Method for predicting maturity of hydrocarbon source rock and crude oil based on Bayesian classification

Abstract

The invention provides a method for predicting the maturity of hydrocarbon source rocks and crude oil based on Bayesian classification, and particularly relates to a method for predicting the maturity of hydrocarbon source rocks and crude oil based on aromatic hydrocarbon biomarker parameters of Bayesian classification machine learning. The method comprises the following steps: firstly, constructing a hydrocarbon source rock and crude oil aromatic hydrocarbon biomarker parameter data set, and dividing the data set into a training set and a test set; then performing data set feature selection processing on the aromatic hydrocarbon biomarker parameter data set to obtain a condition attribute variable set; constructing a neural network model, and training to obtain a sample maturity prediction model based on Bayesian machine classifier learning; and verifying the prediction model through the test set, and finally predicting the maturity of the sample based on the sample aromatic hydrocarbon biomarker parameter prediction model trained by a Bayesian classifier to obtain the maturity of the predicted sample. According to the method, the problems of limitation and multiplicity of solutions of biological standard substance parameter quantitative maturity at present can be solved.

Description

Technical Field

[0001] The invention relates to the field of petroleum geological exploration, and in particular to a method for predicting source rock and crude oil maturity based on Bayesian classification. Background Art

[0002] The maturity of source rocks refers to the degree of thermal evolution of organic matter in source rocks. The main factors affecting maturity are time and temperature. The maturity evaluation of source rocks is an important indicator for source rock evaluation. Regardless of the abundance and type of organic matter in source rocks, only when the organic matter maturity reaches a certain stage can source rocks begin to generate hydrocarbons in large quantities and form large-scale oil reservoirs. The maturity of organic matter is directly related to whether a region is worth exploring and developing.

[0003] Therefore, accurate evaluation of organic matter maturity is very important. Summary of the invention

[0004] The main purpose of the present invention is to provide a method for predicting the maturity of source rocks and crude oil based on Bayesian classification, which can improve the current limitations and multiple solutions of quantitative maturity of biological marker parameters. In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting source rock and crude oil maturity based on Bayesian classification, the method comprising: S1: Construct a data set of source rock and crude oil aromatic biomarker parameters, and evaluate the sample maturity by the vitrinite reflectance Ro of the source rock. The obtained data set is divided into a training set, a test set, and a prediction set according to a ratio of 7:2:1, where the training set and the test set are labeled source rock samples, and the prediction set is an unlabeled crude oil sample; S2: Perform dataset feature selection processing on source rock and crude oil aromatic biomarker parameter datasets to obtain a collection of conditional attribute variables; S3: Construct a neural network model, train the neural network model based on the training set and the conditional attribute variable set based on the Bayesian machine classifier, and obtain a source rock and crude oil maturity prediction model based on the Bayesian machine classifier learning; S4: Use the test set to test the source rock and crude oil maturity prediction model based on the Bayesian classifier learning to obtain the accuracy, and determine whether the accuracy exceeds the set threshold. If it does not exceed, update the attribute variable collection and return to step S3. If it exceeds, execute step S5; S5: Obtain source rock and crude oil aromatic biomarker parameters, input the source rock and crude oil aromatic biomarker parameters into the source rock and crude oil maturity prediction model based on Bayesian machine classifier learning to make predictions, and obtain the maturity judgment of the source rock and crude oil.

[0005] In the preferred embodiment, the data characteristics in the dataset of biomarker parameter data of source rocks and crude oils include 12; Four maturity evaluation criteria are defined according to the vitrinite reflectance Ro of the source rock, specifically: When Ro < 0.7, the maturity belongs to immature - low maturity, and the classification label is 1; When 0.7 ≤ Ro < 1.3, the maturity belongs to maturity, and the classification label is 2; When 1.3 ≤ Ro ≤ 2.0, the maturity belongs to high maturity, and the classification label is 3; When 2.0 < Ro, the maturity belongs to over - maturity, and the classification label is 4.

[0006] In the preferred embodiment, the specific steps in step S2 are: Including: calculating the weight scores of the feature vectors of the training set through the Relief algorithm, and taking the top five features with the highest scores as the conditional attribute variable set, and the method is: a. Randomly select a sample R from the training set; b. Find the nearest neighbor sample H, called Near Hit, from the samples of the same class as sample R, and find the nearest neighbor sample M, called Near Miss, from the samples of different classes from sample R; c. Then update the weight of each feature: If the distance between R and Near Hit on a certain feature is less than the distance between R and Near Miss, it means that this feature is beneficial for distinguishing the nearest neighbors of the same class and different classes, then increase the weight of this feature; If the distance between R and Near Hit on a certain feature is greater than the distance between R and Near Miss, it means that this feature has a negative effect on distinguishing the nearest neighbors of the same class and different classes, then reduce the weight of this feature; d. Return to execute step c until after executing step c N times, and take the weights of each feature after executing step c N times as the weight scores of each feature.

