Shear wave travel time prediction method based on SSA-ELM algorithm

Abstract

The present invention discloses a shear wave travel time prediction method based on the SSA-ELM algorithm, specifically relating to the technical field of oil exploration and development. In the present invention, well logging curves with strong correlation with the shear wave travel time are selected as input data. After preprocessing, the training set and the test set are divided. A reservoir shear wave travel time prediction model with an extreme learning machine is established based on the training set. The sparrow search algorithm is used to find the optimal weights and optimal biases of the extreme learning machine, and the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm is obtained. The prediction model is used to predict the data in the test set, and the root mean square error between the predicted shear wave travel time value and the measured value is analyzed to verify the accuracy of the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm. The present invention optimizes the reservoir shear wave travel time prediction model based on the SSA-ELM algorithm, makes up for the problems of poor stability and insufficient generalization ability of the extreme learning machine, realizes the accurate prediction of the shear wave travel time value, and lays a foundation for the exploration and development of the reservoir.

Description

Technical Field

[0001] The present invention relates to the technical field of oil exploration and development, belongs to the category of intelligent logging interpretation, and particularly relates to a shear wave slowness prediction method based on the SSA-ELM algorithm. Background Technique

[0002] Shear wave velocity logging, as one of the important parameters in seismic exploration, reservoir development, and characterization, is widely used in lithology, porosity, and fluid estimation, four-dimensional seismic research, geomechanics, and wellbore stability research. However, due to cost issues, not all oil and gas wells can obtain shear wave logging data. Therefore, estimating shear wave velocity based on other logging data has become a major research focus.

[0003] Currently, common methods include empirical relationships, multiple regression analysis, and rock physics models. However, with the in-depth exploration and development, especially for shale formations containing sand-mud interlayers, the accuracy of existing methods is insufficient to meet the development needs. In recent years, data-driven machine learning methods have made certain progress in the field of geophysical exploration, and they have obvious advantages in deeply exploring the complex nonlinear relationships between different logging data. Based on this, it is urgent to propose a hybrid model of extreme learning machine (ELM) optimized by the sparrow search algorithm (SSA) to improve the prediction accuracy of traditional methods and single machine learning models to a certain extent, and lay a foundation for the next-step reservoir exploration and development. Summary of the Invention

[0004] The purpose of the present invention is to propose a shear wave slowness prediction method based on the SSA-ELM algorithm, which optimizes the extreme learning machine in the reservoir shear wave slowness prediction model by using the sparrow search algorithm, realizes high-precision prediction of the shear wave slowness value of logging, and lays a foundation for the exploration and development of the reservoir.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] The shear wave slowness prediction method based on the SSA-ELM algorithm includes the following steps:

[0007] s1. By performing a correlation analysis on the shear wave slowness curve and conventional logging curves, select the logging curves that are correlated with the shear wave slowness curve;

[0008] s2. After preprocessing the logging curves, use the measured values corresponding to each depth point on the logging curves as input data, and divide the input data into a training set and a test set to obtain training samples and test samples;

[0009] S3. Establish a reservoir shear wave slowness prediction model based on the training set. An extreme learning machine model is set in the reservoir shear wave slowness prediction model. The sparrow search algorithm is used to optimize the optimal weights and optimal biases of the extreme learning machine model, and the reservoir shear wave slowness prediction model optimized by the sparrow search algorithm is obtained;

[0010] S4. Input the test samples into the reservoir shear wave slowness prediction model optimized by the sparrow search algorithm. Use the reservoir shear wave slowness prediction model optimized by the sparrow search algorithm to predict the shear wave slowness of the test samples, and obtain the predicted values of the shear wave slowness of the test samples. By calculating the root mean square error between the measured values and the predicted values of the shear wave slowness in the test samples, verify the accuracy of the reservoir shear wave slowness prediction model optimized by the sparrow search algorithm.

[0011] Preferably, in the step S2, the preprocessing of the logging curves includes outlier rejection and normalization processing. Among them, the calculation formula for the normalization processing of the logging curves is:

[0012]

[0013] In the formula, X norm is the measured value of the depth point after normalization processing; X is the measured value of the depth point before normalization processing; X max is the maximum measured value of the depth point in the logging curve before normalization processing; X min is the minimum measured value of the depth point in the logging curve before normalization processing.

