A reservoir evaluation method, system, device, and storage medium

Abstract

The embodiment of the present disclosure relates to the field of oil and gas reservoir evaluation and geophysical exploration, and discloses a reservoir evaluation method, system, device and storage medium, the method comprising: acquiring and synchronizing logging data and seismic data; preprocessing the logging data and the seismic data; inputting the preprocessed logging data into a machine learning model to output a machine learning model prediction result; inputting the preprocessed seismic data into a deep learning model to output a deep learning model prediction result; obtaining an artificial confidence index of the logging data and the seismic data; and fusing the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index to obtain a reservoir evaluation result. According to the exemplary embodiment of the present disclosure, the logging / seismic data is comprehensively analyzed by using deep learning, machine learning and ensemble learning, and on this basis, an artificial label is added to ensure the accuracy and reliability of the result.

Description

TECHNICAL FIELD

[0001] The embodiment of the present disclosure relates to the field of oil and gas reservoir evaluation and geophysical exploration, in particular to a reservoir evaluation method, system, device and storage medium based on artificial supervision and well-seismic combination multi-modal learning. BACKGROUND

[0002] Traditional reservoir evaluation methods mainly rely on the interpretation of logging and seismic data. However, this method has obvious limitations in dealing with complex reservoirs with high heterogeneity and microscopic fracture characteristics. Although the introduction of artificial intelligence technology provides a new perspective for reservoir evaluation, most of these technologies are still limited to the application of single machine learning models, such as boosting algorithms, neural networks, convolutional networks, and ensemble algorithms, which leads to the proliferation of evaluation models based on single algorithms and their derivative systems. In order to improve the accuracy and objectivity of intelligent reservoir evaluation, appropriate machine learning algorithms and suitable model parameters must be selected, and the optimization of the training process must be ensured to maximize the prediction accuracy of the model and optimize the efficiency.

[0003] However, there are a series of problems in the existing intelligent reservoir evaluation methods. First, these methods tend to use a single data source, such as relying only on logging or seismic data, ignoring the complementarity between different data sources, which may lead to incomplete evaluation results. Second, although machine learning algorithms can automatically process data, the selection of algorithms and the tuning of model parameters still require a lot of manual preprocessing, which not only increases the complexity of the evaluation process, but also limits the generalization ability and application scenarios of the model. In addition, experts' geological understanding and exploration and development experience in specific work blocks, which are crucial for predicting complex reservoirs, are often ignored in intelligent evaluation. SUMMARY

[0004] The embodiment of the present disclosure provides a reservoir evaluation method, system, device and storage medium to solve or alleviate one or more of the above technical problems in the prior art.

[0005] According to one aspect of the present disclosure, a reservoir evaluation method is provided, comprising:

[0006] obtaining and synchronizing logging data and seismic data;

[0007] preprocessing the logging data and seismic data;

[0008] inputting the preprocessed logging data into a machine learning model to output a machine learning model prediction result;

[0009] inputting the preprocessed seismic data into a deep learning model to output a deep learning model prediction result;

[0010] An artificial confidence index of the logging data and the seismic data is acquired;

[0011] The reservoir evaluation result is obtained by fusing the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index.

[0012] In a possible implementation, acquiring and synchronizing the logging data and the seismic data comprises:

[0013] The logging data is resampled to be the same scale as the seismic data.

[0014] In a possible implementation, the preprocessing of the logging data and the seismic data comprises:

[0015] The abnormal points and noises in the logging data and the seismic data are removed;

[0016] The logging data is divided into a logging data training set, a logging data test set and a logging data validation set;

[0017] The seismic data is divided into a seismic data training set, a seismic data test set and a seismic data validation set;

[0018] The classification variables of the logging data and the seismic data are one-hot encoded to convert the classification variables into numerical types;

[0019] The logging data and the seismic data are standardized;

[0020] Missing values in the logging data and the seismic data are filled.

[0021] In a possible implementation, the logging data after the preprocessing is input into a machine learning model, and a machine learning model prediction result is output, comprising:

[0022] A plurality of machine learning base models are selected;

[0023] The plurality of machine learning base models are respectively trained through the logging data training set;

[0024] The accuracy and the F1 score of the plurality of machine learning base models are respectively calculated through the logging data validation set;

[0025] The performance of the plurality of machine learning base models is respectively evaluated through the logging data test set, and the models are optimized according to the evaluation result;

[0026] Based on the accuracy and the F1 score, the prediction results of the plurality of machine learning base models are integrated through a stacking technology to establish a machine learning prediction model;

[0027] The prediction result is output through the machine learning prediction model.

