An artificial intelligence-based oil exploration method and system

Abstract

The application belongs to the technical field of artificial intelligence, and discloses a petroleum exploration method and system based on artificial intelligence, which comprises the following steps: collecting seismic data, well logging data, exploration environment data and international crude oil price data; performing preliminary processing based on the seismic data and the well logging data to obtain a preliminary data set; performing feature extraction based on the preliminary data set by using an improved PCA algorithm to obtain a new feature data set; predicting petroleum thickness data, porosity data and permeability data by using a DNN model based on the new feature data set; predicting petroleum productivity data by using a regularized exploration environment regression model based on the predicted petroleum thickness data, porosity data and permeability data, and setting a productivity threshold value; and comparing the predicted petroleum productivity data with the productivity threshold value to determine whether to exploit petroleum; the method can accurately predict key parameters in petroleum exploration, improve the efficiency and accuracy of petroleum exploration, and provide scientific and reliable exploitation judgment basis.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence, in particular to an oil exploration method and system based on artificial intelligence. BACKGROUND

[0002] Now there are many problems in the existing oil exploration method and system. First, the existing feature extraction method cannot effectively extract the deep information hidden in the data, especially when facing complex nonlinear relationship, which affects the prediction accuracy of the model and the reliability of the result.

[0003] Secondly, when predicting oil production capacity, it usually relies on simplified physical models or static analysis based on historical data, ignoring the actual impact of dynamic changes of exploration environment on oil production capacity. Changes in geological and environmental conditions can have a significant impact on reservoir permeability, yield, etc., but existing models often fail to consider these changing factors, resulting in deviation and inaccuracy in production capacity prediction.

[0004] In addition, the existing oil production threshold calculation method is often based on fixed historical data and empirical formula, lacking flexible dynamic adjustment mechanism. Once there is a large change in the market or environment, such as a sharp drop or rise in oil prices, the existing method cannot adjust the production threshold in time, which may affect oil exploitation decisions.

[0005] In view of this, the present application proposes an oil exploration method and system based on artificial intelligence to solve the above problems. SUMMARY

[0006] In order to overcome the above-mentioned defects of the prior art, in order to achieve the above-mentioned purpose, the present application provides the following technical scheme, an oil exploration method based on artificial intelligence, comprising:

[0007] Step S1, collecting seismic data, well logging data, exploration environment data and international crude oil price data;

[0008] Step S2, based on the seismic data and well logging data, performing preliminary processing to obtain a preliminary data set; based on the preliminary data set, using an improved PCA algorithm to extract features to obtain a new feature data set;

[0009] Step S3, based on the new feature data set, using a DNN model to predict oil thickness data, porosity data and permeability data;

[0010] Step S4, based on the predicted oil thickness data, porosity data and permeability data, using a regularized exploration environment regression model to predict oil production capacity data, and setting a production capacity threshold, comparing the predicted oil production capacity data and the production capacity threshold, and determining whether to exploit oil.

[0011] Further, the seismic data comprises amplitude characteristic data, wave velocity characteristic data, radiation characteristic data, seismic wave frequency characteristic data and magnetic force characteristic data.

[0012] The well logging data comprises resistivity characteristic data, gamma ray characteristic data, acoustic travel time characteristic data, rock density characteristic data, neutron porosity characteristic data, conductivity characteristic data and natural gamma characteristic data.

[0013] The exploration environment data comprises temperature characteristic data, humidity characteristic data and air pressure characteristic data.

[0014] Further, the obtaining method of the preliminary data set comprises:

[0015] For the seismic data and the well logging data, the missing value is processed by using the mean filling method, the abnormal value is processed by using the interpolation method, the standardization processing is performed by using the Z-Score, and the unified timestamp processing is performed, so as to obtain the preliminary data set.

[0016] Further, the specific method for performing feature extraction on the basis of the preliminary data set by using the improved PCA algorithm to obtain the new feature data set comprises:

[0017] Step A11, for the preliminary data set, a preprocessing matrix is constructed by using method;

[0018] Step A12, for the preprocessing matrix, a covariance matrix is constructed, wherein, is the covariance matrix, is the transpose of the preprocessing matrix, is the preprocessing matrix, and the dimension is n is the total number of samples of the preprocessing matrix, is the total number of features in the preprocessing matrix, and one row in the preprocessing matrix represents one sample, the dimension of ;

[0019] Step A13, the covariance matrix is decomposed into three matrices by using the singular value decomposition method SVD, and the formula is , is the orthogonal matrix of , which contains the left singular vector; is the diagonal matrix of , wherein the elements on the diagonal are singular values; is the orthogonal matrix of , which contains the right singular vector;

[0020] Step A14, the variance contribution degree of the principal component is calculated by using the singular value, and the formula is: wherein, For the first The singular values ​​of the principal components, i.e. the singular values ​​of the principal components. The variance of each principal component, where i is the principal component index. This represents the sum of the singular values ​​of all principal components. For the first The variance contribution of each principal component;

[0021] Step A15: Calculate the average variance contribution of all principal components, using the following formula: , Given the average contribution of all principal components, select the k best principal components whose contribution is greater than the average contribution.