[0014] Preferably, in the step S3, the following steps are included:

[0015] S3.1. Construct an extreme learning machine model, and use the root mean square error between the measured values and the predicted values of the shear wave slowness in the training samples as the fitness function of the sparrow search algorithm, which is used to calculate the fitness of each sparrow in the sparrow population;

[0016] S3.2. Initialize the sparrow search algorithm, set the initial scale of the sparrow population and the maximum number of iterations iter max , and then set the alarm value, safety value, and the ratio of discoverers to followers in the sparrow population;

[0017] S3.3. Calculate the fitness of each sparrow in the sparrow population and sort them to determine the position of the sparrow individual with the best fitness and select discoverers, alarmers, and followers in the sparrow population;

[0018] S3.4. Calculate the alarm value of the sparrow population and update the positions of the discoverers in the sparrow population;

[0019] S3.5. Update the followers according to the updated results of the discoverers;

[0020] S3.6. Select the sparrows that are aware of danger in the sparrow population as the alarmers and update the positions of the alarmers;

[0021] S3.7. Calculate the average fitness of all sparrows in the sparrow population. If the individual fitness of the followers is less than the average fitness of all sparrows, perform wavelet transform on the followers; otherwise, do not process the followers.

[0022] S3.8. Determine whether to continue the loop update according to the number of iterations. If the current number of iterations is less than the maximum number of iterations, return to step S3.4; if the current number of iterations has reached the maximum number of iterations, output the result of the sparrow search as the initial weights and thresholds of the extreme learning machine model, and obtain the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm.

[0023] Preferably, in the step S3.4, the position update formula of the discoverers is shown in Equation (2):

[0024]

[0025] where t is the current number of iterations; iter max is the maximum number of iterations; is the position of the i-th sparrow individual in the j-th dimension at the t-th iteration; is the position of the discoverers after the t-th iteration update; α is a uniformly distributed random number between (0, 1); Q is a random number conforming to the standard normal distribution; L is a 1×d matrix with all internal elements taking values of 1; R2 is the alarm value, R2 ∈ [0, 1]; ST is the safety value, ST ∈ [0.5, 1]; when R2 < ST, it means that there are no predators or other dangers around the alarmers, the search environment is safe, and the discoverers continue to conduct extensive searches; when R2 ≥ ST, predators appear, and the sparrow population exhibits anti-predation behavior, and the discoverers quickly move to a safe area to continue predation.

[0026] Preferably, in the step S3.5, the position update formula of the followers is shown in Equation (3):

[0027]

[0028] where is the position of the followers after the t-th iteration update; is the optimal position of the current discoverers; is the current global optimal position; is the current global worst position; A is a 1×d matrix with internal elements randomly assigned to 1 or -1, A + = A T (AA T ) –1 , AT is the transpose matrix of A.

[0029] Preferably, in the step s3.6, the position update formula of the alarmist is shown in Equation (4):

[0030]

[0031] In the formula, is the position of the alarmist after the t-th iterative update; β is the step size control parameter, set as a random number of a normal distribution with a mean of 0 and a variance of 1; K is a random number between [-1, 1], representing the moving direction of the sparrow; f i represents the fitness of the i-th sparrow individual; f g represents the best fitness of the sparrow individuals in the current sparrow population; f w represents the worst fitness of the sparrow individuals in the current sparrow population; ε is a constant used to avoid the denominator being zero.

[0032] Preferably, in the step s4, the calculation formula of the root mean square error RMSE between the measured value and the predicted value of the shear wave travel time difference in the test sample is:

[0033]

[0034] In the formula, N is the number of training samples; y i is the shear wave travel time difference of the training sample; p i is the shear wave travel time difference calculated by the extreme learning model.

[0035] The present invention has the following beneficial effects:

[0036] The present invention uses the sparrow search algorithm to optimize the extreme learning machine in the reservoir shear wave travel time prediction model, obtains the optimal weight and optimal bias of the extreme learning machine, makes up for the problems of poor stability and insufficient generalization ability of the extreme learning machine due to random allocation, enhances the global search ability of the sparrow search algorithm, and improves the optimization ability and stability of the sparrow search algorithm.

