Method, device and equipment for determining core vitrinite reflectivity and storage medium

Abstract

The invention discloses a core vitrinite reflectance determination method, device and equipment and a storage medium. The method comprises the steps of determining logging data of a target depth section, inputting the logging data into an adversarial neural network model, and outputting a feature vector matched with the logging data through a generator network in the adversarial neural network model; inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value; and determining the core vitrinite reflectivity of the target depth section according to the neural network model predicted value and / or the random forest model predicted value. By using the technical scheme of the invention, the core vitrinite reflectivity of the shale rock can be accurately and quickly determined, so that automatic identification of the graphitized low-resistance shale is realized, and data support is provided for risk exploration of the low-resistance shale.

Description

Technical Field

[0001] The present invention relates to the technical field of oil and gas exploration, and in particular to a method, device, equipment and storage medium for determining the reflectivity of core vitrinite. Background Art

[0002] The organic maturity of graphitized low-resistance shale reservoirs is significantly higher than that of other low-resistance shales, and the gas content is worse than that of other low-resistance shales. The reason may be related to the high degree of organic graphitization. After the organic matter undergoes a large-scale graphitization, the gas production capacity of the shale will be greatly reduced. The core vitrinite reflectance is represented by Ro. The Ro value is used to evaluate the organic maturity of shale rocks and describe the degree of evolution of organic matter in rocks. How to accurately predict the Ro value and screen the development location of graphitized low-resistance shale in a single well based on the Ro value is crucial for shale gas exploration and development.

[0003] Traditional Ro value prediction methods include petrological methods and geochemical methods. The petrological method mainly relies on petrological characteristics (such as color, texture, etc.) for judgment, lacking quantitative analysis; while the geochemical method is to judge by measuring the physical and chemical properties of organic matter in rocks (such as element content, volatile matter, etc.), but it requires sampling and analysis, and has limited accuracy. Summary of the invention

[0004] The present invention provides a method, device, equipment and storage medium for determining the core vitrinite reflectance of shale rock, so as to accurately and quickly determine the core vitrinite reflectance of shale rock, thereby realizing automatic identification of graphitized low-resistance shale and providing data support for risk exploration of low-resistance shale.

[0005] In a first aspect, an embodiment of the present invention provides a method for determining a core vitrinite reflectance, the method comprising:

[0006] Determine the well logging data of the target depth section, and input the well logging data into the adversarial neural network model, and output a feature vector matching the well logging data through a generator network in the adversarial neural network model;

[0007] Inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value;

[0008] The core vitrinite reflectance of the target depth section is determined based on the predicted values ​​of the neural network model and / or the predicted values ​​of the random forest model.

[0009] In a second aspect, an embodiment of the present invention further provides a device for determining a core vitrinite reflectance, the device comprising:

[0010] A feature vector determination module is used to determine the well logging data of the target depth section, and input the well logging data into the adversarial neural network model, and output a feature vector matching the well logging data through a generator network in the adversarial neural network model;

[0011] A prediction module, used for inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value;

[0012] The core vitrinite reflectance determination module is used to determine the core vitrinite reflectance of the target depth section according to the prediction value of the neural network model and / or the prediction value of the random forest model.

[0013] In a third aspect, an embodiment of the present invention further provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, a method for determining the reflectance of core vitrinite as described in any one of the embodiments of the present invention is implemented.

[0014] In a fourth aspect, an embodiment of the present invention further provides a storage medium storing computer executable instructions, wherein the computer executable instructions, when executed by a computer processor, are used to execute a method for determining the reflectance of core vitrinite as described in any one of the embodiments of the present invention.

[0015] The technical solution of the embodiment of the present invention determines the core vitrinite reflectance of the target depth section by inputting the logging data of the target depth section into the adversarial neural network, outputting the feature vector through the generator network of the adversarial neural network, and inputting the feature vector into the neural network prediction model and / or the random forest prediction model to obtain the neural network model prediction value and / or the random forest model prediction value. The problem of low accuracy of the method for determining the core vitrinite reflectance in the prior art is solved, and the core vitrinite reflectance of shale rock can be accurately and quickly determined, thereby realizing the automatic identification of graphitized low-resistance shale and providing data support for the risk exploration of low-resistance shale.

