Inematic rock reservoir fluid identification method based on cascade forest

Abstract

The invention discloses an igneous rock reservoir fluid identification method based on a cascade forest, and relates to the field of rocks, the igneous rock reservoir fluid identification method based on the cascade forest comprises the following steps: step 1, establishing an igneous rock reservoir fluid identification pre-training model data set, and preprocessing data; compared with the prior art, the method has the advantages that the result-oriented parameter optimization process is avoided, the data set is balanced under the condition that the data scale is kept, and the classification accuracy of the unbalanced logging data set is improved; the method has the advantages that the applicability is higher; according to the method, reservoir fluid identification work can be efficiently and accurately completed in igneous rock stratums with complex fluid properties; according to the invention, the prediction rate and the back-judgment rate of fluid identification are improved; the problem of data imbalance in the field of igneous rock fluid identification is effectively solved, and the effect and applicability of the model are improved.

Description

Technical Field

[0001] The invention relates to the field of rocks, and in particular to an igneous rock reservoir fluid identification method based on cascade forests. Background Art

[0002] Igneous oil and gas reservoirs have huge reserves and considerable single-well production, so they have extremely high exploration and development value. However, igneous reservoirs also face many challenges, such as variable lithology and lithofacies, complex reservoir space, severe heterogeneity, complex conductivity mechanism, and difficulty in physical property evaluation and fluid identification. Compared with conventional methods, intelligent algorithms have advantages in data dependence, method complexity and calculation accuracy. Many intelligent methods have been applied to reservoir fluid identification. Ensemble learning algorithm is a method of combining several weak learners into a strong learner. Compared with other intelligent algorithms, it improves the model effect while highlighting the characteristics of weak learners.

[0003] However, the imbalance of data sets from different well logs results in low accuracy in igneous fluid identification, which needs to be improved. Summary of the invention

[0004] The object of the present invention is to provide a cascade forest-based igneous reservoir fluid identification method to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for identifying fluid in an igneous reservoir based on a cascade forest comprises the following steps: Step 1, establishing a pre-trained model data set for fluid identification in igneous reservoirs and pre-processing the data; Step 2, resampling and rebalancing the igneous reservoir fluid identification pre-training model data set to improve the quality of the data set; Step 3: Based on the rebalanced training set, cascade forests are used for training to generate an igneous fluid prediction model. The data to be identified is then input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, which are then input into the igneous fluid prediction model to obtain the final fluid prediction results, thus completing the igneous reservoir fluid prediction.

[0006] As a further solution of the present invention: Step 1 specifically comprises: Step 11, establishing a pre-trained model data set for igneous reservoir fluid identification including well logging data and igneous fluid characteristic data; Step 12, completing the standardization processing of the igneous reservoir fluid identification pre-training model data set.

[0007] As a further solution of the present invention: in step 11, the pre-trained model data set for igneous reservoir fluid identification includes logging data of different fluid characteristic samples; wherein the logging data of different fluid characteristic samples include neutron logging (CNL), density logging (DEN), natural gamma logging (GR), micro lateral resistivity (RLLL), and deep lateral resistivity (RLLD).

[0008] As a further solution of the present invention: in step 12, in order to reduce the impact of data distribution differences on the model, the igneous reservoir fluid identification pre-training model data set is standardized so that the data set obeys the standard normal distribution. The standardization process is as follows: (1); in, represents the standardized data, Represents the original data, represents the mean of the data, and s represents the standard deviation of the data set.

[0009] As a further solution of the present invention: Step 2 specifically includes: Step 21, resampling the igneous reservoir fluid identification pre-training model data set using an adaptive multi-objective group fusion method; Step 22, rebalancing the igneous reservoir fluid identification pre-training model data set.

