Shale oil geological sweet spot prediction method based on multiple parameters

Abstract

The invention discloses a shale oil geological sweet spot prediction method based on multiple parameters, and belongs to the field of unconventional oil and gas development, and the method comprises the steps: obtaining well logging data of a known well, carrying out the preprocessing of the well logging data, and obtaining the preprocessed well logging data; constructing a deep neural network model, inputting the preprocessed logging data into the model for training, and obtaining a trained deep network; applying the trained deep network to a target well based on a transfer learning method to generate a transfer model; obtaining target well data, inputting the target well data into the migration model for prediction, and obtaining related parameters; constructing a geological dessert factor model by using related parameters; geological sweet spot factors are obtained based on the geological sweet spot factor model, and shale oil geological sweet spots are divided into different grades based on the values of the geological sweet spot factors. According to the method, dependence on a large amount of logging data is reduced, prediction precision is improved, and technical support is provided for effective development of unconventional oil and gas resources.

Description

Technical Field

[0001] The present invention belongs to the field of unconventional oil and gas development, and in particular relates to a shale oil geological sweet spot prediction method based on multiple parameters. Background Art

[0002] As an important part of the global energy structure, the development and utilization of shale oil has great strategic significance for coping with the global energy crisis and achieving energy diversification. With the gradual depletion of traditional oil and gas resources, the exploration and development of shale oil has become a hot topic in the international energy field. For a large energy consumer like China, actively developing shale oil resources is not only an effective way to meet the growing domestic energy demand, but also an important measure to reduce dependence on external energy and ensure national energy security. China is also making continuous progress in shale oil development technology. By adopting advanced exploration technology and efficient development models, the development cost of shale oil in China is gradually decreasing and the development efficiency is constantly improving. These technological advances have laid a solid foundation for the large-scale development of shale oil in China. However, the development of shale oil still faces many challenges, including the complexity of geological conditions, the pressure of environmental protection, and the replacement of technical equipment. In order to achieve the sustainable development of shale oil, China needs to continue to increase investment in scientific and technological research and development, promote technological innovation, strengthen international cooperation, introduce and absorb international advanced technologies, and improve the overall level of domestic shale oil development.

[0003] As an important unconventional oil and gas resource, reservoir evaluation of shale oil is the key to its exploration and development. Parameters such as porosity, total organic carbon content, clay mineral content, and free hydrocarbon content are key parameters for reservoir evaluation, and their accurate prediction is crucial for the effective development of shale oil. These parameters not only affect the storage capacity and production of shale oil, but are also directly related to the economic benefits and environmental impact of mining. Porosity determines the size of the shale oil storage space, the total organic carbon content reflects the richness of organic matter in the shale, the clay mineral content affects the mineral composition and storage characteristics of the shale, and the free hydrocarbon content is directly related to the recoverable amount of shale oil. However, traditional geological sweet spot prediction methods rely on a large amount of core data, which is costly and inefficient. In order to reduce the cost of coring, new methods are urgently needed to achieve this goal. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a shale oil geological sweet spot prediction method based on multiple parameters, comprising:

[0005] Acquire well logging data of a known well, and preprocess the well logging data to obtain preprocessed well logging data;

[0006] Constructing a deep neural network model, inputting the preprocessed logging data into the deep neural network model for training, and obtaining a trained deep network;

[0007] Applying the trained deep network to a target well based on a transfer learning method, and fine-tuning using a small amount of well logging data of the target well to generate a transfer model;

[0008] Acquire target well data, input the target well data into the migration model for prediction, and obtain relevant parameters;

[0009] Constructing a geological sweet spot factor model, obtaining a geological sweet spot factor based on the relevant parameters and the geological sweet spot factor model, and dividing shale oil geological sweet spots into different levels based on the value of the geological sweet spot factor;

[0010] Based on different levels of geological sweet spots, the geological sweet spot model is drawn using geological modeling software to obtain the vertical distribution characteristics of the sweet spot.

[0011] Preferably, the process of obtaining pre-processed logging data includes:

[0012] Processing missing values ​​on the well logging data of the known well to obtain data after missing value processing;

[0013] The data after the missing value processing is subjected to outlier processing to obtain processed well logging data.

[0014] Preferably, the process of obtaining the trained deep network includes:

[0015] Based on the Pearson correlation coefficient method and the pre-processed logging data, the logging parameters are obtained, and the logging parameters are input into the deep neural network model as a training set for training to obtain the trained deep network.