[0016] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0018] Figure 1 This is a flow chart of a method for determining the reflectance of core vitrinite provided in Example 1 of the present invention;

[0019] Figure 2 This is a flow chart of a method for determining the reflectance of core vitrinite provided in the second embodiment of the present invention;

[0020] Figure 3 is a schematic structural diagram of a device for determining the reflectance of core vitrinite provided in Embodiment 3 of the present invention;

[0021] Figure 4 It is a structural schematic diagram of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0022] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme 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 only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0023] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0024] Embodiment 1

[0025] Figure 1 A flow chart of a method for determining the core vitrinite reflectance is provided for the first embodiment of the present invention. This embodiment is applicable to the case of predicting the core vitrinite reflectance of low-resistance shale. The method can be executed by a device for determining the core vitrinite reflectance. The device for determining the core vitrinite reflectance can be implemented in the form of hardware and / or software. The device for determining the core vitrinite reflectance can be configured in an electronic device.

[0026] like Figure 1As shown, the method includes:

[0027] S110, determining the logging data of the target depth section, and inputting the logging data into an adversarial neural network model, and outputting a feature vector matching the logging data through a generator network in the adversarial neural network model.

[0028] The target depth segment is a formation depth range consisting of two formation depths, such as 1km-2km. This embodiment does not limit the formation depth ranges corresponding to different depth segments and the division method of different depth segments. Well logging data refers to well logging information obtained by logging the formation to be identified through various well logging methods.

[0029] The adversarial neural network model is used to extract features from the logging data of the target depth segment and output a high-dimensional feature vector. The structure of the adversarial neural network can include a generator network and a discriminator network. The generator network is used to learn the potential distribution of the logging data and convert the logging data into a high-dimensional feature vector representation, while the discriminator network is used to distinguish the authenticity of the data generated by the generator network. Among them, the input of the generator network is the logging data, and the output is a high-dimensional feature vector, which can be expressed as: Among them, x is the logging data, f is the output of the generator network, Denote the parameters of the generator network as θ g The discriminator network can adopt the structure of a convolutional neural network, which consists of multiple convolutional layers and pooling layers, and finally a fully connected layer to output the binary classification results. The input of the discriminator network is a feature vector, and the output is a binary classification result, which can be expressed as: Among them, y represents the binary classification result, Denotes the parameter of the discriminator network as θ d .

[0030] Furthermore, the training process of the adversarial neural network model may include: determining historical logging data and historical core vitrinite reflectivity; inputting the historical logging data into the generator network, and determining the predicted core vitrinite reflectivity that matches the historical logging data through the generator network; judging the authenticity of the predicted core vitrinite reflectivity based on the historical core vitrinite reflectivity and the predicted core vitrinite reflectivity through the discriminator network; calculating the model loss according to the adversarial neural network loss function, and repeatedly inputting the historical logging data into the generator network until the model loss meets the convergence condition.

[0031] Furthermore, in order to enhance the feature extraction capability of the generator network, the generator network is combined with a sparse autoencoder. A sparse autoencoder is an autoencoder that can learn a sparse representation of data, that is, automatically select the most important features in the data and compress them into a sparse vector. In the generator network, the encoder part of the sparse autoencoder is used as the input layer of the generator, and the decoder part is used as the output layer of the generator. The encoder of the sparse autoencoder can be expressed as: h = f(Wx + b), where x is the input logging data, W and b are the parameters of the encoder, f is the activation function, and h is the output of the encoder. The decoder part adopts the structure of a deconvolutional network to restore the output of the encoder to a form that matches the vitrinite reflectance of the historical core.

[0032] Furthermore, a sparsity constraint is introduced into the sparse autoencoder to further improve the feature extraction capability of the generator network. A sparsity penalty term is introduced into the loss function of the sparse autoencoder, which can be expressed by the following formula: Where ρ is the expected average activation, is the average activation of the jth neuron. The purpose of this sparsity penalty term is to encourage the encoder to learn as few activated neurons as possible, thereby obtaining a more sparse representation. The loss function of the sparse autoencoder can be expressed by the following formula: in, Represents the reconstruction error, h represents the output of the encoder, and ||h||1 represents the sparse constraint imposed on the encoded feature vector. The adversarial neural network model can be trained according to the adversarial training method based on the Wasserstein distance, and its goal is to minimize the generator and the discriminator The loss function of the generator and the discriminator The loss function can be expressed by the following formula: Among them, the first and second terms represent the average predicted values ​​of the historical core vitrinite reflectance and the predicted core vitrinite reflectance in the discriminator, respectively. represents the KL divergence term in the sparse autoencoder, and λ represents the weight parameter between the KL divergence term and other terms. The total loss function of the adversarial neural network model is: L = L(θ g ,θ d )+γ·L AE , where γ represents the weight parameter controlling the sparse autoencoder and the adversarial neural network.