[0010] As a further solution of the present invention: in step 21, AMSCO (an adaptive rebalancing model) is used as an adaptive multi-objective group fusion method to resample the igneous reservoir fluid identification pre-trained model data set; The core optimizer of AMSCO is particle swarm optimization (PSO). During the iteration process of particle swarm optimization (PSO), each particle calculates and updates the particle velocity and position. The mathematical process is as follows: (2); (3); in, is a non-negative value and is the inertia weight, is the current spatial dimension, is the current particle number, is the current iteration time, and is a non-negative constant and a velocity factor, and is a random value between 0 and 1, is the particle velocity. In the early stages of the iteration, the velocity and position of each particle are random. is the fitness index, is the collection of all particles.

[0011] As a further solution of the present invention: In step 22, AMSCO constructs two subgroups, optimizes the majority class by group instance selection (SIS), and the minority class by optimized synthetic minority instance oversampling technology (OSMOTE). AMSCO couples the two optimizations together through the proposed rebalancing method as a unified iterator, which gradually enhances the data mixing from the two group optimizations by iterating the optimization results of the crossover two group optimizations. In the process of cooperative evolution, there are three types of evaluation rules, namely ER1, ER2 and ER3: ER1: (4); ER2: (5); ER3: (6).

[0012] As a further solution of the present invention: in step 3, based on the rebalanced training set, training is performed using an adaptive multi-objective group fusion cascade forest to generate an igneous fluid prediction model. During actual training, the training set needs to be divided into two parts according to the labels for training, namely, polyhydrocarbons and polyhydrocarbons; polyhydrocarbons include low oil-yielding layers, water layers, oil-bearing water layers, dry layers and low water-yielding layers; polyhydrocarbons include oil layers and gas layers; next, polyhydrocarbons and polyhydrocarbons are trained separately to convert the multi-classification scenes into multiple binary classification scenes of a single fluid and other fluids; the fluid type with the highest evaluation index reaching the threshold is eliminated, and the preprocessing and fluid identification process are performed again until all fluids are effectively distinguished; the data to be identified is input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, and then input into the igneous fluid prediction model to obtain the final fluid prediction result, thereby completing the igneous reservoir fluid prediction.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: the present invention not only avoids the result-oriented parameter optimization process, but also balances the data set while maintaining the data scale, thereby improving the classification accuracy of the unbalanced logging data set; the present invention can automatically adjust according to the complexity of the data set, and has stronger applicability; the present invention can efficiently and accurately complete the reservoir fluid identification work in igneous rock formations with complex fluid properties, and provide favorable support for reservoir exploration; the igneous fluid identification scheme proposed in the present invention improves the prediction rate and recall rate of fluid identification by optimizing the existing fluid identification process; by performing preprocessing such as standardization on the initial data set and data resampling based on adaptive multi-target group fusion, the data imbalance problem in the field of igneous fluid identification is effectively solved, and the effect and applicability of the model are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 This is a cross plot of conventional logging data for igneous reservoirs.

[0015] Figure 2 This is the decision boundary change diagram in the GR-CNL two-dimensional feature space before resampling.

[0016] Figure 3 This is the decision boundary change diagram in the GR-CNL two-dimensional feature space after resampling.

[0017] Figure 4 This is the result diagram of fluid identification in the simulated well.

[0018] Figure 5 This is the actual igneous rock formation fluid identification result diagram. DETAILED DESCRIPTION

[0019] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0020] A method for identifying fluid in an igneous reservoir based on a cascade forest comprises the following steps: Step 1, establishing a pre-trained model data set for fluid identification in igneous reservoirs and pre-processing the data; Step 2, resampling and rebalancing the igneous reservoir fluid identification pre-training model data set to improve the quality of the data set; Step 3: Based on the rebalanced training set, cascade forests are used for training to generate an igneous fluid prediction model. The data to be identified is then input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, which are then input into the igneous fluid prediction model to obtain the final fluid prediction results, thus completing the igneous reservoir fluid prediction.

[0021] In this embodiment: Step 1 specifically includes: Step 11, establishing a pre-trained model data set for igneous reservoir fluid identification including well logging data and igneous fluid characteristic data; Step 12, completing the standardization processing of the igneous reservoir fluid identification pre-training model data set.