[0037] The present invention uses the curve data with strong correlation with the shear wave travel time difference as the training set to train the extreme learning machine in the reservoir shear wave travel time prediction model, deeply mines the internal relationship between each logging parameter value and the shear wave travel time difference by using the hybrid SSA-ELM model, establishes a reservoir shear wave travel time prediction model, improves the prediction accuracy of the shear wave travel time difference, is conducive to accurately obtaining the shear wave velocity of the reservoir, and lays a foundation for the engineering evaluation and fluid identification of the reservoir. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 is the flow chart of the shear wave travel time prediction method based on the SSA-ELM algorithm of the present invention.

[0039] Figure 2 In the embodiment, it is the convergence situation of the fitness with the number of iterations during the optimization process using the SSA-ELM algorithm.

[0040] Figure 3 In the embodiment, it is the root mean square error between the measured value and the predicted value of the shear wave travel time. Specific implementation manner

[0041] The following takes a shale well in the Songliao Basin as an example in combination with the accompanying drawings to further illustrate the specific implementation manner of the present invention:

[0042] Taking this shale oil well in the Songliao Basin as an example, a shear wave travel time prediction method based on the SSA-ELM algorithm proposed by the present invention is used to predict the shear wave travel time value, as Figure 1 shown, specifically including the following steps:

[0043] s1. Obtain the logging data of the shale oil well. By analyzing the correlation between the conventional logging curves and the shear wave travel time, select the measured values of the longitudinal wave travel time curve AC, the natural gamma curve GR, the neutron porosity curve, and the compensated density curve DEN as input data. There is a good correlation between these curves and the shear wave travel time, and they can be used to predict the shear wave travel time value.

[0044] s2. Preprocess the obtained longitudinal wave travel time curve AC, natural gamma curve GR, neutron porosity curve, and compensated density curve DEN. Remove the outliers on each logging curve respectively and normalize the logging curves using formula (1). Use the measured values corresponding to each depth point on the preprocessed logging curves as input data, and further divide the input data into a training set and a test set. Among them, the training set contains 6000 training samples, and the test set contains 199 test data. The test data are all taken from the shale oil reservoir in the first member of the Qingshankou Formation.

[0045] s3. Based on the MATLAB software platform, train the reservoir shear wave travel time prediction model according to the training set in combination with the sparrow search algorithm. The reservoir shear wave travel time prediction model is set with an extreme learning machine model. Use the sparrow search algorithm to optimize the optimal weight and optimal bias of the extreme learning machine model to obtain the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm, specifically including the following steps:

[0046] s3.1. Construct an extreme learning machine model. The extreme learning machine model contains 2 hidden layer neurons. Use the root mean square error between the measured value and the predicted value of the shear wave travel time in the training samples as the fitness function of the sparrow search algorithm to calculate the fitness of each sparrow in the sparrow population.

[0047] S3.2. Initialize the sparrow search algorithm, set the initial size of the sparrow population to 10 and the maximum number of iterations iter max to 200 times, and then set the alarm value, safety value, and the ratio of discoverers to followers in the sparrow population.

[0048] S3.3. Calculate the fitness of each sparrow in the sparrow population and sort them to determine the position of the sparrow individual with the best fitness. And select discoverers, alarmers, and followers from the sparrow population.

[0049] S3.4. Calculate the alarm value of the sparrow population and update the position of the discoverers in the sparrow population, as shown in Equation (2):

[0050]

[0051] In the formula, t is the current iteration number; iter max is the maximum number of iterations; is the position of the i-th sparrow individual in the j-th dimension at the t-th iteration; is the position of the discoverer after the t-th iteration update; α is a uniform random number between (0,1); Q is a random number conforming to the standard normal distribution; L is a 1×d matrix with all internal elements taking the value of 1; R2 is the alarm value, R2∈[0, 1]; ST is the safety value, ST∈[0.5,1]; when R2 < ST, it means there are no predators or other dangers around the alarmers, and the search environment is safe, and the discoverers continue to conduct extensive searches; when R2 ≥ ST, predators appear, and the sparrow population exhibits anti-predation behavior, and the discoverers quickly move to a safe area to continue predation.