[0007] In the preferred embodiment, the construction of the neural network model in step S3 is specifically: S31. For each maturity level in the training set, calculate its prior probability , and the prior probability can be obtained by calculating the proportion of each category in the training set; At the same time, for each feature and category , calculate the conditional probability , which needs to be statistically analyzed according to the data in the training set; S32. In each iteration, input the samples in the training set into the neural network model one by one. For each input sample, calculate the output result according to the current network parameters, that is, the predicted probability distribution of the maturity level; Then, according to the formula of the Bayesian machine classifier: , calculate the probability value corresponding to each category and compare it with the true label; in, are categories, i.e. different maturity levels, are the characteristics, i.e., aromatic hydrocarbon biomarker parameters; Indicates category The prior probability of Indicates that in the category Characteristics under the condition Probability of occurrence; By calculating the probability value corresponding to each category and then finding the category with the highest probability, it is considered that the item to be classified belongs to this category, thus obtaining a maturity prediction model based on Bayesian machine classifier learning; S33. Use the loss function to measure the difference between the predicted result and the true label, and adjust the weights and biases of the neural network through the back propagation algorithm to minimize the loss function. S34. When the performance of the model on the validation set reaches a satisfactory standard, the training is stopped to obtain a source rock and crude oil maturity prediction model based on Bayesian machine classifier learning. This model can be used to predict the maturity of new source rocks and crude oil samples.

[0008] In the preferred embodiment, the determination of maturity of source rock and crude oil in step S5 is specifically as follows: The source rock and crude oil aromatic biomarker parameters collected by the model sensor are used as the input of the Bayesian classifier, and the maturity level of the source rock and crude oil is I; Assume that the maturity level of source rock and crude oil is JS, the Lth conditional attribute variable is KL, and the conditional probability M of the sample maturity category JS is calculated according to the Bayesian formula:

[0009] Where S is the conditional probability that the conditional attribute variable KL belongs to the maturity category JS. Let the prior probabilities of JS and KL be P(JS) and P(KL) respectively, and calculate the posterior probability Q of KL belonging to JS. The formula is:

[0010] Then, the weight pl is assigned to the conditional attribute KL variable to obtain the weighted maturity feature, and the information entropy q(JS|KL) of KL about the maturity category JS is calculated. The formula is:

[0011] The information entropy of maturity feature KL attributed to each maturity category JS is statistically analyzed, and the maturity level corresponding to the highest information entropy q(JS|KL) is selected as the category of the conditional attribute variable, which is regarded as the maturity of source rock and crude oil.

[0012] The present invention provides a method for predicting the maturity of source rocks and crude oil based on Bayesian classification. The present invention uses aromatic biomarkers as parameters, and classifies and trains a large number of aromatic biomarker parameters based on the Bayesian classification machine learning method to obtain their maturity levels, thereby achieving qualitative analysis of sample maturity. The method can improve the current limitations and multiple solutions of quantitative maturity of biomarker parameters, has more efficient data utilization and higher accuracy, and can promote the widespread use of aromatic biomarkers.

[0013] Comparing with the existing methods and technologies, it can be seen that the current scholars mainly use conventional single biomarker parameters to qualitatively and quantitatively determine maturity, and the method based on Bayesian classification and machine learning has not been effectively used. The Bayesian classification and machine learning method used in the present invention to predict maturity is a typical artificial intelligence method, which has higher applicability and accuracy for the maturity evaluation of source rocks and crude oil, and provides a good method for maturity evaluation, with good application prospects and promotion value.