[0022] As a further solution of the present invention: please refer to Figure 1In step 11, the igneous reservoir fluid identification pre-training model data set includes logging data of different fluid characteristic samples; wherein the logging data of different fluid characteristic samples include neutron logging (CNL), density logging (DEN), natural gamma logging (GR), micro lateral resistivity (RLLL), and deep lateral resistivity (RLLD).

[0023] The conventional well logging crossplot matrix for igneous reservoirs is shown in Figure 1 , including dry layers, low oil-yield layers, oil layers, water layers, low water-yield layers, gas layers, etc.

[0024] In this embodiment: In step 12, in order to reduce the impact of data distribution differences on the model, the igneous reservoir fluid identification pre-training model data set is standardized so that the data set obeys the standard normal distribution. Because the data set used is tabular data, each column of data is standardized separately. The standardization process of each column of data is as follows: (1); in, represents the mean of the data, and s represents the standard deviation of the data set.

[0025] In this embodiment: Step 2 specifically includes: Step 21, resampling the igneous reservoir fluid identification pre-training model data set using an adaptive multi-objective group fusion method; Step 22, rebalancing the igneous reservoir fluid identification pre-training model data set.

[0026] In this embodiment: in step 21, Conventional well logging data sets are typical unbalanced data sets. Since in an unbalanced data set, the information from the majority class exceeds the information from the minority class, the classifier will become overfitting the majority class. In view of the imbalance problem of such conventional well logging curve data, the present invention uses the AMSCO method to regenerate a balanced data set.

[0027] AMSCO is an adaptive rebalancing model (swarm fusion), the overall idea of ​​which is to smash the original data set and use a combination of qualified instances selected by two swarm optimization processes to solve the imbalanced classification problem. During the rebalancing process, the parameters are automatically and adaptively controlled. The core optimizer of AMSCO is particle swarm optimization (PSO), which is a classic swarm intelligence algorithm that imitates the foraging process of birds. Compared with other intelligent algorithms, it has the advantages of easy implementation, faster convergence speed and fewer parameters.

[0028] AMSCO (an adaptive rebalancing model) is used as an adaptive multi-objective population fusion method to resample the pre-trained model dataset for fluid identification in igneous reservoirs; During the iteration process of particle swarm optimization (PSO), each particle calculates and updates the particle velocity and position. The mathematical process is as follows: (2); (3); in, is a non-negative value and is the inertia weight, is the current spatial dimension, is the current particle number, is the current iteration time, and is a non-negative constant and a velocity factor, and is a random value between 0 and 1, is the particle velocity. In the early stages of the iteration, the velocity and position of each particle are random. is the fitness index, is the collection of all particles.

[0029] In this embodiment: In step 22, AMSCO constructs two subgroups, optimizes the majority class by group instance selection (SIS), and the minority class by optimized synthetic minority instance oversampling technique (OSMOTE). AMSCO couples the two optimizations together through the proposed rebalancing method as a unified iterator, which gradually enhances the data mixing from the two group optimizations by iterating the optimization results of the crossover two group optimizations. In the process of cooperative evolution, there are three types of evaluation rules, namely ER1, ER2 and ER3: ER1: (4); ER2: (5); ER3: (6).

[0030] in Represents particles The best fit, Represents particles The best fitness of the subgroup, Represents particles The best fitness of the group. In the above process, each subgroup divides and searches the space in parallel and shares the best information to find the best solution in the shortest time.

[0031] See also Figure 2 and Figure 3In order to intuitively compare the classification effect of the rebalanced data set, the GR-CNL intersection diagram was made using the natural gamma (GR) and neutron porosity (CNL) data with the highest feature importance ranking. The SVM method is used as a learner to show the changes in the decision boundary of the two-dimensional feature space. In the figure, the blue scattered points represent the majority class dry layer data, the blue background is the dry layer classification range, the red scattered points represent the minority class oil layer data, and the red background is the oil layer classification range. After rebalancing, the oil layer classification range becomes larger and the decision boundary moves toward the dry layer. At the same time, the classification accuracy rate increased from 76.596% to 82.879%, and the F1 coincidence rate remained almost unchanged. This shows that after the AMSCO rebalanced data set, the minority class range becomes larger, the decision boundary is compressed to the majority class, and the classification accuracy rate is improved.