[0022] Step A16: Construct the optimal principal component matrix from the k optimal principal components. , preprocessing matrix Projecting onto the new feature matrix, the calculation formula is: ,in, The new feature matrix after projection is transformed into a new feature dataset using the pandas.DataFrame method.

[0023] Furthermore, the specific methods for using a DNN model to predict oil thickness data, porosity data, and permeability data based on the obtained new feature dataset include:

[0024] Step D11: Input the sample set, which includes... Each set of samples includes a new feature dataset and corresponding oil thickness data, porosity data, and permeability data.

[0025] Step D12: Set the number of neurons in the input layer (VB) to be equal to the number of features in the new feature dataset;

[0026] Use grid search to set the number of hidden layers to OP, with each hidden layer having 2*VB neurons;

[0027] The number of neurons in the output layer is set to 3, which corresponds to the predicted oil thickness data, porosity data, and permeability data.

[0028] Initialize the model weights and biases, and set the number of iterations;

[0029] Step D13: In the PR-th iteration, the sample is passed from the input layer to the first hidden layer, and the output of the first hidden layer is calculated using the hybrid activation function;

[0030] The output of the first hidden layer is passed to the next hidden layer as its input, and the mixed activation function is repeatedly used to calculate the output of the hidden layer until the last hidden layer is reached.

[0031] passing the output of the last hidden layer to the output layer to obtain the output of the PRth iteration, the output including predicted oil thickness data, predicted porosity data, and output permeability data;

[0032] Step D14, using the mean square error as a loss function to calculate the error between the PRth iteration predicted value and the true value for the output PRth iteration predicted oil thickness data, predicted porosity data, and output permeability data, and the error of the predicted oil thickness data, predicted porosity data, and output permeability data is weighted and summed to obtain a total loss function error, wherein the true value refers to the corresponding oil thickness data, porosity data, and permeability data in step D11;

[0033] Step D15, using the chain rule to calculate the gradient of the weight and bias of each layer through the total loss function error, and using the Adam optimizer to update the weight and bias of each layer;

[0034] Step D16, when the set number of iterations is reached, the iteration is stopped; the predicted oil thickness data, porosity data, and permeability data are output.

[0035] Further, the specific way of calculating the output of the first hidden layer using the mixed activation function includes:

[0036] In the DNN model, the activation function includes activation function, activation function, activation function, and activation function.

[0037] The mixed activation function is calculated using the weighted average method, and the formula is: wherein, , , and are the weight coefficients of activation function, activation function, activation function, and activation function, respectively, and .

[0038] Further, the specific way of predicting the oil production data based on the predicted oil thickness data, porosity data, and permeability data using the regularized exploration environment regression model includes:

[0039] Step ff1, input Each set of oil samples includes predicted oil thickness characteristics, porosity characteristics, and permeability characteristics, as well as corresponding oil production capacity data.

[0040] Step ff2: Set the number of iterations; initialize the intercept. Initialize regression coefficients , and ,in, , and These are the regression coefficients for the predicted oil thickness, porosity, and permeability data, respectively.

[0041] Step ff3: Calculate the predicted value of oil production capacity based on the initialized regression coefficients and intercepts. The formula is as follows: ,in, , and This represents the predicted oil thickness, porosity, and permeability data for the s-th oil sample, where s is the oil sample index. This represents the predicted oil production capacity data for the s-th oil sample group.

[0042] Step ff4: Use regularization and exploration environment data to adjust the loss function;

[0043] Step ff5: Calculate the gradient of the loss function with respect to the regression coefficients and intercept through backpropagation; then update the regression coefficients and intercept using gradient descent.

[0044] Step ff6: Repeat steps ff3 to ff5 until the set number of iterations is reached, and then stop to obtain the final regression coefficients and intercepts.

[0045] Furthermore, the specific methods for using regularization and exploration environment data to modify the loss function include:

[0046] The loss function is calculated using the average error, as shown in the formula: ,in, For the oil production capacity data in the s-th oil sample;

[0047] The loss function is corrected using regularization and exploration environment data to obtain the corrected loss function, as shown in the formula: ,in, For regularization parameters, The L2 norm of the regression coefficients, For the index of regression coefficients, For the first One regression coefficient, Let be the correction factor for the exploration environment data of the s-th oil sample at the same time. As a regulating factor;

[0048] The correction factor for the exploration environment data was calculated using a weighted average method. ,in, , and The weights of temperature, humidity, and air pressure data within the exploration environment data are divided. , and These are the temperature, humidity, and air pressure characteristic data values ​​obtained at the same time as the s-th group of oil samples.

[0049] Furthermore, the specific method for setting a production capacity threshold and comparing the predicted oil production capacity data with the production capacity threshold to determine whether to extract oil includes:

[0050] The average and standard deviation of historical oil production capacity are calculated, and the initial production capacity threshold is set based on the average and standard deviation of historical oil production capacity data.