[0014] This method is more advanced than current methods, and it is difficult for competitors or the public to crack or imitate it through legal means such as disassembly, measurement, and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] The present invention will be further described below in conjunction with the accompanying drawings and embodiments: Figure 1 It is the flow chart of model design of the present invention; Figure 2 is a flow chart of the ReliefF algorithm of the present invention; Figure 3 It is a simple Bayesian network structure diagram of the present invention; Figure 4 This is a graph of Ro results of source rock samples of the Yanghugou Formation and the Ulalike Formation in the Ordos Basin of the present invention; DETAILED DESCRIPTION Example 1 like Figures 1 to 4 As shown, a method for predicting source rock and crude oil maturity based on Bayesian classification includes: The present invention relates to the field of petroleum geological exploration, and specifically to a method for predicting source rock and crude oil maturity using aromatic biomarker parameters based on Bayesian classification machine learning. The method comprises: Firstly, a data set of source rock and crude oil aromatic biomarker parameters was constructed, and the data set was divided into a training set and a test set. Then, the aromatic hydrocarbon biomarker parameter data set is processed with data set feature selection to obtain a collection of conditional attribute variables; Then, a neural network model is constructed and trained to obtain a prediction model of sample maturity based on Bayesian machine classifier learning; The prediction model was verified by the test set, and finally the sample maturity was predicted based on the sample aromatic biomarker parameter prediction model trained by the Bayesian classifier to obtain the predicted sample maturity.

[0016] In order to achieve the above object, the present invention provides a method for predicting the maturity of source rocks and crude oil based on Bayesian classification machine learning. The specific technical scheme is as follows: S1: Construct a data set of source rock and crude oil aromatic biomarker parameters, and evaluate the sample maturity by the vitrinite reflectance Ro of the source rock. The obtained data set is divided into a training set, a test set, and a prediction set according to a ratio of 7:2:1, where the training set and the test set are labeled source rock samples, and the prediction set is an unlabeled crude oil sample; S2: Perform dataset feature selection processing on source rock and crude oil aromatic biomarker parameter datasets to obtain a collection of conditional attribute variables; S3: Construct a neural network model, train the neural network model based on the training set and the conditional attribute variable set based on the Bayesian machine classifier, and obtain a source rock and crude oil maturity prediction model based on the Bayesian machine classifier learning; S4: Use the test set to test the source rock and crude oil maturity prediction model based on the Bayesian classifier learning to obtain the accuracy, and determine whether the accuracy exceeds the set threshold. If it does not exceed, update the attribute variable collection and return to step S3. If it exceeds, execute step S5; S5: Obtain hydrocarbon source rock and crude oil biomarker parameters, input the hydrocarbon source rock and crude oil biomarker parameters into the hydrocarbon source rock and crude oil maturity prediction model based on Bayesian machine classifier learning to make predictions, and obtain the hydrocarbon source rock and crude oil maturity judgments.

[0017] In the preferred scheme, the data features of the source rock and crude oil aromatic biomarker parameter data set include 12, which are: IMPI-1=(1.5*(3-MP+2-MP)) / (P+1-MP+9-MP); IIMPR2=(3*2-MP) / (P+1-MP+9-MP); MNR=2-MN / 1-MN; ENR = 2-ethylnaphthalene / 1-ethylnaphthalene; DNR=(2,6-DMN+2,7-DMN) / 1,5-DMN; Rc1 = 0.09 * DNR - 1 + 0.49; TNR - 1 = 2,3,6 - TMN / (1,4,6 - TMN + 1,3,5 - TMN); TNR - 2 = (2,3,6 - TMN + 1,3,7 - TMN) / (1,4,6 - TMN + 1,3,5 - TMN + 1,3,6 - TMN); Rc2 = 0.6 * TNR - 2 + 0.4; TMNr = 1,3,7 - / (1,3,7 - + 1,2,5 - )TMN; TeMNr = 1,3,6,7 - / (1,3,6,7 - + 1,2,5,6 - + 1,2,3,5 - )TeMN; PMNr = 1,2,4,6,7 - / (1,2,4,6,7 - + 1,2,3,5,6 - )PMN. Wherein: phenanthrene (P), methylphenanthrene (MP), methylnaphthalene (MN), dimethylnaphthalene (DMN), trimethylnaphthalene (TMN), tetramethylnaphthalene (TeMN) and pentamethylnaphthalene (PMN).

[0018] Four maturity evaluation criteria are defined according to the vitrinite reflectance Ro of source rocks, specifically: When Ro < 0.7, the maturity belongs to immature - low maturity, and the classification label is 1; When 0.7 ≤ Ro < 1.3, the maturity belongs to maturity, and the classification label is 2; When 1.3 ≤ Ro ≤ 2.0, the maturity belongs to high maturity, and the classification label is 3; When 2.0 < Ro, the maturity belongs to over - maturity, and the classification label is 4.