[0032] In this embodiment: In step 3, based on the rebalanced training set, the adaptive multi-objective group fusion cascade forest is used for training to generate an igneous fluid prediction model. In actual training, the training set needs to be divided into two parts according to the labels for training, namely, multi-hydrocarbons and minor hydrocarbons; minor hydrocarbons include low oil-yielding layers, water layers, oil-bearing water layers, dry layers and low water-yielding layers; multi-hydrocarbons include oil layers and gas layers; then, multi-hydrocarbons and minor hydrocarbons are trained separately to convert the multi-classification scenes into multiple binary classification scenes of a single fluid and other fluids; the fluid type with the highest F1 compliance rate reaching the threshold is eliminated, and the preprocessing and fluid identification process are performed again until all fluids are effectively distinguished; the data to be identified is input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, and then input into the igneous fluid prediction model to obtain the final fluid prediction result, thereby completing the igneous reservoir fluid prediction.

[0033] The final igneous fluid identification result can be obtained by putting the data to be identified into the resampled igneous reservoir fluid identification pre-training model. In order to verify the igneous fluid identification effect of this method, a simulated well was constructed using the validation set data to analyze the fluid identification effect, and fluid identification was performed in the actual igneous formation.

[0034] See also Figure 4 It can be clearly seen from the simulated well fluid identification result graph that the adaptive multi-target group fusion cascade forest method produced identification errors in the oil layer and the junction of the water layer and the water-bearing oil layer. Combined with the conventional curve, the curve has corresponding fluctuations in the incorrectly identified layer segment. In addition, the identification of other fluids is very accurate. At the same time, the identification results of the control group are inferior to the fluid identification method proposed in the present invention, including XDF (cascade forest), XGBoost (extreme gradient boosting), and SVM (support vector machine) in the control group.

[0035] See also Figure 5, there are some differences between the actual igneous rock formation fluid identification results and the formation fluid category boundary depth of the label channel, and other places are basically consistent. The reason for this phenomenon is that the difference between adjacent logging data is small, resulting in identification errors, which are acceptable. In general, the adaptive multi-objective group fusion cascade forest proposed by the present invention has a good application effect in wells.

[0036] The present invention first establishes a pre-trained model data set for fluid identification of igneous reservoirs and a training data set for a fluid prediction model, then uses the logging data in the data set as features and the fluid type as a label, uses cascade forests for training, and saves the mineral pre-trained model. Then, the AMSCO method is used to resample the training set, and based on the resampled training set, the cascade forest is used for training to generate an igneous fluid prediction model, and then the data to be identified is input into the resampled igneous reservoir fluid identification pre-trained model to generate fluid features, and then input into the igneous fluid prediction model to obtain the final fluid prediction result.

[0037] The present invention not only avoids the result-oriented parameter optimization process, but also balances the data set while maintaining the data scale, thereby improving the classification accuracy of the unbalanced logging data set; The present invention can automatically adjust according to the complexity of the data set and has stronger applicability; The present invention can efficiently and accurately complete reservoir fluid identification in igneous rock formations with complex fluid properties, providing favorable support for reservoir exploration; The igneous fluid identification scheme proposed in the present invention improves the prediction rate and the recall rate of fluid identification by optimizing the existing fluid identification process; By performing preprocessing such as standardization on the initial data set and resampling the data based on adaptive multi-target group fusion, the data imbalance problem in the field of igneous fluid identification is effectively solved, and the effect and applicability of the model are improved.

[0038] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, the embodiments should be considered as exemplary and non-restrictive in all respects.