[0033] Furthermore, the adversarial neural network model that introduces the sparse autoencoder can be trained by alternating optimization, that is, in the model training, the parameters of the generator network and the discriminator network, and the parameters of the sparse autoencoder are updated and optimized in turn. Specifically, the training process may include: determining the historical logging data and the historical core vitrinite reflectance; inputting the historical logging data into the generator network, and determining the predicted core vitrinite reflectance that matches the historical logging data through the generator network; judging the authenticity of the predicted core vitrinite reflectance based on the historical core vitrinite reflectance and the predicted core vitrinite reflectance through the discriminator network; calculating the model loss according to the adversarial neural network loss function, and updating the parameters of the discriminator network. Then the historical logging data is input into the generator network, and the predicted core vitrinite reflectance that matches the historical logging data is determined by the generator network. The authenticity of the predicted core vitrinite reflectance is judged again by the discriminator network based on the historical core vitrinite reflectance and the predicted core vitrinite reflectance, and the model loss is calculated according to the adversarial neural network loss function, and the parameters of the generator network are updated. Finally, the sparse autoencoder is used to reconstruct the data, and the loss function of the reconstruction error and the KL divergence term is calculated. Then, the back propagation algorithm is used to update the parameters of the sparse autoencoder. Model training by alternating optimization can improve the model performance of the trained adversarial neural network.

[0034] S120, inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value.

[0035] This embodiment provides two prediction methods for core vitrinite reflectance (hereinafter represented by Ro value), and the Ro value prediction is performed through a neural network prediction model and / or a random forest prediction model.

[0036] Among them, the neural network prediction model is a regression model based on a neural network, which takes the feature vector as input and outputs the Ro value prediction result. The neural network prediction model can adopt a multi-layer perceptron structure, which consists of an input layer, several hidden layers and an output layer, and the back propagation algorithm can be used to train the model. Specifically, if the feature vector x output by the generator network of the adversarial neural network model is n-dimensional, the neural network prediction model calculates the prediction result of the Ro value through the following formula: Where g(·) is the activation function; W h and b h is the weight matrix and bias vector of the hidden layer; W o and b o are the weight matrix and bias vector of the output layer; f(·) is the activation function of the output layer, usually the Sigmoid function.

[0037] Furthermore, in order to improve the stability of the neural network prediction model, the Dropout technology can be used in the model training process. This technology can randomly discard a certain proportion of neurons in each iteration, which helps to reduce overfitting and improve the generalization ability of the model.

[0038] Among them, the random forest prediction model constructs multiple decision trees by randomly selecting features and combines them for prediction. When selecting features, common indicators such as the Gini index or information gain can be used.

[0039] In this embodiment, Ro value prediction is performed through two prediction models. When predicting the Ro value, one or two of them can be flexibly selected according to the actual needs of the prediction. This setting solves the problem of low prediction accuracy in traditional Ro value prediction, and has higher robustness and generalization ability.

[0040] S130. Determine the core vitrinite reflectance of the target depth segment according to the predicted value of the neural network model and / or the predicted value of the random forest model.

[0041] In this embodiment, when a prediction model is selected for Ro value prediction, the Ro prediction value of the prediction model is directly used as the Ro value of the target depth segment. When the neural network prediction model and the random forest prediction model are used for Ro value prediction at the same time, the prediction values ​​obtained by the two prediction methods are combined to determine the Ro value of the target depth segment.

[0042] Specifically, the predicted values ​​obtained by the two prediction methods and the logging data can be combined to determine whether there are outliers in the predicted values. If there are outliers, the outliers are removed, and the Ro predicted value without the abnormality is used as the Ro value of the target depth segment. The abnormal cause of the prediction is analyzed and predicted based on the outlier, and the prediction model is optimized. If there are no outliers, the average value of the predicted values ​​can be calculated based on the predicted values ​​of the neural network model and the predicted values ​​of the random forest model, and the average value can be used as the Ro value of the target depth segment. The maximum value of the two predicted values ​​can also be used as the Ro value of the target depth segment. This embodiment does not limit the specific method of determining the final Ro value of the target depth segment based on the two predicted values.

[0043] Furthermore, after determining the core vitrinite reflectance of the target depth segment, the method further includes: if it is determined that the core vitrinite reflectance of the target depth segment is greater than or equal to a preset reflectance threshold, determining that the target depth segment is a graphitized low-resistance shale development depth segment.