[0051] The average price is calculated based on historical international crude oil price data. The initial production capacity threshold is then adjusted by the ratio of the historical average international crude oil price data to the current international crude oil price data to obtain the production capacity threshold.

[0052] When the oil production capacity data is greater than or equal to the production capacity threshold data, it means that the explored oil can be extracted; when the oil production capacity data is less than the production capacity threshold data, it means that the explored oil cannot be extracted.

[0053] The formula for the average historical oil production capacity data is: ,in, This represents the average historical oil production capacity. This represents the total number of historical oil production data points. It serves as an index for historical oil production data. For the first Historical oil production capacity data;

[0054] The formula for the standard deviation of historical oil production data is: , The standard deviation of historical oil production capacity;

[0055] The initial capacity threshold formula is: , This is the initial capacity threshold. For adjustment coefficients;

[0056] The formula for the average price of historical international crude oil data is: wherein, is an index of historical international crude oil prices, is the price of the historical international crude oil, is the average price of historical international crude oil;

[0057] The capacity threshold formula is: wherein, is the current international crude oil price data.

[0058] An oil exploration system based on artificial intelligence is applied to the oil exploration method based on artificial intelligence, and comprises:

[0059] A data acquisition module: acquiring seismic data, well logging data, exploration environment data and international crude oil price data;

[0060] A data processing module: performing preliminary processing on the seismic data and well logging data to obtain a preliminary data set; based on the preliminary data set, using an improved PCA algorithm to extract features to obtain a new feature data set;

[0061] A data prediction module: using a DNN model to predict oil thickness data, porosity data and permeability data based on the new feature data set;

[0062] An oil exploration module: using a regularized exploration environment regression model to predict oil capacity data based on the predicted oil thickness data, porosity data and permeability data, and setting a capacity threshold, comparing the predicted oil capacity data with the capacity threshold, and determining whether to exploit oil.

[0063] The technical effects and advantages of the oil exploration method and system based on artificial intelligence are as follows:

[0064] The present application realizes the intelligentization and high efficiency of oil exploration through deep analysis of data and model prediction; first, by acquiring seismic data, well logging data, exploration environment data and international crude oil price data, the system can comprehensively obtain multi-dimensional information affecting oil capacity, improving the prediction accuracy;

[0065] Then, the improved PCA algorithm is used for feature extraction, which effectively reduces data redundancy, improves the accuracy of feature data, reduces the amount of calculation, while retaining the core information of the data and enhancing the learning ability of the model;

[0066] Next, based on the deep neural network DNN model, the oil thickness data, porosity data and permeability data are accurately predicted, which provides a basis for subsequent oil capacity prediction;

[0067] Finally, combined with the regularized exploration environment regression model, the system can comprehensively consider the influence of the exploration environment and accurately predict the oil production capacity.

[0068] The historical oil production capacity data and international oil price data are used to correct the production capacity threshold, so that the production capacity prediction is more dynamic and flexible, can respond to market changes in real time, reduces the risk of blind exploitation by accurately predicting the production capacity and intelligently judging whether to exploit or not, and avoids resource waste caused by over-exploitation or inefficient exploitation. BRIEF DESCRIPTION OF DRAWINGS

[0069] Figure 1 It is a schematic diagram of the oil exploration method based on artificial intelligence.

[0070] Figure 2 It is a schematic diagram of the oil exploration system based on artificial intelligence. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0072] Embodiment 1

[0073] Please refer to Figure 1 The oil exploration method based on artificial intelligence comprises:

[0074] Step S1, collecting seismic data, well logging data, exploration environment data and international crude oil price data;

[0075] Step S2, based on the seismic data and the well logging data, performing preliminary processing to obtain a preliminary data set; based on the preliminary data set, using an improved PCA algorithm to perform feature extraction to obtain a new feature data set;

[0076] Step S3, based on the new feature data set, using a DNN model to predict oil thickness data, porosity data and permeability data;

[0077] Step S4, based on the predicted oil thickness data, porosity data and permeability data, using a regularized exploration environment regression model to predict oil production capacity data, and setting a production capacity threshold, comparing the predicted oil production capacity data with the production capacity threshold, and judging whether to exploit oil.

[0078] The seismic data comprises amplitude feature data, wave velocity feature data, radiation feature data, seismic wave frequency feature data and magnetic force feature data.

[0079] Well logging data includes resistivity characteristic data, gamma ray characteristic data, sonic transit time characteristic data, rock density characteristic data, neutron porosity characteristic data, electrical conductivity characteristic data, and natural gamma characteristic data;

[0080] Exploration environment data includes temperature characteristic data, humidity characteristic data, and air pressure characteristic data.

[0081] Sensors are installed in the borehole to acquire data, specifically including: acquiring amplitude characteristic data, wave velocity characteristic data, radiation characteristic data, and seismic wave frequency characteristic data through seismic detectors; and acquiring magnetic characteristic data through magnetometers.