[0028] In a possible implementation, the inputting the preprocessed seismic data into the deep learning model and outputting a deep learning model prediction result comprises:

[0029] selecting a deep learning model according to data requirements;

[0030] training the deep learning model through a seismic data training set and optimizing a training process by using a callback function;

[0031] After the model training is completed, the deep learning model architecture can be visualized, and the deep learning model performance can be evaluated.

[0032] In a possible implementation, the fusing the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index to obtain a reservoir evaluation result comprises:

[0033] converting the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index into numerical data;

[0034] stacking the numerical data;

[0035] training a meta-model by using a support vector machine, and optimizing and evaluating the meta-model accuracy and F1 score.

[0036] According to one aspect of the present disclosure, a reservoir evaluation system is provided, comprising:

[0037] a first acquisition unit configured to acquire and synchronize logging data and seismic data;

[0038] a preprocessing unit configured to preprocess the logging data and the seismic data;

[0039] a first prediction unit configured to input the preprocessed logging data into a machine learning model and output a machine learning model prediction result;

[0040] a second prediction unit configured to input the preprocessed seismic data into a deep learning model and output a deep learning model prediction result;

[0041] a second acquisition unit configured to acquire an artificial confidence index of the logging data and the seismic data;

[0042] a fusion unit configured to fuse the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index to obtain a reservoir evaluation result.

[0043] In a possible implementation, the acquisition unit comprises:

[0044] a resampling module configured to resample the logging data to be the same size as the seismic data.

[0045] In a possible implementation, the preprocessing unit comprises:

[0046] a clearing module configured to clear abnormal points and noise in the well logging data and the seismic data;

[0047] a first division module configured to divide the well logging data into a well logging data training set, a well logging data test set and a well logging data validation set;

[0048] a second division module configured to divide the seismic data into a seismic data training set, a seismic data test set and a seismic data validation set;

[0049] an encoding module configured to perform one-hot encoding on a classification variable of the well logging data and the seismic data, and convert the classification variable into a numerical type;

[0050] a standardization module configured to perform standardization processing on the well logging data and the seismic data;

[0051] a filling module configured to fill in missing values in the well logging data and the seismic data.

[0052] In a possible implementation, the first prediction unit comprises:

[0053] a first selection module configured to select a plurality of machine learning base models;

[0054] a first training module configured to train the plurality of machine learning base models respectively by using the well logging data training set;

[0055] a calculation module configured to calculate accuracy and F1 scores of the plurality of machine learning base models respectively by using the well logging data validation set;

[0056] a first evaluation module configured to evaluate performance of the plurality of machine learning base models respectively by using the well logging data test set, and optimize the models according to the evaluation results;

[0057] an integration module configured to integrate prediction results of the plurality of machine learning base models by using a stacking technology based on the accuracy and the F1 scores, and establish a machine learning prediction model;

[0058] an output module configured to output a prediction result by using the machine learning prediction model.

[0059] In a possible implementation, the inputting of the preprocessed seismic data into the deep learning model and the outputting of a deep learning model prediction result comprise:

[0060] a second selection module configured to select a deep learning model according to a data requirement;

[0061] The second training module is configured to train the deep learning model by using the seismic data training set and optimize the training process by using a callback function.

[0062] The second evaluation module is configured to visualize the deep learning model architecture and evaluate the performance of the deep learning model after the model training is completed.

[0063] In a possible implementation, the reservoir evaluation result is obtained by fusing the machine learning model prediction result, the deep learning model prediction result and the artificial confidence indicator.

[0064] The conversion module is configured to convert the machine learning model prediction result, the deep learning model prediction result and the artificial confidence indicator into numerical data.

[0065] The stacking module is configured to stack the numerical data.

[0066] The optimization module is configured to train the meta-model by using a support vector machine, and optimize and evaluate the accuracy and F1 score of the meta-model.

[0067] According to an aspect of the present disclosure, there is provided a device comprising:

[0068] a processor and a memory;

[0069] The memory is configured to store a computer program, and the processor is configured to invoke the computer program stored in the memory to execute the reservoir evaluation method according to any one of the preceding aspects.

[0070] According to an aspect of the present disclosure, there is provided a computer readable storage medium having a computer program stored therein, and when the computer program is executed by a processor, the processor is enabled to execute the reservoir evaluation method according to any one of the preceding aspects.

[0071] The exemplary embodiments of the present disclosure have the following beneficial effects: The exemplary embodiments of the present disclosure comprehensively analyze well logging / seismic data by using deep learning, machine learning and ensemble learning, and on this basis, add artificial labels to ensure the accuracy and reliability of the results. Finally, the reservoir is intelligently analyzed by using multi-modal learning, thereby improving the accuracy and efficiency of the oil and gas reservoir evaluation under complex geological structures.