[0044] The preset reflectivity threshold can be set to 3.5%, but this embodiment does not limit the specific value and setting method of the preset reflectivity threshold. The depth section with Ro value greater than or equal to 3.5% is used as the distribution interval of graphitized low-resistance shale, and the development position, development depth and thickness of graphitized low-resistance shale are automatically identified, which greatly improves the evaluation accuracy and efficiency of low-resistance shale and provides an effective basis for low-resistance shale risk exploration.

[0045] The technical solution of the embodiment of the present invention determines the core vitrinite reflectance of the target depth section by inputting the logging data of the target depth section into the adversarial neural network, outputting the feature vector through the generator network of the adversarial neural network, and inputting the feature vector into the neural network prediction model and / or the random forest prediction model to obtain the neural network model prediction value and / or the random forest model prediction value. The problem of low accuracy of the method for determining the core vitrinite reflectance in the prior art is solved, and the core vitrinite reflectance of shale rock can be accurately and quickly determined, thereby realizing the automatic identification of graphitized low-resistance shale and providing data support for the risk exploration of low-resistance shale.

[0046] Embodiment 2

[0047] Figure 2 A flow chart of a method for determining the reflectivity of core vitrinite provided in Example 2 of the present invention. Based on the above embodiments, the training process of the adversarial neural network model, the neural network prediction model and the random forest prediction model is further specified in this embodiment of the present invention, and a judgment process of low-resistance shale graphitization based on the reflectivity of core vitrinite is added.

[0048] like Figure 2 As shown, the method includes:

[0049] S210. Determine historical logging data and historical core vitrinite reflectance.

[0050] In this embodiment, the well logging data corresponding to the well section with known Ro value and its Ro value are used as sample data to train the adversarial neural network model, the neural network prediction model and the random forest prediction model.

[0051] Furthermore, S210 may include:

[0052] S1. Determine historical well logging curves and historical core vitrinite reflectance of at least two well logging data types.

[0053] Among them, the logging data types may include natural gamma (GR), uranium-free gamma (KTH), uranium (U), thorium (TH), potassium (K), acoustic time difference (AC), neutron (CNL), density (DEN), deep lateral resistivity (RT) and shallow lateral resistivity (RXO), etc. This embodiment does not limit the specific content and quantity of the logging data types.

[0054] S2. Determine the correlation between each logging data type and the core vitrinite reflectivity based on the historical logging curves of each logging data type and the historical core vitrinite reflectivity.

[0055] Among them, correlation is used to characterize the contribution of logging data types to the Ro value prediction results. Correlation analysis of different types of logging data can reduce data redundancy.

[0056] S3. Determine historical logging data according to historical logging curves of logging data types whose correlations meet preset correlation conditions.

[0057] Meeting the preset correlation condition may be to sort the correlations between different logging data types and Ro values ​​from high to low, and select the first preset number of logging data types with the highest correlations as the logging data types whose correlations meet the preset correlation condition. Alternatively, a correlation threshold may be pre-set, and the logging data types whose correlations with the Ro value are greater than or equal to the correlation threshold are used as the logging data types whose correlations meet the preset correlation condition. This embodiment does not limit the specific content of the preset correlation condition.

[0058] Taking the correlation analysis of the above-mentioned ten logging curves of different logging data types, namely GR, KTH, U, TH, K, AC, CNL, DEN, RT and RXO, as an example, it is found that the nine logging curves of DEN, RT, RXO, CNL, K, AC, GR, U and KTH are more important in the prediction of Ro value. Therefore, the above nine types of logging curves can be selected as the input of Ro value prediction.

[0059] Furthermore, after screening out various logging data types that meet the preset correlation conditions, data preprocessing can be performed to improve data redundancy, missingness, errors, etc. It can be understood that during the model training process and after the subsequent training to obtain the model, in the process of predicting the Ro value of the target depth section based on the logging data of the target depth section, data preprocessing can be performed to improve data quality and ensure the accuracy of Ro value prediction.

[0060] Specifically, for redundant or repeated data, direct deletion can be performed. For missing values, mean filling can be used to process them. Mean filling is to analyze the characteristics of the logging data type and determine whether the logging data type is a numerical feature or a non-numerical feature. If it is a numerical feature, the average value of the other remaining values ​​corresponding to the logging data type is calculated, and the average value is interpolated to the vacancy. If it is a non-numerical feature, the number of occurrences of the other remaining objects corresponding to the logging data type can be counted according to the majority principle, and the object with the largest number of occurrences can be selected to interpolate to the vacancy. For error values, the chi-square test method of non-parametric verification can be used for detection, and the deviation value between suspected error data and normal data can be counted. The larger the deviation value, the higher the possibility of data abnormality, and the smaller the deviation value, the greater the possibility of data being normal data. Abnormal data can be directly eliminated. The data can also be normalized. Normalization can better reflect the relationship between logging data and prediction results, and reduce the impact of different magnitudes on prediction results. The range normalization method can be used to normalize the data.