[0016] Preferably, the deep neural network model includes a network architecture design module and a model training module;

[0017] The network architecture design module is used to design the architecture of a deep neural network, including an input layer, a hidden layer, and an output layer;

[0018] The model training module is used to forward propagate the input data through the neurons of each layer during the training process, calculate the predicted value of the output layer, and update the weights and bias of the network through the back-propagation algorithm according to the gradient of the loss function, so that the network can continuously reduce the error between the predicted value and the actual value.

[0019] Preferably, the process of generating the migration model includes:

[0020] Obtain well logging data and porosity, total organic carbon content, clay mineral content, and free hydrocarbon content labels for target wells;

[0021] Freezing some layers of the trained deep network, fine-tuning the frozen deep network using a small amount of data from the target well, then inputting the target well data, performing forward propagation through the network, and calculating a predicted value;

[0022] According to the gradient of the loss function, back propagation is performed through the network, the weights of the unfrozen layers are updated, and then the network is saved to obtain the migration model.

[0023] Preferably, the process of obtaining relevant parameters includes:

[0024] The logging data of the target well is obtained, and the logging data of the target well is input into the migration model for forward propagation calculation to obtain the predicted values ​​of porosity, total organic carbon content, clay mineral content and free hydrocarbon content.

[0025] Preferably, the process of classifying shale oil geological sweet spots into different levels based on the value of the geological sweet spot factor comprises:

[0026] Calculating a geological sweet spot factor based on the predicted values ​​of the porosity, total organic carbon content, clay mineral content and free hydrocarbon content and the geological sweet spot factor model;

[0027] According to the cumulative frequency of the size of the geological sweet spot factor, shale oil geological sweet spots are divided into different levels.

[0028] Compared with the prior art, the present invention has the following advantages and technical effects:

[0029] The present invention selects logging parameters suitable for deep neural networks as input, and obtains a deep neural network model that can well predict source well data after training; then the deep neural network of the source well is used as a pre-trained network, and a small amount of data is used to fine-tune it on the target well to obtain a migration deep neural network model that predicts the reservoir porosity, total organic carbon content, clay mineral content, and free hydrocarbon content of the target well; then the geological sweet spot factor model is used to calculate the geological sweet spot factor, and then the sweet spot is divided into different levels; finally, the geological sweet spot model is drawn using geological modeling software. The present invention proposes an innovative method for obtaining shale reservoir parameters and modeling geological sweet spots. Compared with traditional technologies, the present invention can accurately predict geological sweet spots based on limited logging data, significantly reducing the cost of oil and gas exploration and development, and has important practical value. Through deep transfer learning technology, the present invention not only reduces the dependence on a large amount of logging data, but also improves the prediction accuracy, providing technical support for the effective development of unconventional oil and gas resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0031] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;

[0032] Figure 2 A structural diagram of a deep neural network according to an embodiment of the present invention;

[0033] Figure 3 This is a structural diagram of a migration deep neural network according to an embodiment of the present invention. DETAILED DESCRIPTION

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0036] Embodiment 1

[0037] like Figure 1 As shown, this embodiment provides a shale oil geological sweet spot prediction method based on multiple parameters, including:

[0038] Acquire all well logging data of known wells, and pre-process the well logging data to obtain processed well logging data;

[0039] Using the Pearson correlation coefficient method to select the required well logging data, inputting the selected well logging data into a deep neural network, and constructing a deep neural network model;

[0040] Using transfer learning, the deep neural network model is applied to the target well and fine-tuned using a small amount of well logging data from the target well;

[0041] The adjusted model is used to predict the porosity, total organic carbon content, brittle mineral content and free hydrocarbon content of the target well to obtain specific values;

[0042] Using the geological sweet spot factor model and combining the related parameters, the geological sweet spot factor is calculated, and the geological sweet spots are divided into different levels according to the value of the geological sweet spot factor;

[0043] Using geological modeling software, based on the geological sweet spot factors, a geological sweet spot model is drawn to obtain the vertical distribution characteristics of the sweet spot.

[0044] Furthermore, obtaining all logging data of known wells includes:

[0045] Obtain all relevant logging parameter data of 5 known wells;

[0046] Further, preprocessing the well logging data includes:

[0047] The well logging data are processed for missing values ​​in turn. There are two processing strategies for missing values: when all well logging attribute values ​​are missing or more than 50% are missing, the entire well logging attribute is deleted; when a small number of well logging attribute values ​​are missing, the mean value of the attribute is used to fill the missing values.