[0072] The details of one or more embodiments of the application are set forth in the accompanying drawings and the description below. Other features and advantages of the application will become apparent from the description that follows. It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the application. BRIEF DESCRIPTION OF DRAWINGS

[0073] The accompanying drawings, which are incorporated herein and constitute part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, further serve to explain the principles of the present disclosure. It is apparent that the accompanying drawings are only some embodiments of the present disclosure, and other drawings can be obtained from the following description by those of ordinary skill in the art without creative effort.

[0074] Figure 1 is a flow chart of a reservoir evaluation method of the present exemplary embodiment;

[0075] Figure 2 is an integrated learning structure schematic diagram of the present exemplary embodiment;

[0076] Figure 3 is a multi-layer perceptron learning framework schematic diagram of the present exemplary embodiment;

[0077] Figure 4 is a total model fusion framework schematic diagram of the present exemplary embodiment;

[0078] Figure 5 is a B-3 well prediction diagram of the present exemplary embodiment;

[0079] Figure 6 is a B-4 well fitting result diagram of the present exemplary embodiment;

[0080] Figure 7 is a block diagram of a reservoir evaluation system of the present exemplary embodiment;

[0081] Figure 8 is a structure schematic diagram of an apparatus of the present exemplary embodiment. DETAILED DESCRIPTION

[0082] Example implementations are now described with reference to the drawings. Example implementations can be implemented in any number of manners, and are not limited to the examples described herein; rather, these examples are provided to give a full and enabling disclosure of the present disclosure, and to fully convey the scope of the example implementations to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more implementations. In the following description, numerous specific details are provided to give a thorough understanding of implementations of the present disclosure. One skilled in the relevant art will recognize, however, that the implementations of the present disclosure can be practiced without one or more of the specific details, or with other methods, components, materials, and so forth. In other instances, well-known structures have not been described in order to avoid obscuring the application.

[0083] Further, the accompanying drawings are included to provide a further understanding of the present disclosure and are incorporated in and constitute a part of this specification. The drawings are not necessarily to scale, the dimensions of certain features being exaggerated from others for illustrative purposes. Like reference numerals represent like elements throughout the detailed description. The drawings illustrate exemplary embodiments of the present disclosure and, as such, should not be considered limiting of its scope. The drawings include the following figures:

[0084] The flowcharts shown in the drawings are merely illustrative and do not necessarily include all steps. For example, some steps can be further divided, and some steps can be combined or partially combined, so the actual execution order can be changed according to the actual situation.

[0085] The terms "first", "second", and the like in the specification and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily describe a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0086] In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device that includes a series of steps or sub-modules does not necessarily have to be limited to those steps or sub-modules clearly listed, but can include other steps or sub-modules that are not clearly listed or inherent to these processes, methods, products or devices.

[0087] In order to overcome the limitations in the prior art and improve the accuracy and comprehensiveness of reservoir evaluation, the present embodiment develops a comprehensive intelligent evaluation method that can comprehensively utilize multiple data sources, does not require excessive manual preprocessing, and fully utilizes geological expert knowledge. The present embodiment innovatively improves and integrates existing machine learning techniques, and also pays more attention to comprehensive analysis of data and digitization of geological knowledge during model development, achieving more accurate and comprehensive evaluation of complex reservoirs.

[0088] Figure 1 A flowchart of a reservoir evaluation method according to an exemplary embodiment of the present disclosure is shown in FIG. 1, and the exemplary embodiment of the present disclosure provides a reservoir evaluation method, which includes: Figure 1

[0089] S1 acquiring and synchronizing logging data and seismic data;

[0090] S2 preprocessing the logging data and seismic data;

[0091] ​S3 inputs the preprocessed logging data into the machine learning model, and outputs a machine learning model prediction result;

[0092] S4 inputs the preprocessed seismic data into the deep learning model, and outputs a deep learning model prediction result;

[0093] S5 obtains an artificial confidence index of the logging data and the seismic data;

[0094] S6 fuses the machine learning model prediction result, the deep learning model prediction result, and the artificial confidence index to obtain a reservoir evaluation result.

