode of the data histogram of the standard layer section in the key wells of the oilfield; Making a histogram of the same layer section of the processing well and comparing it with the standard mode. If the coincidence is poor, it indicates that there is a systematic error, and the error curve is systematically corrected by histogram translation; For the three porosity-related logging data of density, acoustic wave, and neutron, normalization and the calculation of the red-green mode are performed before standardization; Checking whether the measured values of the logging density, acoustic wave, and neutron curves are generally too large or too small based on the red-green model data and performing standardization correction; The selection of sensitive logging curves of low-permeability reservoir characteristic parameters includes the steps: Using the XGBoost data mining technology, first linearly transforming the original logging data into a set of linearly independent data in each dimension, and extracting the main feature components of the data; Rotating the data coordinate axis to the direction where the data angle is important, determining the number of characteristic attributes to be retained through eigenvalue analysis, and discarding other attributes to achieve data dimensionality reduction; Selecting sensitive logging curves related to grain size and soluble mineral content by measuring the importance of features from the perspective of importance measurement; The reconstruction of sensitive curves based on the artificial intelligence algorithm includes the steps: Normalizing the logging curves selected through the sensitivity analysis of grain size and soluble mineral content, using them as the input of the neural network, and using the reservoir grain size value and soluble mineral content value obtained through core analysis as the output to train the prediction model.

6. A computer-readable storage medium stores computer instructions, and the computer instructions are used to implement the method for constructing the logging quality parameters of low-permeability reservoirs as described in any one of claims 1 - 4 when executed by a processor.

Images