[0082] Resistivity characteristics of formations are measured using resistivity logging tools; gamma ray characteristics and natural gamma ray characteristics of formations are measured using gamma ray logging tools; sonic transit time characteristics of formations are measured using sonic logging tools; rock density characteristics are measured using density logging tools; neutron porosity characteristics of formations are measured using neutron logging tools; and electrical conductivity characteristics are measured using electrical conductivity logging tools.

[0083] Temperature characteristic data, humidity characteristic data, and air pressure characteristic data are obtained by installing temperature sensors, humidity sensors, and air pressure sensors.

[0084] Obtain international crude oil price data through financial websites or commodity trading platforms.

[0085] The methods for obtaining the initial dataset include:

[0086] For seismic and well logging data, the mean imputation method was used to handle missing values, the interpolation method was used to handle outliers, the Z-score was used for standardization, and a unified timestamp was applied to obtain a preliminary dataset.

[0087] Based on the initial dataset, the improved PCA algorithm is used to extract features, and the specific methods for obtaining a new feature dataset include:

[0088] Step A11: For the initial dataset, use... The method constitutes the preprocessing matrix;

[0089] Step A12: Construct the covariance matrix for the preprocessed matrix. ,in, Let covariance matrix be the variance matrix. For the transpose of the preprocessed matrix, The preprocessed matrix has dimensions of . n is the total number of preprocessed matrix samples. This represents the total number of features in the preprocessing matrix, where each row in the preprocessing matrix represents a single sample. The dimension is ;

[0090] Step A13: Use the Singular Value Decomposition (SVD) method to decompose the covariance matrix into three matrices, as shown in the formula. , for An orthogonal matrix containing left singular vectors; for A diagonal matrix, where the elements on the diagonal are singular values; for An orthogonal matrix containing right singular vectors;

[0091] Step A14: Calculate the variance contribution of the principal components using singular values. The formula is: ,in, For the first The singular values ​​of the principal components, i.e. the singular values ​​of the principal components. The variance of each principal component, where i is the principal component index. This represents the sum of the singular values ​​of all principal components. For the first The variance contribution of each principal component;

[0092] Step A15: Calculate the average variance contribution of all principal components, using the following formula: , Given the average contribution of all principal components, select the k best principal components whose contribution is greater than the average contribution.

[0093] Step A16: Construct the optimal principal component matrix from the k optimal principal components. , preprocessing matrix Projecting onto the new feature matrix, the calculation formula is: ,in, To transform the projected new feature matrix into a new feature dataset, use the pandas.DataFrame method.

[0094] Using an improved PCA model for feature extraction can effectively reduce data dimensionality, retain the most important features, thereby reducing noise and redundant information and improving the accuracy of subsequent analysis.

[0095] The PCA model usually relies on calculating the eigenvalues and eigenvectors of the covariance matrix, while the improved PCA model directly decomposes the data matrix through the singular value decomposition method SVD, avoiding the process of solving the covariance matrix and reducing the computational complexity, especially when dealing with large-scale data sets; SVD can efficiently capture the principal components of the data and perform better in numerical stability and calculation speed;

[0096] In addition, the new feature matrix is converted into a dataset using the pandas.DataFrame method in the Pandas library, which facilitates subsequent data processing and analysis, making the entire process more efficient and practical.

[0097] Based on the obtained new feature dataset, the specific method of using a DNN model to predict oil thickness data, porosity data and permeability data includes:

[0098] Step D11, input the sample set, which includes group samples, each group sample including a new feature dataset and corresponding oil thickness data, porosity data and permeability data;

[0099] Step D12, set the number of neurons VB of the input layer equal to the number of features in the new feature dataset;

[0100] Use grid search to set the number of hidden layers OP, and set the number of neurons in each hidden layer to 2*VB;

[0101] Set the number of neurons in the output layer to 3, which corresponds to the predicted oil thickness data, porosity data and permeability data;

[0102] Initialize model weights and biases, and set the number of iterations;

[0103] Step D13, in the PRth iteration, pass the sample from the input layer to the first hidden layer, and use the hybrid activation function to calculate the output of the first hidden layer, the formula is: wherein, is the output of the first hidden layer, is the weight of the first hidden layer, is the input sample, is the bias of the first hidden layer, is the hybrid activation function;

[0104] Pass the output of the first hidden layer to the next hidden layer as the input of the next hidden layer, and repeatedly use the hybrid activation function to calculate the output of the hidden layer until the last hidden layer is finished, the formula is: wherein, is the output of the first hidden layer, is the output of the first hidden layer, for the first hidden layer, weights of the layer, for the first hidden layer, output of the layer, for the first hidden layer, bias of the layer;

[0105] passing the output of the last hidden layer to the output layer to obtain the output of the PRth iteration, and the calculation formula is wherein, output of the output layer, the output includes predicted oil thickness data, predicted porosity data, and output permeability data, weights of the output layer, bias of the output layer;

[0106] Step D14, using the mean square error as the loss function to calculate the error between the predicted value and the true value of the predicted oil thickness data, the predicted porosity data, and the output permeability data of the PRth iteration, and the errors of the predicted oil thickness data, the predicted porosity data, and the output permeability data are weighted and summed to obtain the total loss function error, wherein the true value refers to the corresponding oil thickness data, porosity data, and permeability data in step D11;

[0107] Step D15, using the chain rule to calculate the gradient of the weight and bias of each layer through the total loss function error, and using the Adam optimizer to update the weight and bias of each layer;

[0108] Step D16, when the set number of iterations is reached, the iteration is stopped; and the predicted oil thickness data, porosity data, and permeability data are outputted.