[0095] To solve the limitations in the traditional reservoir evaluation method, the embodiment proposes an innovative method. In view of the problems that the traditional reservoir interpretation method usually faces strong subjectivity, cannot efficiently process massive data, especially when it needs to combine well-seismic data for comprehensive interpretation, it will face multiple solutions, and the deficiencies in efficiency and batch processing, the embodiment proposes an artificial supervision well-seismic combined multi-modal learning reservoir evaluation method. The embodiment combines machine learning to analyze logging data and deep learning to process seismic data, and comprehensively interprets the reservoir distribution and characteristics. The core innovation lies in using expert geological knowledge and work experience in the work area to artificially calibrate the algorithm prediction, and by fusing the prediction results of the three models, the final model prediction result is accurately corrected, and accurate reservoir evaluation is realized. By integrating machine learning, deep learning, and artificial supervision, the embodiment can improve the efficiency and accuracy of reservoir evaluation, especially in analyzing large-scale complex data sets. The embodiment not only improves the scientificity and reliability of data processing, but also provides a more accurate and efficient technical solution for oil and gas exploration and development.

[0096] Specifically, obtaining and synchronizing logging data and seismic data includes:

[0097] Resampling the logging data to the same scale as the seismic data.

[0098] Specifically, preprocessing the logging data and seismic data includes:

[0099] Removing abnormal points and noise in the logging data and seismic data;

[0100] Dividing the logging data into a logging data training set, a logging data test set, and a logging data validation set;

[0101] Dividing the seismic data into a seismic data training set, a seismic data test set, and a seismic data validation set;

[0102] One-hot encoding the classification variables of the logging data and seismic data to convert the classification variables into numerical types;

[0103] standardizing the logging data and the seismic data;

[0104] filling in missing values in the logging data and the seismic data.

[0105] Specifically, the preprocessed logging data is input into a machine learning model, and the output of the machine learning model prediction result includes:

[0106] selecting a plurality of machine learning base models;

[0107] training the plurality of machine learning base models respectively through a logging data training set;

[0108] calculating the accuracy and F1 score of the plurality of machine learning base models respectively through a logging data validation set;

[0109] evaluating the performance of the plurality of machine learning base models respectively through a logging data test set, and optimizing the model according to the evaluation result;

[0110] integrating the prediction results of the plurality of machine learning base models through stacking technology based on the accuracy and F1 score, and establishing a machine learning prediction model;

[0111] outputting a prediction result through the machine learning prediction model.

[0112] As Figure 2 shown, after systematic investigation and multiple rounds of algorithm testing, the embodiment tests the artificial intelligence algorithm for logging and seismic data. The embodiment is widely applied in three oilfields in a certain block in Africa. By analyzing and applying logging and seismic data, and comprehensively considering the characteristics of the data and processing requirements, the most suitable algorithm combination is determined.

[0113] In order to accurately process logging data, the embodiment tests a plurality of machine learning algorithms. After comprehensive comparison of performance and applicability, SVM, RF and XGBoost show excellent accuracy and reliability in logging data analysis.

[0114] The prediction of SVM is:

[0115]

[0116] In the formula, x represents the feature vector of a single logging data sample. These features can include values extracted from various logging curves (acoustic travel time, density, neutron porosity, natural gamma ray, deep invasion resistivity, porosity, water saturation, permeability, shale content). y i represents the class label of each sample in the training data set - reservoir property, such as (oil, gas, water, dry layer, gas-water layer, etc.).a i ​This represents the Lagrange multiplier determined during SVM training, used to find the optimal boundary. Only data points (i.e., support vectors) located at or near the decision boundary will receive a non-zero 'a'. i Value. K(x) i ,x j () represents a kernel function, used to compute data points x in a higher-dimensional space when the data cannot be effectively separated in the original input space. i x from the training set and new data points j The similarity between the values. b represents the bias term, a parameter of the SVM model used to adjust the decision boundary for better classification of the training data. ι represents the summation from i to ι.

[0117] Based on these characteristics of well logging data, this embodiment uses the RBF kernel to predict the final result, in the form of:

[0118] K(x,z)=exp(-γ‖xz|| 2 );

[0119] In the formula, γ represents the parameter of the kernel function, ||xz|| 2 This represents the square of the Euclidean distance between two points x and z. By adjusting a... i Using b and γ, SVM can find the optimal decision boundary for reservoir classification.

[0120] Random forest prediction for:

[0121]

[0122] In the formula, the sample is x, and each tree is T. i The prediction is T i (x), where I is an indicator function, which has a value of 1 when its internal condition is true and 0 otherwise; C is the category; and K is the total number of decision trees.

[0123] The integrated prediction results are as follows:

[0124]

[0125] Where T is the number of decision trees, and yt is the prediction of the t-th tree.

[0126] The classification process of random forests involves voting on the independent reservoir prediction results of multiple decision trees, and finally selecting the reservoir category with the most votes as the predicted classification for the input sample.