[0052] S3.5. Update the followers according to the update results of the discoverers, as shown in Equation (3):

[0053] The position update formula for the followers is as shown in Equation (3):

[0054]

[0055] In the formula, is the position of the followers after the t-th iteration update; is the optimal position of the current discoverer; is the current global optimal position; is the current global worst position; A is a 1×d matrix with internal elements randomly assigned to 1 or -1, A + = A T (AA T ) –1 , A T is the transpose matrix of A.

[0056] S3.6. Select sparrows that are aware of danger in the sparrow population as alarmers, and update the positions of the alarmers as shown in Equation (4):

[0057]

[0058] In the formula, is the position of the alarmers after the t-th iteration update; β is the step size control parameter, set as a random number of normal distribution with a mean of 0 and a variance of 1; K is a random number between [-1, 1], representing the moving direction of the sparrows; f i represents the fitness of the i-th sparrow individual; f g represents the best fitness of the sparrow individuals in the current sparrow population; f w represents the worst fitness of the sparrow individuals in the current sparrow population; ε is a constant used to avoid the denominator being zero.

[0059] S3.7. Calculate the average value of the fitness of all sparrows in the sparrow population. If the individual fitness of the followers is less than the average value of the fitness of all sparrows, perform wavelet transform on the followers, otherwise do not process the followers.

[0060] S3.8. Determine whether to continue the loop update according to the number of iterations. If the current number of iterations is less than the maximum number of iterations, return to step S3.4; if the current number of iterations has reached the maximum number of iterations, output the result of the sparrow search as the initial weights and thresholds of the extreme learning machine model, and obtain the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm.

[0061] Figure 2 The following shows the convergence of the fitness value with the number of iterations when training the extreme learning machine model using the sparrow search algorithm in this embodiment.

[0062] S4. Input the test samples into the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm, use the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm to predict the shear wave travel time of the test samples, obtain the predicted values of the shear wave travel time of the test samples, and calculate the root mean square error between the measured value and the predicted value of the shear wave travel time in the test samples using Equation (5), as Figure 3 shown. By comparing the measured values and predicted values of the shale shear wave travel time in the test set, it is obtained that the prediction results are basically consistent with the changes in the actual shear wave logging curve. The root mean square error between the predicted value and the actual measured value is 21.87. The predicted shear wave travel time value using the method of the present invention is relatively close to the actual measured value, and the prediction effect is good, verifying the accuracy of the shear wave travel time prediction method based on the SSA-ELM algorithm in the present invention.

[0063] Certainly, the above description is not a limitation of the present invention, and the present invention is not limited to the above examples. Changes, modifications, additions, or substitutions made by those skilled in the art within the scope of the essence of the present invention should also fall within the protection scope of the present invention.

Claims

1. A shear wave travel time prediction method based on the SSA-ELM algorithm, characterized in that, It includes the following steps: S1. By performing a correlation analysis on the shear wave travel time curve and conventional logging curves, select the logging curves that are correlated with the shear wave travel time curve; S2. After preprocessing the logging curves, use the measured values corresponding to each depth point on the logging curves as input data, and divide the input data into a training set and a test set to obtain training samples and test samples; S3. Based on the training set, establish a reservoir shear wave travel time prediction model. An extreme learning machine model is set in the reservoir shear wave travel time prediction model. Use the sparrow search algorithm to optimize the optimal weights and optimal biases of the extreme learning machine model to obtain the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm; S4. Input the test samples into the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm. Use the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm to predict the shear wave travel time of the test samples to obtain the predicted values of the shear wave travel time of the test samples. By calculating the root mean square error between the measured values and the predicted values of the shear wave travel time in the test samples, verify the accuracy of the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm; In the step S3, it includes the following steps: S3.

1. Construct an extreme learning machine model. Use the root mean square error between the measured values and the predicted values of the shear wave travel time in the training samples as the fitness function of the sparrow search algorithm to calculate the fitness of each sparrow in the sparrow population; S3.2 Initialize the sparrow search algorithm, set the initial size of the sparrow population and the maximum number of iterations iter max , then set the alarm value, safety value, and the ratio of discoverers to followers in the sparrow population; S3.