[0109] The DNN model can effectively capture complex nonlinear relationships and perform well in processing large-scale data, with good generalization ability, and can adapt to different data types and feature distributions. These characteristics enable the DNN to provide more accurate and comprehensive prediction results in oil exploration, thereby improving the efficiency of exploration and the accuracy of decision-making.

[0110] The specific way of calculating the output of the first hidden layer using the mixed activation function includes:

[0111] In the DNN model, the activation function includes the activation function, the activation function, the activation function, and the activation function.

[0112] The mixed activation function is calculated using the weighted average method, and the formula is: wherein, , , and are the weight coefficients of the activation functions, activation function, activation function, activation function, and activation function, and , , , and are each set to 0.25;

[0113] The mixed activation function is used to calculate the output of the hidden layer in the DNN model. By combining the characteristics of multiple activation functions, the complex patterns of the input data can be more effectively captured. The mixed activation function integrates the advantages of different activation functions using the weighted average method, thereby improving the non-linear expression ability of the model. This approach not only enhances the adaptability of the model to different features, but also improves the gradient propagation and reduces the problem of gradient vanishing, ultimately improving the learning efficiency and prediction accuracy of the model.

[0114] The specific way of predicting the oil production capacity data based on the predicted oil thickness data, porosity data, and permeability data using the regularized exploration environment regression model includes:

[0115] Step ff1, input a group of oil samples, each group of oil samples including a predicted oil thickness feature data, a porosity feature data, and a permeability feature data, and corresponding oil production capacity data;

[0116] Step ff2, set the number of iterations; initialize the intercept , initialize the regression coefficients , and , wherein , and are the regression coefficients of the input predicted oil thickness data, porosity data, and permeability data, respectively;

[0117] Step ff3, based on the initialized regression coefficients and intercept, calculate the predicted value of the oil production capacity data, the formula is: , wherein , and represent the predicted oil thickness data, porosity data, and permeability data of the s-th group of oil samples, s is the index of the oil sample, represents the predicted oil production capacity data of the s-th group of oil samples;

[0118] Step ff4, use the regularization and exploration environment data to correct the loss function;

[0119] Step ff5: Calculate the gradient of the loss function with respect to the regression coefficients and intercept through backpropagation; then update the regression coefficients and intercept using gradient descent.

[0120] Step ff6: Repeat steps ff3 to ff5 until the set number of iterations is reached, and then stop to obtain the final regression coefficients and intercepts;

[0121] Using predicted oil thickness, porosity, and permeability data to predict oil production capacity through a regularized exploration environment regression model offers higher accuracy and reliability compared to directly predicting oil production capacity using new feature datasets. Firstly, it can more deeply capture the complex relationship between geological features and production capacity, and regularization reduces overfitting, improving the model's generalization ability. Furthermore, utilizing intermediate prediction results as features allows for better integration and utilization of information from different data sources, making the final production capacity prediction more scientific. This not only improves the quality of the prediction but also provides stronger support for decision-making, helping to optimize oil resource development strategies.

[0122] Specific ways to use regularization and exploration environment data to modify the loss function include:

[0123] The loss function is calculated using the average error, as shown in the formula: ,in, For the oil production capacity data in the s-th oil sample;

[0124] The loss function is corrected using regularization and exploration environment data to obtain the corrected loss function, as shown in the formula: ,in, These are regularization parameters, obtained through empirical rules, used to control the weight of the regularization terms. The L2 norm of the regression coefficients, For the index of regression coefficients, For the first One regression coefficient, Let be the correction factor for the exploration environment data of the s-th oil sample at the same time. The adjustment factor is set using a random method;

[0125] The correction factor for the exploration environment data was calculated using a weighted average method. ,in, , and The exploration environment data is divided into three weights: temperature data, humidity data, and air pressure data, each set as follows: , , and These are the temperature, humidity, and air pressure characteristic data values ​​obtained at the same time as the s-th group of oil samples.

[0126] By using regularization and exploration environment data to correct the loss function, the robustness and generalization ability of the model are effectively improved. Regularization controls the complexity of the regression coefficients and prevents overfitting. At the same time, the correction coefficients of exploration environment data are combined to enable the model to better reflect the impact of the exploration environment on oil production capacity. By integrating temperature, humidity and air pressure exploration environment data through the weighted average method, more comprehensive contextual information is provided, thereby improving the accuracy and reliability of the prediction.