[0019] In the preferred solution, the specific steps in step S2 are: Including: calculating the weight scores for the feature vectors of the training set through the Relief algorithm, and taking the top five features with the highest scores as the conditional attribute variable set, the method is: a. Randomly select a sample R from the training set; b. Find the nearest neighbor sample H, called Near Hit, from the samples of the same class as sample R, and find the nearest neighbor sample M, called Near Miss, from the samples of different classes from sample R; c. Then update the weight of each feature: If the distance between R and Near Hit on a certain feature is less than the distance between R and Near Miss, it means that this feature is beneficial for distinguishing the nearest neighbors of the same class and different classes, then increase the weight of this feature; If the distance between R and Near Hit on a certain feature is greater than the distance between R and Near Miss, it means that the feature has a negative effect on distinguishing the nearest neighbors of the same and different classes, so the weight of the feature is reduced; d. Return to step c and execute until step c is executed N times, and use the weight of each feature after executing step c N times as the weight score of each feature.

[0020] In the preferred solution, the neural network model is constructed in step S3 as follows: S31. For each maturity level in the training set, calculate its prior probability , the prior probability can be obtained by calculating the proportion of each category in the training set; At the same time, for each feature and categories , calculate the conditional probability , it is necessary to perform statistical analysis based on the data in the training set; S32. In each iteration, the samples in the training set are input into the neural network model one by one, and for each input sample, the output result, i.e., the predicted maturity level probability distribution, is calculated according to the current network parameters; Then, according to the formula of the Bayesian machine classifier: , calculate the probability value corresponding to each category and compare it with the true label; in, are categories, i.e. different maturity levels, are the characteristics, i.e., aromatic hydrocarbon biomarker parameters; Indicates category The prior probability of Indicates that in the category Characteristics under the condition Probability of occurrence; By calculating the probability value corresponding to each category and then finding the category with the highest probability, it is considered that the item to be classified belongs to this category, thus obtaining a maturity prediction model based on Bayesian machine classifier learning; S33. Use the loss function to measure the difference between the predicted result and the true label, and adjust the weights and biases of the neural network through the back propagation algorithm to minimize the loss function. S34. When the performance of the model on the validation set reaches a satisfactory standard, the training is stopped to obtain a source rock and crude oil maturity prediction model based on Bayesian machine classifier learning. This model can be used to predict the maturity of new source rocks and crude oil samples.

[0021] In the preferred embodiment, the determination of maturity of source rock and crude oil in step S5 is specifically as follows: The source rock and crude oil aromatic biomarker parameters collected by the model sensor are used as the input of the Bayesian classifier, and the maturity level of the source rock and crude oil is I; Assume that the maturity level of source rock and crude oil is JS, the Lth conditional attribute variable is KL, and the conditional probability M of the sample maturity category JS is calculated according to the Bayesian formula:

[0022] Where S is the conditional probability that the conditional attribute variable KL belongs to the maturity category JS. Let the prior probabilities of JS and KL be P(JS) and P(KL) respectively, and calculate the posterior probability Q of KL belonging to JS. The formula is:

[0023] Then, the weight pl is assigned to the conditional attribute KL variable to obtain the weighted maturity feature, and the information entropy q(JS|KL) of KL about the maturity category JS is calculated. The formula is:

[0024] The information entropy of maturity feature KL attributed to each maturity category JS is statistically analyzed, and the maturity level corresponding to the highest information entropy q(JS|KL) is selected as the category of the conditional attribute variable, which is regarded as the maturity of source rock and crude oil.

[0025] The following is a more detailed supplement to the above: Step 5: Predict sample maturity based on the model In step S5, the determination of the maturity of source rock and crude oil is specifically as follows: the source rock and crude oil aromatic biomarker parameters collected by the model sensor are used as inputs based on the Bayesian machine classifier. These parameters are key feature data obtained after processing and screening in the previous steps, which can reflect the relevant characteristics of source rock and crude oil and are used to predict their maturity.

[0026] The number of maturity levels of source rocks and crude oil is I. Assuming that the maturity level of source rocks and crude oil is JS, the Lth conditional attribute variable is KL.

[0027] The conditional probability M of the sample maturity category JS is calculated according to the Bayesian formula: Here It represents the conditional probability that the conditional attribute variable KL belongs to the maturity category JS. The meaning of this formula is to comprehensively consider all possible maturity levels I and all conditional attribute variables N, and calculate the joint conditional probability that the sample belongs to a maturity category JS under these conditions.