[0061] Furthermore, S3 may include:

[0062] S31. Determine a combination of well logging data types among the well logging data types whose correlations satisfy preset correlation conditions according to a multi-objective feature selection method.

[0063] In this embodiment, after screening various logging data types that meet the preset correlation conditions and performing data preprocessing, the optimal combination of logging data types can be further determined, thereby improving the accuracy of Ro value prediction.

[0064] A multi-objective feature selection method, in which one objective function is to minimize the error and the other objective function is to minimize the number of features, is used to determine the optimal combination of logging data types, which enables the model to achieve a balance between prediction accuracy and the number of features.

[0065] Specifically, for the preprocessed logging data set D = {(x1, y1), (x2, y2), ..., (x m ,y m )}, where x i is an n-dimensional feature vector, y i is a k-dimensional target vector. Select an optimal feature subset Where |S| = s. Two objective functions are defined, one is to minimize the mean square error (MSE): Another is to minimize the number of features: minimizef2(S) = |S|, where is the prediction result of the jth target value of the i-th sample by the model trained with feature subset S, where |S| represents the size of the feature subset. The two objective functions are solved to obtain the optimal combination of well logging data types.

[0066] S32. Determine historical logging data according to historical logging curves of each logging data type in the logging data type combination.

[0067] S220. Training the adversarial neural network model based on historical logging data, historical core vitrinite reflectivity, and adversarial neural network loss function.

[0068] Among them, the adversarial neural network loss function is obtained according to the loss functions of the generator network and the discriminator network and the sparse autoencoder loss function based on the sparsity penalty term.

[0069] The specific functions of the generator network, the discriminator network, and the sparse autoencoder, the loss functions of the generator network and the discriminator network, the sparsity penalty term, the sparse autoencoder loss function, and the adversarial neural network loss function, as well as the training process of the adversarial neural network model are all explained in the above embodiments and will not be repeated in this embodiment.

[0070] S230, inputting the historical logging data into the adversarial neural network model to obtain a feature vector matching the historical logging data.

[0071] S240, inputting the feature vector matching the historical logging data into a neural network prediction model to obtain a predicted core vitrinite reflectance matching the historical logging data.

[0072] Assume that historical logging data is input into the adversarial neural network model to obtain an n-dimensional feature vector x. Input x into the neural network prediction model to obtain the predicted Ro value: Among them, g(·) is the activation function, usually using the ReLU function; W h and b h is the weight matrix and bias vector of the hidden layer; W o and b o are the weight matrix and bias vector of the output layer; f(·) is the activation function of the output layer, usually the Sigmoid function.

[0073] S250, training the neural network prediction model according to the predicted core vitrinite reflectance, the historical core vitrinite reflectance and the neural network prediction model loss function based on the regularization term.

[0074] In this embodiment, in order to further improve the prediction accuracy, a regularization term, such as L1 or L2 regularization, is introduced to avoid overfitting. The loss function of the neural network prediction model based on the regularization term can be expressed by the following formula: Where m is the number of input feature vectors, y i and are the historical core vitrinite reflectance and the predicted core vitrinite reflectance corresponding to the i-th eigenvector, respectively; p is the number of parameters in the model, w j is the jth parameter; λ is the regularization coefficient, which is used to balance the contribution of the error term and the regularization term.

[0075] To train the model, the Adam algorithm can be used to minimize the objective function J. Specifically, the model parameters are updated using the following formula: Among them, α is the learning rate, is the objective function J with respect to w j The partial derivative of .

[0076] S260: Determine, according to the decision tree weights of the decision trees, the feature weights of the features of each dimension of the feature vector that matches the historical well logging data.

[0077] In this embodiment, in order to improve the prediction performance of the random forest prediction model, an adaptive weight feature selection method is introduced into the random forest prediction model to learn the importance of each dimension feature and introduce the importance weight into the random forest. When predicting, the random forest will perform weighted processing according to the importance of each feature, thereby improving the prediction performance.

[0078] Specifically, the GBDT (Gradient Boosting Decision Tree) method can be used to determine the weight of each decision tree. After obtaining the weight of each decision tree, it is introduced into the random forest. For each feature x k , we can define its feature weight as: Where T is the number of decision trees in the random forest, is the t-th decision tree for feature x k The weight of .