[0127] Xgboost's prediction is:

[0128] Objective function (used in the t-th iteration):

[0129]

[0130] Model prediction (for new input sample x):

[0131]

[0132] Finally, the prediction probability is output by the sigmoid function:

[0133]

[0134] where: l is the loss function. is the objective function including the regularization term. T is the number of leaf nodes. w j is the weight of the leaf node, and γ and λ are regularization parameters. is the original prediction value of the model for the new input sample x, which is the sum of the predictions of all K trees. σ is the sigmoid function, which converts the original prediction value into a probability between 0 and 1.

[0135] The classification process of XGBoost is to add tree models step by step, each tree reduces the residual left by the previous tree, and finally the output of all trees is converted into reservoir classification probability through an activation function (such as Sigmoid or Softmax).

[0136] On this basis, the embodiment applies an ensemble learning method to improve the accuracy and robustness of the prediction, integrates the prediction results of different models through stacking technology, effectively combines the advantages of each model, and creates a more powerful and reliable prediction model. This strategy is particularly important for improving the accuracy of well logging data analysis. If a linear model is used as a meta-model, the mathematical expression can be expressed as:

[0137]

[0138] where: β0, β1,..., β M The parameters are obtained by minimizing the prediction error of the meta-model, and f m (x) is the prediction of the m-th base model on the input feature x.

[0139] Specifically, the preprocessed seismic data is input into the deep learning model, and the deep learning model prediction result is output, which includes:

[0140] Select a deep learning model according to data requirements;

[0141] Train the deep learning model through the seismic data training set, and optimize the training process using a callback function;

[0142] After the model training is completed, the deep learning model architecture can be visualized, and the deep learning model performance can be evaluated.

[0143] like Figure 3 As shown, in earthquake data analysis, the Multilayer Perceptron (MLP) was selected due to its superior performance in training speed and accuracy. This embodiment designs a neural network model with a feedforward structure. Exemplarily, its model structure is as follows: an input layer with 46 neurons, corresponding to the number of features in the earthquake attribute data; three hidden layers, each with 128 neurons, using the ReLU (Rectified Linear Unit) function as the activation function to enhance the model's ability to handle nonlinear problems; and a final output layer that precisely maps to the prediction target (it should be noted that the model structure in this embodiment is only one example and does not limit this disclosure).

[0144] Multilayer perceptron prediction for:

[0145]

[0146] Where: x i The input value is represented by i, which ranges from 1 to n (the number of input features). i Representative and input x i The associated weights, b is the bias term, which provides additional flexibility to the model, and f is the activation function.

[0147] This embodiment innovatively introduces the concept of "human confidence level," aiming to quantify the degree of confidence that humans have in the predicted data. Based on this confidence level, a value between 0 and 1 is assigned, which is then incorporated into the model as a feature. This feature, along with predictions from deep learning and machine learning models, is input as a new feature into a hyperparameter-tuned Support Vector Machine (SVM) model. This method not only combines the advantages of artificial intelligence but also improves the comprehensiveness and accuracy of predictions, opening up new avenues for processing complex geological data.

[0148] Specifically, by integrating the prediction results of the machine learning model, the prediction results of the deep learning model, and the manual confidence index, the reservoir evaluation results include:

[0149] Convert machine learning model predictions, deep learning model predictions, and human confidence metrics into numerical data;

[0150] Stack numerical data;

[0151] The meta-model was trained using support vector machines, and its accuracy and F1 score were optimized and evaluated.

[0152] like Figure 4As shown, the present embodiment adopts an innovative model fusion technology, integrating the prediction outputs of deep learning and machine learning models and the expert's artificial confidence index as input features into an adjusted support vector machine (SVM) model. The design of this meta-learner not only collects the prediction results of different models, but also integrates the expert's subjective evaluation of the data, with the aim of providing a comprehensive evaluation and classification. This fusion strategy realizes comprehensive and accurate prediction of geological characteristics by combining the advantages of multiple models and expert evaluation.

[0153] One embodiment of the present disclosure includes the following steps:

[0154] S10 Data collection and synchronization:

[0155] Collect data: Obtain necessary well logging and seismic data.

[0156] Data synchronization: Resample the logging data to ensure consistency with the time or depth scale of the seismic data, synchronize the seismic and well location data using software such as Petrel / GeoGraphix, and extract seismic attributes such as reflection intensity and wave velocity.

[0157] S20 Data preprocessing:

[0158] Cleaning and segmentation: Identify and remove outliers, reduce noise interference, and divide the data set into training and test sets.