[0028] Assume that the prior probabilities of JS and KL are P(JS) and P(KL) respectively. The prior probability P(JS) is the probability that the sample belongs to the maturity category JS when the specific conditional attribute variable KL is not considered. It can be estimated by the proportion of samples of each maturity level in the training set data. P(KL) is the probability of the conditional attribute variable KL itself appearing, which can also be obtained by statistical analysis of the training set data.

[0029] The formula for calculating the KL-attributed JS posterior probability Q is: The posterior probability Q reflects the updated probability value of the sample belonging to the maturity category JS when the conditional attribute variable KL is known. It combines the prior probability P(JS) and the conditional probability M as well as the information of the conditional attribute variable KL itself.

[0030] Then, the weight pl is assigned to the conditional attribute KL variable to obtain the weighted maturity feature. The allocation of the weight pl may be based on the evaluation of the importance of each conditional attribute variable, and different weights will affect the final maturity determination result.

[0031] Calculate the information entropy q(JS|KL) of KL about the maturity category JS, the formula is Information entropy is an indicator used to measure uncertainty. Here, it is used to measure the degree of uncertainty of the conditional attribute variable KL in determining the maturity category JS. By calculating the information entropy, we can better understand the role of each conditional attribute variable in the maturity determination.

[0032] The information entropy of the statistical maturity feature KL attributed to each maturity category JS. For each conditional attribute variable KL, it will have different information entropy values ​​under different maturity categories JS.

[0033] The maturity level corresponding to the highest information entropy q(JS|KL) is selected as the category of the conditional attribute variable. Because the highest information entropy means that under this maturity category, the conditional attribute variable KL provides the most information about maturity determination, so it is regarded as the maturity level of source rock and crude oil.

[0034] Example 2 Further illustrate with reference to Example 1, Figure 4 The structure shown in the figure shows that the technical method described in the present invention has been put into practice in the qualitative work of the maturity of the source rocks and crude oil of the Yanghugou Formation and the Uralik Formation in the Ordos Basin and has achieved good results. This case uses more than 80 source rocks and crude oil samples of the Yanghugou Formation and the Uralik Formation as materials. By using the 12 aromatic parameters mentioned above as training data, the equivalent maturity model constructed and trained basically matches the actual maturity of the source rocks and crude oil of the Yanghugou Formation and the Uralik Formation in the region, and realizes the accurate qualitative determination of the maturity of the source rocks and crude oil.

[0035] Table 1 Equivalent maturity of aromatics in the Yanghugou Formation and the Ulalike Formation in the Ordos Basin

[0036] Bayesian classification machine learning was used to evaluate the maturity of the Yanghugou Formation and the Ulalike Formation in the Ordos Basin. The results in Table 1 show that the classification results are consistent with the measured Ro conclusions, which has achieved good results in the region and can be used as a good basis for the maturity evaluation of source rocks and crude oil.

[0037] The above embodiments are only preferred technical solutions of the present invention and should not be regarded as limiting the present invention. The protection scope of the present invention shall be the technical solutions recorded in the claims, including equivalent replacement solutions of the technical features in the technical solutions recorded in the claims. That is, equivalent replacement improvements within this scope are also within the protection scope of the present invention.

Claims

1. A method for predicting source rock and crude oil maturity based on Bayesian classification, characterized by: The method includes: S1: Construct a dataset of biomarker parameters of source rocks and crude oils. Evaluate the maturity of samples by the vitrinite reflectance Ro of source rocks, and divide the obtained dataset into a training set, a test set, and a prediction set at a ratio of 7:2:

1. Among them, the training set and the test set are source rock samples with labels, and the prediction set is crude oil samples without labels; S2: Perform dataset feature selection processing on the dataset of biomarker parameters of source rocks and crude oils to obtain a set of conditional attribute variables; S3: Construct a neural network model. Train the neural network model based on the Bayesian machine classifier according to the training set and the set of conditional attribute variables to obtain a prediction model for the maturity of source rocks and crude oils learned based on the Bayesian machine classifier; S4: Use the test set to test the prediction model for the maturity of source rocks and crude oils learned based on the Bayesian machine classifier to obtain the accuracy rate. Determine whether the accuracy rate exceeds the set threshold. If it does not exceed, update the set of attribute variables and return to execute step S3. If it exceeds, execute step S5; S5: Obtain the biomarker parameters of source rocks and crude oils. Input the biomarker parameters of source rocks and crude oils into the prediction model for the maturity of source rocks and crude oils learned based on the Bayesian machine classifier for prediction to obtain the determination of the maturity of source rocks and crude oils.