[0048] The data after the missing value processing is subjected to outlier processing, and the outliers are directly eliminated to obtain processed logging data.

[0049] Furthermore, the process of building a deep neural network model includes:

[0050] The Pearson correlation coefficient method is used to select AC, DEN, DTC, GR, and PE, a total of five logging attributes as the input of the deep neural network. The calculation model of the Pearson correlation coefficient is:

[0051]

[0052] r(X,Y): Correlation coefficient between two variables

[0053] Var(): standard deviation of a variable

[0054] cov(X,Y): covariance between two variables

[0055] Designing deep neural network architectures, such as Figure 2 As shown, it includes an input layer, a hidden layer and an output layer. The number of hidden layers is set to 3, and the number of neuron nodes in each first, second and third hidden layer is 12, 12 and 5 respectively.

[0056] The selected data is input into the deep neural network to train the network model.

[0057] Furthermore, using transfer learning, the pre-trained model is applied to the target well. The process of fine-tuning using a small amount of well logging data from the target well includes:

[0058] like Figure 3As shown, the neural network input layer, the first and second hidden layers are frozen, and the third hidden layer and the output layer parameters are trained, so that the parameters are updated to suit the geological conditions of the target well;

[0059] The process of building the deep neural network model for the remaining four known wells is the same as the above process. By repeating the above process, the deep neural network models for the five known wells can be obtained. Based on the deep neural network models of the five known wells, the relevant logging parameters of one unknown well can be obtained by using transfer learning. By repeating the above deep transfer learning process, other key reservoir parameters of the 15 target wells can be well predicted, such as porosity, total organic carbon content, brittle mineral content and free hydrocarbon content.

[0060] Furthermore, the geological sweet spot factor is calculated by using the geological sweet spot factor model in combination with the related parameters. According to the value of the geological sweet spot factor, the geological sweet spots are divided into different levels, including:

[0061] The shale oil sweet spot factor is the product of four parameters: reservoir performance, oil content, brittleness, and shale oil mobility. Porosity is selected as a parameter to reflect reservoir performance, total organic carbon content is selected as a parameter to evaluate reservoir oil content, brittle mineral content is used to evaluate reservoir brittleness, and the difference between free hydrocarbon content and 100 times total organic carbon content is used to determine the amount of movable oil. The sweet spot factor model is:

[0062] U=U1U2U3U4

[0063] U1=(φ-φ min ) / (φ max -φ min )

[0064] U2=(S1-S 1min ) / (S 1max -S 1min )

[0065] U3=(EE min ) / (E max -E min )

[0066] U4=(S lm -S lm min ) / (S lm max -S lm min )

[0067] U, U1, U2, U3, U4: shale oil sweet spot factor, reservoir performance factor, reservoir oil content factor, reservoir brittleness factor and shale oil mobility factor, respectively;

[0068] φ,φ max ,φ min : are reservoir porosity, maximum porosity and minimum porosity, respectively, %;

[0069] S1,S 1max ,S 1min : are the free hydrocarbon content, maximum free hydrocarbon content and minimum free hydrocarbon content of the reservoir, mg / g;

[0070] E,E max ,E min : are reservoir brittleness index, maximum brittleness index and minimum brittleness index respectively;

[0071] S 1m ,S 1mmax ,S 1mmin : Reservoir movable oil content, maximum movable oil content and minimum movable oil content, mg / g;

[0072] After calculating the geological sweet spot factor, the shale oil sweet spot factor with a cumulative frequency greater than 80% is recorded as a Class I shale oil geological sweet spot, 60%-80% is recorded as a Class II shale oil geological sweet spot, and less than 60% is recorded as a Class III shale oil geological sweet spot. Among them, the Class I shale oil geological sweet spot is the most favorable sweet spot.

[0073] Furthermore, using geological modeling software, a predicted cross-section is drawn to visually display the distribution of the sweet spot;

[0074] Analyze the predicted profile to obtain the vertical distribution characteristics of the sweet spot. This includes the continuity, thickness, location, etc. of the sweet spot. By analyzing parameters such as porosity, permeability, and gas saturation, the enrichment area of ​​the high-quality "sweet spot" reservoir is determined. In this way, areas with higher sweet spot factor values ​​can be identified, which may have better reservoir performance and oil content.