[0159] Data encoding and standardization: One-hot encoding for categorical variables, standardization to ensure consistent variable dimensions, missing value imputation, and noise addition or dimensionality reduction as needed.

[0160] S30 Machine learning model training:

[0161] Model selection and training: Use algorithms such as SVC, RandomForest, XGBClassifier, and perform parameter optimization through Grid / RandomSearchCV.

[0162] Performance evaluation: Calculate the accuracy and F1 score of the model on the validation set.

[0163] S40 Deep learning model training:

[0164] Model construction: Select MLP, CNN, or LSTM models based on data requirements, and use callback functions to optimize the training process.

[0165] Performance visualization: After model training is complete, visualize the model architecture and training history to evaluate model performance.

[0166] S50 Application of artificial confidence:

[0167] Confidence score: Geological experts score the data based on their experience in the work area, and these scores are combined with the original data as features or weights.

[0168] S60 Model fusion:

[0169] Prediction result preparation: Ensure that there are prediction results from deep learning and machine learning models and artificial confidence, and convert the results to numerical data for stacking.

[0170] Fusion training: Train the meta-model using support vector machines, optimize and evaluate accuracy and F1 score.

[0171] S70 Result verification and application:

[0172] Model verification: Verify the accuracy and robustness of the final model using the test set, and form the reservoir property prediction.

[0173] According to Figure 5 ( Figure 5 In this embodiment, "new" means artificial intelligence prediction results, "old" means well logging interpretation results, DT means acoustic time difference; RHOB means density; NPHI means neutron porosity; GR means natural gamma ray; ILD means deep invasion resistivity; POR means porosity; SW means water saturation; PERM means permeability; VSH means shale content; yellow stripes represent poor gas layers; yellow represents gas layers; white represents dry layers; blue represents water layers; blue and yellow represent gas and water layers) The analysis results show that the method used in this embodiment is highly consistent with the artificial prediction results in most intervals. Except in the 1680-1690 interval, especially in the prediction of poor gas layers and gas-water layers, this embodiment has a little deviation from the artificial prediction.

[0174] According to Figure 6 ( Figure 6 The schematic diagram of the perfect fitting result of well B-4) Comparison analysis shows that the results obtained by the artificial intelligence prediction technology used in this embodiment are basically consistent with the results obtained by the traditional prediction method. Through experimental verification in multiple blocks, this embodiment shows its efficiency and accuracy in reservoir distribution prediction, effectively supporting the accurate division of reservoirs. In addition, this method has the advantages of batch processing of data while ensuring the accuracy of prediction, which significantly optimizes the work flow, saves time cost, and improves the overall work efficiency. This achievement not only proves the application value of this embodiment in the field of geological exploration and evaluation, but also provides reliable method support for the formulation of future reservoir management and development strategies.

[0175] Figure 7 is a block diagram of a reservoir evaluation system of the present exemplary embodiment, as shown in Figure 7As shown, the exemplary embodiments of the present disclosure provide a reservoir evaluation system, comprising: a first acquisition unit configured to acquire and synchronize logging data and seismic data;

[0176] a first acquisition unit 10 configured to acquire and synchronize logging data and seismic data;

[0177] a preprocessing unit 20 configured to preprocess the logging data and the seismic data;

[0178] a first prediction unit 30 configured to input the preprocessed logging data into a machine learning model and output a machine learning model prediction result;

[0179] a second prediction unit 40 configured to input the preprocessed seismic data into a deep learning model and output a deep learning model prediction result;

[0180] a second acquisition unit 50 configured to acquire an artificial confidence index of the logging data and the seismic data;

[0181] a fusion unit 60 configured to fuse the machine learning model prediction result, the deep learning model prediction result and the artificial confidence index to obtain a reservoir evaluation result.

[0182] Specifically, the acquisition unit comprises:

[0183] a resampling module configured to resample the logging data to the same scale as the seismic data.

[0184] Specifically, the preprocessing unit comprises:

[0185] a cleaning module configured to clean abnormal points and noise in the logging data and the seismic data;

[0186] a first division module configured to divide the logging data into a logging data training set, a logging data test set and a logging data validation set;

[0187] a second division module configured to divide the seismic data into a seismic data training set, a seismic data test set and a seismic data validation set;

[0188] an encoding module configured to one-hot encode the classification variables of the logging data and the seismic data, and convert the classification variables into numerical types;

[0189] a standardization module configured to standardize the logging data and the seismic data;

[0190] a filling module configured to fill in missing values in the logging data and the seismic data.