[0127] Setting production capacity thresholds and comparing predicted oil production capacity data with these thresholds to determine whether to proceed with oil extraction includes the following specific methods:

[0128] The average and standard deviation of historical oil production capacity are calculated, and the initial production capacity threshold is set based on the average and standard deviation of historical oil production capacity data.

[0129] The average price is calculated based on historical international crude oil price data. The initial production capacity threshold is then adjusted by the ratio of the historical average international crude oil price data to the current international crude oil price data to obtain the production capacity threshold.

[0130] When the oil production capacity data is greater than or equal to the production capacity threshold data, it means that the explored oil can be extracted; when the oil production capacity data is less than the production capacity threshold data, it means that the explored oil cannot be extracted.

[0131] The formula for the average historical oil production capacity data is: ,in, This represents the average historical oil production capacity. This represents the total number of historical oil production data points. It serves as an index for historical oil production data. For the first Historical oil production capacity data;

[0132] The formula for the standard deviation of historical oil production data is: , The standard deviation of historical oil production capacity;

[0133] The initial capacity threshold formula is: , This is the initial capacity threshold. For adjustment coefficients;

[0134] The formula for the average price of historical international crude oil data is: ,in, As an index of historical international crude oil prices, For the first a historical international crude oil price, an average price of the historical international crude oil;

[0135] The capacity threshold formula is: wherein, the current international crude oil price data;

[0136] By obtaining the capacity threshold through historical oil production capacity data and international crude oil price data, the decision-making can be more scientific and data-driven. First, using historical data to calculate the average and standard deviation provides a reasonable basis for the setting of the capacity threshold, which helps to identify the normal capacity range. Second, combining the proportion of international crude oil prices to correct the initial threshold makes the threshold more flexible, which can timely reflect market fluctuations and changes in the economic environment, improve the accuracy of resource exploitation judgment, optimize resource allocation, and reduce the risk of exploitation, thereby more effectively supporting oil exploration and development decisions.

[0137] In this embodiment, through deep analysis of data and model prediction, intelligent and efficient oil exploration is realized. First, by collecting seismic data, well logging data, exploration environment data and international crude oil price data, the system can comprehensively obtain multi-dimensional information affecting oil production capacity, improving the prediction accuracy.

[0138] Then, through the improved PCA algorithm for feature extraction, the data redundancy is effectively reduced, the accuracy of feature data is improved, the calculation amount is reduced, the core information of the data is retained, and the learning ability of the model is enhanced.

[0139] Next, based on the deep neural network DNN model, the oil thickness data, porosity data and permeability data are accurately predicted, providing a basis for subsequent oil production capacity prediction.

[0140] Finally, combined with the regularized exploration environment regression model, the system can comprehensively consider the influence of the exploration environment to accurately predict the oil production capacity.

[0141] Using historical oil production data and international oil price data to correct the capacity threshold makes the capacity prediction more dynamic and flexible, which can respond to market changes in real time. By accurately predicting the capacity and intelligently judging whether to exploit or not, the risk of blind exploitation is reduced, and resource waste caused by over-exploitation or inefficient exploitation is avoided.

[0142] Embodiment 2

[0143] Please refer to Figure 2 The part not described in detail in this embodiment can be seen from the description of embodiment 1. An oil exploration system based on artificial intelligence is provided, which comprises:

[0144] The data acquisition module acquires seismic data, logging data, exploration environment data and international crude oil price data;

[0145] The data processing module performs preliminary processing on the seismic data and the logging data to obtain a preliminary data set, and performs feature extraction on the preliminary data set by using an improved PCA algorithm to obtain a new feature data set;

[0146] The data prediction module uses a DNN model to predict the oil thickness data, the porosity data and the permeability data based on the new feature data set.

[0147] The oil exploration module uses a regularized exploration environment regression model to predict the oil production capacity data based on the predicted oil thickness data, the porosity data and the permeability data, sets a production capacity threshold, compares the predicted oil production capacity data with the production capacity threshold, and determines whether to exploit the oil.

[0148] Embodiment 3

[0149] The embodiment discloses an electronic device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the operation mode of the above-mentioned oil exploration method and system based on artificial intelligence is realized.

[0150] Since the electronic device introduced in the embodiment is used to implement the oil exploration method and system based on artificial intelligence in the embodiment, the specific implementation mode of the electronic device and various changes thereof can be understood by those skilled in the art based on the oil exploration method and system based on artificial intelligence introduced in the embodiment. Therefore, how the electronic device implements the method in the embodiment will not be described in detail. As long as the electronic device used to implement the oil exploration method and system based on artificial intelligence in the embodiment is implemented by those skilled in the art, it belongs to the scope of protection of the present application.

[0151] The above formulas are dimensionless values calculated, and the formulas are obtained by software simulation of a large amount of data to obtain a formula of the nearest real situation. The preset parameters and threshold values in the formula are set by those skilled in the art according to the actual situation.