2. The method for predicting source rock and crude oil maturity based on Bayesian classification according to claim 1, characterized in that: There are also multiple data features in the dataset of biomarker parameters of source rocks and crude oils; Four maturity evaluation criteria are defined according to the vitrinite reflectance Ro of source rocks. Specifically: Ro < 0.7, the maturity belongs to immature - low maturity, and the classification label is 1; 0.7 ≤ Ro < 1.3, the maturity belongs to mature, and the classification label is 2; 1.3 ≤ Ro ≤ 2.0, the maturity belongs to high maturity, and the classification label is 3; 2.0 < Ro, the maturity belongs to over - mature, and the classification label is 4.

3. The method for predicting source rock and crude oil maturity based on Bayesian classification according to claim 1, characterized in that: The specific steps in step S2 are: Include: Calculate the weight scores of the feature vectors of the training set through the Relief algorithm, and take the top five features with the highest scores as the set of conditional attribute variables. The method is: a. Randomly select a sample R from the training set; b. Find the nearest neighbor sample H, called Near Hit, from the samples of the same class as sample R, and find the nearest neighbor sample M, called Near Miss, from the samples of different classes from sample R; c. Then update the weight of each feature: If the distance between R and Near Hit on a certain feature is less than the distance between R and Near Miss, it means that this feature is beneficial for distinguishing the nearest neighbors of the same class and different classes, then increase the weight of this feature; If the distance between R and Near Hit on a certain feature is greater than the distance between R and Near Miss, it means that this feature plays a negative role in distinguishing the nearest neighbors of the same class and different classes, then reduce the weight of this feature; d. Return to execute step c until after executing step c N times, take the weights of each feature after executing step c N times as the weight scores of each feature.

4. The method for predicting source rock and crude oil maturity based on Bayesian classification according to claim 1, characterized in that: The specific construction of the neural network model in step S3 is: S31. For each maturity level in the training set, calculate its prior probability , the prior probability can be obtained by calculating the proportion of each category in the training set; At the same time, for each feature and categories , calculate the conditional probability , it is necessary to perform statistical analysis based on the data in the training set; S32. In each iteration, the samples in the training set are input into the neural network model one by one, and for each input sample, the output result, i.e., the predicted maturity level probability distribution, is calculated according to the current network parameters; Then, according to the formula of the Bayesian machine classifier: , calculate the probability value corresponding to each category and compare it with the true label; in, are categories, i.e. different maturity levels, are the characteristics, i.e., aromatic hydrocarbon biomarker parameters; Indicates category The prior probability of Indicates that in the category Characteristics under the condition Probability of occurrence; By calculating the probability value corresponding to each category and then finding the category with the highest probability, it is considered that the item to be classified belongs to this category, thus obtaining a maturity prediction model based on Bayesian machine classifier learning; S33. Use the loss function to measure the difference between the predicted result and the true label, and adjust the weights and biases of the neural network through the back propagation algorithm to minimize the loss function. S34. When the performance of the model on the validation set reaches a satisfactory standard, the training is stopped to obtain a source rock and crude oil maturity prediction model based on Bayesian machine classifier learning. This model can be used to predict the maturity of new source rocks and crude oil samples.

5. The method for predicting source rock and crude oil maturity based on Bayesian classification according to claim 1, characterized in that: The determination of the maturity of source rock and crude oil in step S5 is specifically as follows: The source rock and crude oil aromatic biomarker parameters collected by the model sensor are used as the input of the Bayesian classifier, and the maturity level of the source rock and crude oil is I; Assume that the maturity level of source rock and crude oil is JS, the Lth conditional attribute variable is KL, and the conditional probability M of the sample maturity category JS is calculated according to the Bayesian formula: Where S is the conditional probability that the conditional attribute variable KL belongs to the maturity category JS. Let the prior probabilities of JS and KL be P(JS) and P(KL) respectively, and calculate the posterior probability Q of KL belonging to JS. The formula is: Then, the weight pl is assigned to the conditional attribute KL variable to obtain the weighted maturity feature, and the information entropy q(JS|KL) of KL about the maturity category JS is calculated. The formula is: The information entropy of maturity feature KL attributed to each maturity category JS is statistically analyzed, and the maturity level corresponding to the highest information entropy q(JS|KL) is selected as the category of the conditional attribute variable, which is regarded as the maturity of source rock and crude oil.

Images