[0191] Specifically, the first prediction unit comprises:

[0192] The first selection module is configured to select a plurality of machine learning base models;

[0193] The first training module is configured to train the plurality of machine learning base models respectively by using a well logging data training set;

[0194] The calculation module is configured to calculate the accuracy and F1 score of the plurality of machine learning base models respectively by using a well logging data validation set;

[0195] The first evaluation module is configured to evaluate the performance of the plurality of machine learning base models respectively by using a well logging data test set, and optimize the models according to the evaluation results;

[0196] The integration module is configured to integrate the prediction results of the plurality of machine learning base models by using a stacking technology based on the accuracy and F1 score, and establish a machine learning prediction model;

[0197] The output module is configured to output the prediction results by using the machine learning prediction model.

[0198] Specifically, the inputting of the preprocessed seismic data into the deep learning model and the outputting of the deep learning model prediction results comprise:

[0199] The second selection module is configured to select a deep learning model according to data requirements;

[0200] The second training module is configured to train the deep learning model by using a seismic data training set, and optimize the training process by using a callback function;

[0201] The second evaluation module is configured to visualize the deep learning model architecture after the model training is completed, and evaluate the performance of the deep learning model.

[0202] Specifically, the fusion of the machine learning model prediction results, the deep learning model prediction results and the artificial confidence index to obtain the reservoir evaluation results comprises:

[0203] The conversion module is configured to convert the machine learning model prediction results, the deep learning model prediction results and the artificial confidence index into numerical data;

[0204] The stacking module is configured to stack the numerical data;

[0205] The optimization module is configured to train a meta-model by using a support vector machine, and optimize and evaluate the accuracy and F1 score of the meta-model.

[0206] Figure 8 is a structural schematic diagram of a device of the present exemplary embodiment. As shown in Figure 8As shown, the present disclosure also provides a device corresponding to the reservoir evaluation method provided above. Since the embodiments of the device are similar to the above-mentioned method embodiments, they are described relatively simply, and the relevant parts are described in the above-mentioned method embodiment part. The device described below is only illustrative. The device can include a processor 1, a memory 2, a communication bus (i.e. the above-mentioned device bus), and a lookup engine. The processor 1 and the memory 2 can communicate with each other through the communication bus and communicate with the outside through the communication interface. The processor 1 can call the logic instructions in the memory 2 to execute the reservoir evaluation method.

[0207] In addition, the logic instructions in the above-mentioned memory 2 can be realized in the form of a software function unit and sold or used as an independent product. Based on such understanding, the technical solutions of the present disclosure essentially or the part that contributes to the prior art or part of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, including a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the embodiments of the present disclosure. The foregoing storage medium includes a storage chip, a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0208] On the other hand, the embodiments of the present disclosure also provide a processor-readable storage medium, and the processor-readable storage medium stores a computer program 3. When the computer program 3 is executed by the processor 1, the reservoir evaluation method provided by the above-mentioned embodiments is realized.

[0209] The processor-readable storage medium can be any available medium or data storage device that the processor 1 can access, including but not limited to a magnetic storage (such as a floppy disk, a hard disk, a magnetic tape, a magneto-optical disk (MO), etc.), an optical storage (such as a CD, a DVD, a BD, a HVD, etc.), and a semiconductor storage (such as a ROM, an EPROM, an EEPROM, a non-volatile memory (NAND FLASH), a solid state disk (SSD)), etc.

[0210] The above is only the preferred embodiment of the present disclosure, and the protection scope of the present disclosure is not limited to the above-mentioned embodiments. Any technical solution that belongs to the technical solutions of the present disclosure should be considered as the protection scope of the present disclosure. It should be noted that some improvements and refinements without departing from the principles of the present disclosure should be considered as the protection scope of the present disclosure.

Claims

1. A reservoir evaluation method, characterized in that, include: Acquire and synchronize well logging data and seismic data; The well logging data and seismic data are preprocessed; The preprocessed well logging data is input into the machine learning model, and the machine learning model prediction results are output. The preprocessed seismic data is input into the deep learning model, and the deep learning model's prediction results are output. Human confidence index for acquiring well logging and seismic data; By integrating the prediction results of the machine learning model, the prediction results of the deep learning model, and the human confidence index, the reservoir evaluation results are obtained. Acquiring and synchronizing well logging and seismic data includes: Resample the well logging data to the same scale as the seismic data; By integrating the prediction results of the machine learning model, the prediction results of the deep learning model, and the manual confidence index, the reservoir evaluation results are obtained, including: Convert machine learning model predictions, deep learning model predictions, and human confidence metrics into numerical data; Stack numerical data; The meta-model was trained using support vector machines, and its accuracy and F1 score were optimized and evaluated.