[0075] The technical effects of this embodiment are:

[0076] This embodiment proposes a new machine learning concept - deep transfer learning, and proposes a new geological sweet spot optimization method based on this concept. By using the Pearson correlation coefficient method to optimize the logging parameters suitable for the deep neural network as input, a deep neural network model that can well predict the source well data is obtained after training; then the deep neural network of the source well is used as a pre-trained network, and a small amount of data is used to fine-tune it on the target well to obtain a migration deep neural network model that predicts the reservoir porosity, total organic carbon content, clay mineral content, and free hydrocarbon content of the target well; then the geological sweet spot factor model is used to calculate the geological sweet spot factor, and then the sweet spot is divided into different levels; finally, the geological modeling software is used to draw the geological sweet spot model. The present invention proposes an innovative shale reservoir parameter acquisition and geological sweet spot modeling method. Compared with traditional technologies, the present invention can accurately predict geological sweet spots based on limited logging data, significantly reduce the cost of oil and gas exploration and development, and has important practical value. Through deep transfer learning technology, the present invention not only reduces the dependence on a large amount of logging data, but also improves the prediction accuracy, providing technical support for the effective development of unconventional oil and gas resources.

[0077] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A shale oil geological sweet spot prediction method based on multiple parameters, characterized in that: include: Acquire well logging data of a known well, and preprocess the well logging data to obtain preprocessed well logging data; Constructing a deep neural network model, inputting the preprocessed logging data into the deep neural network model for training, and obtaining a trained deep network; Applying the trained deep network to a target well based on a transfer learning method, and fine-tuning using a small amount of well logging data of the target well to generate a transfer model; Acquire target well data, input the target well data into the migration model for prediction, and obtain relevant parameters; Constructing a geological sweet spot factor model, obtaining a geological sweet spot factor based on the relevant parameters and the geological sweet spot factor model, and dividing shale oil geological sweet spots into different levels based on the value of the geological sweet spot factor; Based on different levels of geological sweet spots, the geological sweet spot model is drawn using geological modeling software to obtain the vertical distribution characteristics of the sweet spot.

2. The method according to claim 1, characterized in that The process of obtaining pre-processed logging data includes: Processing missing values ​​on the well logging data of the known well to obtain data after missing value processing; The data after the missing value processing is subjected to outlier processing to obtain processed well logging data.

3. The method according to claim 1, characterized in that The process of obtaining the trained deep network includes: Based on the Pearson correlation coefficient method and the pre-processed logging data, the logging parameters are obtained, and the logging parameters are input into the deep neural network model as a training set for training to obtain the trained deep network.

4. The method according to claim 3, characterized in that: The deep neural network model includes a network architecture design module and a model training module; The network architecture design module is used to design the architecture of a deep neural network, including an input layer, a hidden layer, and an output layer; The model training module is used to forward propagate the input data through the neurons of each layer during the training process, calculate the predicted value of the output layer, and update the weights and bias of the network through the back-propagation algorithm according to the gradient of the loss function, so that the network can continuously reduce the error between the predicted value and the actual value.

5. The method according to claim 1, characterized in that The process of generating the migration model includes: Obtain well logging data and porosity, total organic carbon content, clay mineral content, and free hydrocarbon content labels for target wells; Freezing some layers of the trained deep network, fine-tuning the frozen deep network using a small amount of data from the target well, then inputting the target well data, performing forward propagation through the network, and calculating a predicted value; According to the gradient of the loss function, back propagation is performed through the network, the weights of the unfrozen layers are updated, and then the network is saved to obtain the migration model.

6. The method according to claim 1, characterized in that The process of obtaining relevant parameters includes: The logging data of the target well is obtained, and the logging data of the target well is input into the migration model for forward propagation calculation to obtain the predicted values ​​of porosity, total organic carbon content, clay mineral content and free hydrocarbon content.

7. The method according to claim 6, characterized in that The process of classifying shale oil geological sweet spots into different levels based on the value of the geological sweet spot factor includes: Calculating a geological sweet spot factor based on the predicted values ​​of the porosity, total organic carbon content, clay mineral content and free hydrocarbon content and the geological sweet spot factor model; According to the cumulative frequency of the size of the geological sweet spot factor, the shale oil geological sweet spots are divided into different levels.

Images