[0152] The above-mentioned only is the preferred embodiment of the present application, the protection scope of the present application is not only limited to the above-mentioned embodiment, all technical solutions under the idea of the present application belong to the protection scope of the present application. It should be pointed out that, for the ordinary technical user in this technical field, some improvements and decorations without departing from the principle of the present application are also regarded as the protection scope of the present application.

Claims

1. An artificial intelligence-based method for oil exploration, characterized by, The application relates to a method for predicting oil production capacity, and belongs to the field of oil exploration. The method comprises the following steps: S1, collecting seismic data, well logging data, exploration environment data and international crude oil price data; S2, performing preliminary processing based on the seismic data and the well logging data to obtain a preliminary data set; performing feature extraction based on the preliminary data set by using an improved PCA algorithm to obtain a new feature data set; S3, using a DNN model to predict oil thickness data, porosity data and permeability data based on the new feature data set; S4, using a regularized exploration environment regression model to predict oil production capacity data based on the predicted oil thickness data, porosity data and permeability data, setting a production capacity threshold, comparing the predicted oil production capacity data with the production capacity threshold, and determining whether to exploit oil; Step ff1, input The petroleum samples are grouped, each group of petroleum samples including a predicted petroleum thickness characteristic data, a porosity characteristic data and a permeability characteristic data, and corresponding petroleum productivity data. Step ff2, set iteration number; initialize intercept , initialize regression coefficients 、 and wherein, 、 and are the regression coefficients for the input predicted oil thickness data, porosity data, and permeability data, respectively; Step ff3, performing prediction value calculation of oil production data based on the initialized regression coefficients and intercept, the formula is: wherein, , and represents the predicted oil thickness data, porosity data and permeability data of the s-th group of oil samples, s is the index of the oil sample, represents the predicted oil production data of the s-th group of oil samples; The specific method for predicting oil production capacity data by using the regularized exploration environment regression model based on the predicted oil thickness data, porosity data and permeability data comprises the following steps: ff4, using regularization and exploration environment data to correct a loss function; ff5, calculating the gradient of the loss function on regression coefficients and intercepts by back propagation, and then updating the regression coefficients and intercepts by using a gradient descent method; 2. The artificial intelligence based oil exploration method as claimed in claim 1 wherein, ff6, repeatedly executing steps ff3 to ff5 until a set iteration number is reached to stop, and finally obtaining the regression coefficients and intercepts. The seismic data comprises amplitude feature data, wave velocity feature data, radiation feature data, seismic wave frequency feature data and magnetic force feature data; The well logging data comprises resistivity feature data, gamma ray feature data, acoustic travel time feature data, rock density feature data, neutron porosity feature data, conductivity feature data and natural gamma feature data; 3. The artificial intelligence based oil exploration method as claimed in claim 2, wherein, The exploration environment data comprises temperature feature data, humidity feature data and air pressure feature data. The method for obtaining the preliminary data set comprises the following steps:

4. The artificial intelligence based oil exploration method as claimed in claim 3, wherein, For the seismic data and the well logging data, missing value processing is performed by using a mean filling method, abnormal value processing is performed by using an interpolation method, standardization processing is performed by using Z-Score, and unified timestamp processing is performed to obtain the preliminary data set. Step A11, for the preliminary data set, using The method constitutes a pre-processing matrix; Step A12, constructing a covariance matrix for the preprocessed matrix, wherein, is a covariance matrix, is a transpose of the preprocessed matrix, is a preprocessed matrix, with dimensions n is a total number of samples of the preprocessed matrix, is a total number of features in the preprocessed matrix, one row of the preprocessed matrix representing one sample, has dimensions ; Step A13: Use the Singular Value Decomposition (SVD) method to decompose the covariance matrix into three matrices, as shown in the formula. , for An orthogonal matrix containing left singular vectors; for A diagonal matrix, where the elements on the diagonal are singular values; for An orthogonal matrix containing right singular vectors; Step A14, calculate the variance contribution of principal components by singular values, formula is: Wherein, is the singular value of the i th principal component, that is, the variance of the i th principal component, i is the principal component index, represents the sum of all principal component singular values, is the variance contribution of the i th principal component; Step A15, calculate the average variance contribution of all principal components, the formula is: , The average contribution of all principal components is selected, and k best principal components with a contribution greater than the average contribution are selected. Step A16, constructing the best principal components into a best principal component matrix , the pre-processing matrix is projected into a new feature matrix, and the calculation formula is: , wherein is the new feature matrix after projection, and the new feature matrix is converted into a new feature dataset using the pandas.DataFrame method.