[0079] S270, determining the predicted core vitrinite reflectance matching the historical well logging data according to each decision tree and the feature weight of each dimensional feature of the feature vector matching the historical well logging data.

[0080] Specifically, for any input sample x, the prediction result of the random forest prediction model on its output can be expressed as: Among them, f t (x) is the prediction result of the tth decision tree for the input sample x.

[0081] S280, training a random forest prediction model according to the predicted core vitrinite reflectance, the historical core vitrinite reflectance and the mean square error function.

[0082] Specifically, the mean square error function can be expressed as: Among them, y i is the true historical core vitrinite reflectance, x i is the input feature vector, is the model's feature vector x for the input i Predicted core vitrinite reflectance.

[0083] In this embodiment, the prediction accuracy of the random forest prediction model is measured by the mean square error function. When the value of the mean square error function is greater than a preset threshold, it can be considered that the prediction accuracy of the random forest prediction model is good, and the training of the random forest prediction model is stopped.

[0084] It should be noted that, in this embodiment, the feature vector obtained by extracting features from historical well logging data according to the adversarial neural network model is used to train the neural network prediction model and the random forest prediction model in sequence. The neural network prediction model and the random forest prediction model can also be trained at the same time, or the random forest prediction model can be trained first, and then the neural network prediction model can be trained. At the same time, when the Ro value is predicted according to only one prediction model, only the prediction model can be trained, and this embodiment does not limit this.

[0085] S290, determining the logging data of the target depth segment, and inputting the logging data into an adversarial neural network model, and outputting a feature vector matching the logging data through a generator network in the adversarial neural network model.

[0086] S2100. Input the feature vector into a neural network prediction model to obtain a neural network model prediction value, and input the feature vector into a random forest prediction model to obtain a random forest model prediction value.

[0087] S2110. Determine the core vitrinite reflectance of the target depth segment based on the predicted value of the neural network model and / or the predicted value of the random forest model.

[0088] S2120: If the core vitrinite reflectance of the target depth segment is determined to be greater than or equal to a preset reflectance threshold, the target depth segment is determined to be a graphitized low-resistance shale development depth segment.

[0089] After training the adversarial neural network model, the neural network prediction model and the random forest prediction model, the feature extraction of the logging data is performed according to the adversarial neural network model, the Ro value is predicted according to the neural network prediction model and the random forest prediction model, the Ro value of the target depth segment is determined according to the predicted Ro value, and the graphitized low-resistance shale is marked according to the Ro value of the target depth segment. This has been explained in the above embodiments and will not be repeated in this embodiment.

[0090] The technical solution of this embodiment performs Ro value correlation analysis on different logging data types, and then performs data preprocessing to solve data redundancy, missing, errors, etc. The data preprocessing can ensure the quality of the data used to meet the subsequent learning model building tasks. Then, a feature combination is selected based on a multi-objective feature selection method to find a set of optimal feature combinations so that the model strikes a balance between prediction accuracy and feature number. Then, data feature extraction is performed based on an adversarial neural network, which includes two neural networks: a generator network and a discriminator network. The generator network is used to learn the potential distribution of logging data and convert the logging data into a high-dimensional feature vector representation, while the discriminator network is used to distinguish between false data and real data generated by the generator network. In order to enhance the feature extraction capability of the generator network, the generator network is combined with a sparse autoencoder, and sparsity constraints are introduced through the sparse autoencoder to further enhance the feature extraction capability of the generator network. Finally, Ro value prediction is performed using a regression model based on a neural network and an improved random forest. It solves the problem of low prediction accuracy of traditional Ro value prediction, has high robustness and generalization ability, can automatically identify the development location, development depth and thickness of graphitized low-resistance shale, greatly improves the evaluation accuracy and efficiency of low-resistance shale, and provides an effective basis for low-resistance shale risk exploration.

[0091] Embodiment 3

[0092] Figure 3 This is a schematic diagram of the structure of a device for determining the reflectance of core vitrinite provided in the third embodiment of the present invention. Figure 3 As shown, the device includes: a feature vector determination module 310, a prediction module 320 and a core vitrinite reflectance determination module 330. Among them:

[0093] A feature vector determination module 310 is used to determine the well logging data of the target depth section, and input the well logging data into the adversarial neural network model, and output a feature vector matching the well logging data through a generator network in the adversarial neural network model;

[0094] A prediction module 320, configured to input the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or input the feature vector into a random forest prediction model to obtain a random forest model prediction value;

[0095] The core vitrinite reflectance determination module 330 is used to determine the core vitrinite reflectance of the target depth segment according to the prediction value of the neural network model and / or the prediction value of the random forest model.