3. Calculate the fitness of each sparrow in the sparrow population and sort them to determine the position of the sparrow individual with the best fitness. Then select discoverers, alarmists, and followers from the sparrow population. S3.

4. Calculate the alarm value of the sparrow population and update the positions of the discoverers in the sparrow population; S3.

5. Update the followers according to the updated results of the discoverers; S3.

6. Select the sparrows that are aware of danger in the sparrow population as the alarmers and update the positions of the alarmers; S3.

7. Calculate the average value of the fitness of all sparrows in the sparrow population. If the individual fitness of the followers is less than the average value of the fitness of all sparrows, perform wavelet transform on the followers, otherwise do not process the followers; S3.

8. Determine whether to continue the loop update according to the number of iterations. If the current number of iterations is less than the maximum number of iterations, return to step S3.4; if the current number of iterations has reached the maximum number of iterations, output the result of the sparrow search as the initial weights and thresholds of the extreme learning machine model to obtain the reservoir shear wave travel time prediction model optimized by the sparrow search algorithm.

2. The shear wave travel time prediction method based on the SSA-ELM algorithm according to claim 1, characterized in that, In the step S2, the preprocessing of the logging curves includes outlier removal and normalization processing. Among them, the calculation formula for the normalization processing of the logging curves is: where X norm is the measured value of the depth point after normalization; X is the measured value of the depth point before normalization; X max is the maximum measured value of the depth point in the logging curve before normalization; X min is the minimum measured value of the depth point in the logging curve before normalization.

3. The shear wave travel time prediction method based on the SSA-ELM algorithm according to claim 1, characterized in that, In the step S3.4, the position update formula of the discoverers is shown in Equation (2): where \(t\) is the current iteration number; iter max is the maximum iteration number; is the position of the \(i\)-th sparrow individual in the \(j\)-th dimension at the \(t\)-th iteration; is the position of the discoverer after the \(t\)-th iteration update; \(\alpha\) is a uniformly distributed random number between \((0, 1)\); \(Q\) is a random number conforming to the standard normal distribution; \(L\) is a \(1\times d\) matrix with all elements inside being \(1\); \(R_2\) is the alarm value, \(R_2\in[0, 1]\); \(ST\) is the safety value, \(ST\in[0.5, 1]\); when \(R_2 \lt ST\), it means there are no predators or other dangers around the alarm caller, the search environment is safe, and the discoverer continues to conduct extensive searches; when \(R_2\geq ST\), a predator appears, the sparrow population exhibits anti-predation behavior, and the discoverer quickly moves to a safe area and continues to hunt.

4. The shear wave travel time prediction method based on the SSA-ELM algorithm according to claim 1, characterized in that, In the step S3.5, the position update formula of the followers is shown in Equation (3): Wherein, is the position of the follower after the t-th iterative update; is the optimal position of the current discoverer; is the current global optimal position; is the current global worst position; A is a 1×d matrix with internal elements randomly assigned to 1 or -1, A + = A T (AA T ) –1 , A T is the transpose matrix of A.

5. The shear wave travel time prediction method based on the SSA-ELM algorithm according to claim 1, characterized in that, In the step S3.6, the position update formula of the alarmers is shown in Equation (4): wherein is the position of the alarmist after the t-th iterative update; β is the step size control parameter, set as a random number of a normal distribution with a mean of 0 and a variance of 1; K is a random number between [-1, 1], representing the moving direction of the sparrow; f i represents the fitness of the i-th sparrow individual; f g represents the best fitness of the sparrow individuals in the current sparrow population; f w represents the worst fitness of the sparrow individuals in the current sparrow population; ε is a constant used to avoid a zero denominator.

6. The shear wave travel time prediction method based on the SSA-ELM algorithm according to claim 1, characterized in that, In the step S4, the calculation formula for the root mean square error RMSE between the measured values and the predicted values of the shear wave travel time in the test samples is: where N is the number of training samples; y i is the shear wave time difference of the training sample; p i is the shear wave time difference calculated by the extreme learning machine model.

Images