5. The artificial intelligence based oil exploration method as claimed in claim 4, wherein, The specific method for performing feature extraction based on the preliminary data set by using the improved PCA algorithm to obtain the new feature data set comprises the following steps: Step D11, inputting a sample set, the sample set including group samples, each group sample including a new feature data set and corresponding oil thickness data, porosity data, and permeability data; The specific method for predicting oil thickness data, porosity data and permeability data based on the obtained new feature data set by using the DNN model comprises the following steps: D12, setting the number of neurons VB of an input layer to be equal to the number of features in the new feature data set; using a grid search to set the number of hidden layers to be OP, and setting the number of neurons in each hidden layer to be 2*VB; setting the number of neurons in an output layer to be 3, i.e. corresponding to the predicted oil thickness data, porosity data and permeability data; initializing model weights and biases, and setting the number of iterations; Step D13, in the PRth iteration, the sample is passed from the input layer to the first hidden layer, and the output of the first hidden layer is calculated using a hybrid activation function; the output of the first hidden layer is passed to the next hidden layer as the input of the next hidden layer, and the output of the hidden layer is repeatedly calculated using a hybrid activation function until the last hidden layer ends; the output of the last hidden layer is passed to the output layer to obtain the output of the PRth iteration, which includes predicted oil thickness data, predicted porosity data, and output permeability data; Step D14, for the predicted oil thickness data, predicted porosity data, and output permeability data of the PRth iteration output, the mean square error is used as the loss function to calculate the error between the predicted value and the true value of the PRth iteration, and the errors of the predicted oil thickness data, predicted porosity data, and output permeability data are weighted and summed to obtain the total loss function error, wherein the true value refers to the corresponding oil thickness data, porosity data, and permeability data in step D11; Step D15, by using the total loss function error, the gradient of the weight and bias of each layer is calculated using the chain rule, and the weight and bias of each layer are updated using the Adam optimizer; Step D16, when the set number of iterations is reached, the iteration is stopped; the predicted oil thickness data, porosity data, and permeability data are output.

6. The artificial intelligence-based oil exploration method according to claim 5, characterized in that, The specific way of calculating the output of the first hidden layer using a hybrid activation function includes: In the DNN model, the activation function includes the activation function, the activation function, the activation function, and the activation function; The hybrid activation function is calculated using the weighted average method, and the formula is: wherein, , , and are the weight coefficients of the activation function, the activation function, the activation function, the activation function, and the activation function, and .

7. The artificial intelligence-based oil exploration method according to claim 6, characterized in that, The specific way of using regularization and exploration environment data to correct the loss function includes: The loss function is calculated using the mean squared error, which is given by: where, is the oil production data in the s-th group of oil samples; The loss function is corrected using regularization and exploration environment data to obtain the corrected loss function, as shown in the formula: ,in, For regularization parameters, The L2 norm of the regression coefficients, For the index of regression coefficients, For the first One regression coefficient, Let be the correction factor for the exploration environment data of the s-th oil sample at the same time. As a regulating factor; The correction coefficient of the exploration environment data is calculated by a weighted average method, wherein, , and are the weights of the temperature data, the humidity data and the air pressure data in the exploration environment data, , and are the temperature feature data value, the humidity feature data value and the air pressure feature data obtained at the same time as the s-th group of oil samples, respectively.

8. The artificial intelligence based oil exploration method as claimed in claim 7, wherein, The specific way of setting a production threshold and comparing the predicted oil production data with the production threshold to determine whether to exploit the oil includes: Based on the historical oil production data, the mean and standard deviation are calculated, and the initial production threshold is set according to the mean and standard deviation of the historical oil production data; The price average is calculated according to the historical international crude oil price data, and the initial production threshold is corrected by the ratio of the historical international crude oil average price data to the current international crude oil price data to obtain the production threshold; When the oil production data is greater than or equal to the production threshold data, it indicates that the explored oil can be exploited; when the oil production data is less than the production threshold data, it indicates that the explored oil cannot be exploited; The formula for the average of historical oil production data is: wherein, is the average of historical oil production data, is the total number of historical oil production data, is the index of historical oil production data, is the historical oil production data at the th index. The standard deviation formula for historical oil production capacity data is: , is the standard deviation of historical oil production capacity; The initial production capacity threshold formula is: , is the initial production capacity threshold, is the adjustment coefficient; The price average formula for historical international crude oil price data is: wherein, is an index of historical international crude oil prices, is the price of the historical international crude oil, is the average price of historical international crude oil; The capacity threshold formula is: wherein, is the current international original price data.

9. An artificial intelligence-based oil exploration system for implementing the artificial intelligence-based oil exploration method according to any one of claims 1 to 8, characterized in that, It includes: Data acquisition module: collect seismic data, logging data, exploration environment data, and international crude oil price data; Data processing module: preliminary processing of seismic data and logging data to obtain a preliminary data set; based on the preliminary data set, feature extraction is performed using an improved PCA algorithm to obtain a new feature data set; Data prediction module: using a DNN model to predict oil thickness data, porosity data, and permeability data based on the new feature data set; Oil exploration module: based on the predicted oil thickness data, porosity data, and permeability data, use a regularized exploration environment regression model to predict oil production data, and set a production threshold to compare the predicted oil production data with the production threshold to determine whether to exploit the oil.

Images