[0096] The technical solution of the embodiment of the present invention determines the core vitrinite reflectance of the target depth section by inputting the logging data of the target depth section into the adversarial neural network, outputting the feature vector through the generator network of the adversarial neural network, and inputting the feature vector into the neural network prediction model and / or the random forest prediction model to obtain the neural network model prediction value and / or the random forest model prediction value. The problem of low accuracy of the method for determining the core vitrinite reflectance in the prior art is solved, and the core vitrinite reflectance of shale rock can be accurately and quickly determined, thereby realizing the automatic identification of graphitized low-resistance shale and providing data support for the risk exploration of low-resistance shale.

[0097] On the basis of the above embodiment, the adversarial neural network model further includes a discriminator network, and the generator network adopts a sparse autoencoder structure;

[0098] The device further comprises:

[0099] A sample data determination module, used to determine historical logging data and historical core vitrinite reflectance;

[0100] An adversarial neural network training module is used to train the adversarial neural network model based on historical logging data, historical core vitrinite reflectance, and adversarial neural network loss function;

[0101] Among them, the adversarial neural network loss function is obtained according to the loss functions of the generator network and the discriminator network and the sparse autoencoder loss function based on the sparsity penalty term.

[0102] Based on the above embodiment, the device further includes:

[0103] A feature vector determination module, used to input historical well logging data into an adversarial neural network model to obtain a feature vector matching the historical well logging data;

[0104] A first core vitrinite reflectance prediction module, used for inputting a feature vector matching the historical well logging data into a neural network prediction model to obtain a predicted core vitrinite reflectance matching the historical well logging data;

[0105] The neural network prediction model training module is used to train the neural network prediction model according to the predicted core vitrinite reflectance, the historical core vitrinite reflectance and the neural network prediction model loss function based on the regularization term.

[0106] Based on the above embodiment, the device further includes:

[0107] A decision tree weight determination module, used to determine at least two decision trees and a decision tree weight of each decision tree according to a feature vector matched with the historical well logging data through a random forest prediction model;

[0108] A feature weight determination module, used to determine the feature weight of each dimension of the feature vector matching the historical well logging data according to the decision tree weight of each decision tree;

[0109] A second core vitrinite reflectance prediction module, used to determine the predicted core vitrinite reflectance matching the historical well logging data according to each decision tree and the feature weights of each dimension of the feature vector matching the historical well logging data;

[0110] The random forest prediction model training module is used to train the random forest prediction model according to the predicted core vitrinite reflectance, the historical core vitrinite reflectance and the mean square error function.

[0111] Based on the above embodiment, the sample data determination module includes:

[0112] A well logging data determination unit, used to determine historical well logging curves and historical core vitrinite reflectance of at least two well logging data types;

[0113] A feature correlation determination unit, used to determine the correlation between each logging data type and the core vitrinite reflectance according to the historical logging curves and the historical core vitrinite reflectance of each logging data type;

[0114] The historical logging data determination unit is used to determine the historical logging data according to the historical logging curves of the logging data type whose correlation meets the preset correlation condition.

[0115] Based on the above embodiment, the historical logging data determination unit is specifically used for:

[0116] Among the well logging data types whose correlations satisfy the preset correlation conditions, determining a combination of well logging data types according to a multi-objective feature selection method;

[0117] The historical logging data is determined according to the historical logging curves of each logging data type in the logging data type combination.

[0118] Based on the above embodiment, the device further includes:

[0119] The graphitization marking module is used to determine that the target depth section is a graphitized low-resistance shale development depth section if the core vitrinite reflectance of the target depth section is greater than or equal to a preset reflectance threshold.

[0120] The device for determining the reflectance of core vitrinite provided in the embodiment of the present invention can execute the method for determining the reflectance of core vitrinite provided in any embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method.

[0121] Embodiment 4

[0122] Figure 4 A schematic diagram of the structure of an electronic device 10 that can be used to implement an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0123] like Figure 4 As shown, the electronic device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0124] A number of components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0125] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a method for determining the reflectance of core vitrinite.

[0126] In some embodiments, the method for determining the core vitrinite reflectance may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the method for determining the core vitrinite reflectance described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the method for determining the core vitrinite reflectance by any other appropriate means (e.g., by means of firmware).

[0127] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0128] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that when the computer program is executed by the processor, the functions / operations specified in the flow chart and / or block diagram are implemented. The computer program may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0129] In the context of the present invention, a computer-readable storage medium may be a tangible medium that may contain or store a computer program for use by or in combination with an instruction execution system, device or equipment. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0130] To provide interaction with a user, the systems and techniques described herein may be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices may also be used to provide interaction with the user; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user may be received in any form (including acoustic input, voice input, or tactile input).