2. The reservoir evaluation method according to claim 1, characterized in that, Preprocessing of the well logging and seismic data includes: Remove outliers and noise from the well logging and seismic data; The well logging data is divided into a well logging data training set, a well logging data test set, and a well logging data validation set. The earthquake data is divided into an earthquake data training set, an earthquake data test set, and an earthquake data validation set. The categorical variables of the well logging data and seismic data are one-hot encoded to convert the categorical variables into numerical values. The well logging data and seismic data are standardized. Fill in the missing values ​​in the well logging and seismic data.

3. The reservoir evaluation method according to claim 2, characterized in that, The preprocessed well logging data is input into the machine learning model, and the output machine learning model prediction results include: Choose multiple machine learning base models; Multiple machine learning base models were trained using the well logging data training set. The accuracy and F1 score of multiple machine learning base models were calculated using a well logging data validation set. The performance of multiple machine learning base models was evaluated using a well logging data test set, and the models were optimized based on the evaluation results. Based on accuracy and F1 score, the prediction results of multiple machine learning base models are integrated through stacking technology to establish a machine learning prediction model; The prediction results are output through the machine learning prediction model.

4. The reservoir evaluation method according to claim 2, characterized in that, The process of inputting the preprocessed seismic data into the deep learning model and outputting the deep learning model's prediction results includes: Select a deep learning model based on data requirements; The deep learning model is trained using a seismic data training set, and the training process is optimized using a callback function. After the model is trained, visualize the deep learning model architecture and evaluate the performance of the deep learning model.

5. A reservoir evaluation system, characterized in that, include: The first acquisition unit is used to acquire and synchronize well logging data and seismic data; A preprocessing unit is used to preprocess the well logging data and seismic data; The first prediction unit is used to input the preprocessed well logging data into the machine learning model and output the prediction results of the machine learning model. The second prediction unit is used to input the preprocessed seismic data into the deep learning model and output the prediction results of the deep learning model. The second acquisition unit is used to acquire artificial confidence indices for well logging data and seismic data; The fusion unit is used to fuse the prediction results of the machine learning model, the prediction results of the deep learning model, and the human confidence index to obtain the reservoir evaluation results. The acquisition unit includes: The resampling module is used to resample well logging data to the same scale as seismic data; By integrating the prediction results of the machine learning model, the prediction results of the deep learning model, and the manual confidence index, the reservoir evaluation results are obtained, including: The conversion module is used to convert machine learning model predictions, deep learning model predictions, and human confidence metrics into numerical data. The stacking module is used to stack numerical data. The optimization module is used to train a meta-model using a support vector machine, optimize and evaluate the meta-model's accuracy and F1 score.

6. The reservoir evaluation system according to claim 5, characterized in that, The preprocessing unit includes: The cleaning module is used to remove outliers and noise from the well logging data and seismic data. The first partitioning module is used to divide the logging data into a logging data training set, a logging data test set, and a logging data validation set. The second partitioning module is used to divide the earthquake data into an earthquake data training set, an earthquake data test set, and an earthquake data validation set. The encoding module is used to perform one-hot encoding on the categorical variables of the well logging data and seismic data, converting the categorical variables into numerical values; A standardization module is used to standardize the well logging data and seismic data; The filling module is used to fill in missing values ​​in the well logging data and seismic data.

7. The reservoir evaluation system according to claim 6, characterized in that, The first prediction unit includes: The first selection module is used to select multiple machine learning base models; The first training module is used to train multiple machine learning base models using well logging data training sets. The calculation module is used to calculate the accuracy and F1 score of multiple machine learning base models using a well logging data validation set. The first evaluation module is used to evaluate the performance of multiple machine learning base models using a well logging data test set, and optimize the models based on the evaluation results. The integration module is used to integrate the prediction results of multiple machine learning base models based on accuracy and F1 score through stacking technology to build a machine learning prediction model. The output module is used to output the prediction results through the machine learning prediction model.

8. The reservoir evaluation system according to claim 6, characterized in that, The process of inputting the preprocessed seismic data into the deep learning model and outputting the deep learning model's prediction results includes: The second selection module is used to select a deep learning model based on data requirements; The second training module is used to train the deep learning model using the earthquake data training set and to optimize the training process using a callback function. The second evaluation module is used to visualize the deep learning model architecture and evaluate the performance of the deep learning model after training is completed.

9. A device, characterized in that, include: Processor and memory; The memory is used to store computer programs, and the processor calls the computer programs stored in the memory to execute the reservoir evaluation method according to any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the processor to perform the reservoir evaluation method according to any one of claims 1 to 4.

Images