[0131] The systems and techniques described herein may be implemented in a computing system that includes backend components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes frontend components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such backend components, middleware components, or frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0132] A computing system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The client and server relationship is generated by computer programs running on the corresponding computers and having a client-server relationship with each other. The server may be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services.

[0133] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution of the present invention can be achieved, and this document does not limit this.

[0134] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for determining the reflectance of core vitrinite, characterized in that: include: Determine the well logging data of the target depth section, and input the well logging data into the adversarial neural network model, and output a feature vector matching the well logging data through a generator network in the adversarial neural network model; Inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value; The core vitrinite reflectance of the target depth section is determined based on the predicted values ​​of the neural network model and / or the predicted values ​​of the random forest model.

2. The method according to claim 1, characterized in that: The adversarial neural network model also includes a discriminator network, and the generator network adopts a sparse autoencoder structure; Before determining the logging data of the target depth section, it also includes: Determine historical well log data and historical core vitrinite reflectance; The adversarial neural network model is trained based on historical logging data, historical core vitrinite reflectance, and adversarial neural network loss function; Among them, the adversarial neural network loss function is obtained according to the loss functions of the generator network and the discriminator network and the sparse autoencoder loss function based on the sparsity penalty term.

3. The method according to claim 2, characterized in that Before determining the logging data of the target depth section, it also includes: Inputting historical well logging data into an adversarial neural network model to obtain a feature vector matching the historical well logging data; Inputting the feature vector matching the historical well logging data into a neural network prediction model to obtain a predicted core vitrinite reflectance matching the historical well logging data; The neural network prediction model is trained according to the predicted core vitrinite reflectance, historical core vitrinite reflectance and the neural network prediction model loss function based on the regularization term.

4. The method according to claim 2, characterized in that: Before determining the logging data of the target depth section, it also includes: Inputting historical well logging data into an adversarial neural network model to obtain a feature vector matching the historical well logging data; Determine at least two decision trees according to the feature vectors matching the historical well logging data through a random forest prediction model, and determine a decision tree weight of each decision tree; Determining the feature weights of each dimension of the feature vector matching the historical well logging data according to the decision tree weights of each decision tree; Determine the predicted core vitrinite reflectance matching the historical well logging data according to each decision tree and the feature weight of each dimension of the feature vector matching the historical well logging data; The random forest prediction model was trained based on the predicted core vitrinite reflectance, historical core vitrinite reflectance and mean square error function.

5. The method according to claim 2, characterized in that: Determine historical well log data and historical core vitrinite reflectance, including: Determine historical well logs and historical core vitrinite reflectance for at least two well log data types; According to the historical logging curves of each logging data type and the historical core vitrinite reflectance, the correlation between each logging data type and the core vitrinite reflectance is determined; The historical logging data are determined according to the historical logging curves of the logging data types whose correlations meet the preset correlation conditions.

6. The method according to claim 5, characterized in that According to the historical logging curves of the logging data types whose correlation meets the preset correlation conditions, the historical logging data are determined, including: Among the well logging data types whose correlations satisfy the preset correlation conditions, determining a combination of well logging data types according to a multi-objective feature selection method; The historical logging data is determined according to the historical logging curves of each logging data type in the logging data type combination.

7. The method according to claim 1, characterized in that After determining the core vitrinite reflectance of the target depth section, it also includes: If the core vitrinite reflectance of the target depth section is determined to be greater than or equal to a preset reflectance threshold, the target depth section is determined to be a graphitized low-resistance shale development depth section.

8. A device for determining the reflectance of core vitrinite, characterized in that: include: A feature vector determination module is used to determine the well logging data of the target depth section, and input the well logging data into the adversarial neural network model, and output a feature vector matching the well logging data through a generator network in the adversarial neural network model; A prediction module, used for inputting the feature vector into a neural network prediction model to obtain a neural network model prediction value, and / or inputting the feature vector into a random forest prediction model to obtain a random forest model prediction value; The core vitrinite reflectance determination module is used to determine the core vitrinite reflectance of the target depth section according to the prediction value of the neural network model and / or the prediction value of the random forest model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for determining the core vitrinite reflectance as described in any one of claims 1 to 7 is implemented.

10. A storage medium storing computer executable instructions, characterized in that: When the computer executable instructions are executed by a computer processor, they are used to perform the method for determining the core vitrinite reflectance as described in any one of claims 1 to 7.

Images