[0039] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. A method for identifying fluid in igneous reservoirs based on cascade forests, characterized in that: The cascade forest-based igneous reservoir fluid identification method comprises the following steps: Step 1, establishing a pre-trained model data set for fluid identification in igneous reservoirs and pre-processing the data; Step 2, resampling and rebalancing the igneous reservoir fluid identification pre-training model data set to improve the quality of the data set; Step 3: Based on the rebalanced training set, cascade forests are used for training to generate an igneous fluid prediction model. The data to be identified is then input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, which are then input into the igneous fluid prediction model to obtain the final fluid prediction results, thus completing the igneous reservoir fluid prediction.

2. The igneous reservoir fluid identification method based on cascade forests according to claim 1 is characterized in that: Step 1 specifically includes: Step 11, establishing a pre-trained model data set for igneous reservoir fluid identification including well logging data and igneous fluid characteristic data; Step 12, completing the standardization processing of the igneous reservoir fluid identification pre-training model data set.

3. The igneous reservoir fluid identification method based on cascade forest according to claim 2 is characterized in that: In step 11, the igneous reservoir fluid identification pre-training model data set includes logging data of different fluid characteristic samples; wherein the logging data of different fluid characteristic samples include neutron logging, density logging, natural gamma logging, micro lateral resistivity, and deep lateral resistivity.

4. The igneous reservoir fluid identification method based on cascade forests according to claim 2 or 3, characterized in that: In step 12, in order to reduce the impact of data distribution differences on the model, the igneous reservoir fluid identification pre-training model data set is standardized so that the data set obeys the standard normal distribution. The standardization process is as follows: （1）； in, Represents the original data, represents the mean of the data, and s represents the standard deviation of the data set.

5. The igneous reservoir fluid identification method based on cascade forest according to claim 1, characterized in that: Step 2 specifically includes: Step 21, resampling the igneous reservoir fluid identification pre-training model data set using an adaptive multi-objective group fusion method; Step 22, rebalancing the igneous reservoir fluid identification pre-training model data set.

6. The igneous reservoir fluid identification method based on cascade forests according to claim 5 is characterized in that: In step 21, AMSCO is used as an adaptive multi-objective group fusion method to resample the igneous reservoir fluid identification pre-trained model dataset; The core optimizer of AMSCO is particle swarm optimization. During the iterative process of particle swarm optimization, each particle will calculate and update the particle velocity and position. The mathematical process is as follows: （2）； （3）； in, is a non-negative value and is the inertia weight, is the current spatial dimension, is the current particle number, is the current iteration time, and is a non-negative constant and a velocity factor, and is a random value between 0 and 1, is the particle velocity. In the early stages of the iteration, the velocity and position of each particle are random. is the fitness index, is the collection of all particles.

7. The igneous reservoir fluid identification method based on cascade forests according to claim 5 or 6, characterized in that: In step 22, AMSCO constructs two subgroups, optimizes the majority class through group instance selection, and optimizes the minority class through the optimized synthetic minority instance oversampling technique. AMSCO couples the two optimizations together through the proposed rebalancing method as a unified iterator, which gradually enhances the data mixing from the two group optimizations by iterating the optimization results of the crossover two group optimizations. In the process of cooperative evolution, there are three types of evaluation rules, namely ER1, ER2 and ER3: ER1 : (4); ER2 : (5); ER3: (6).

8. The igneous reservoir fluid identification method based on cascade forests according to claim 1, characterized in that: In step 3, based on the rebalanced training set, the igneous fluid prediction model can be generated by training based on the adaptive multi-objective group fusion cascade forest. In actual training, the training set needs to be divided into two parts according to the labels for training, namely, more hydrocarbons and less hydrocarbons; The less hydrocarbons include low oil-yielding layers, water layers, oil-water layers, dry layers, and low water-yielding layers; the more hydrocarbons include oil layers and gas layers; Next, the more hydrocarbons and less hydrocarbons are trained separately to convert the multi-classification scenes into multiple binary classification scenes of a single fluid and other fluids; The fluid type with the highest evaluation index reaching the threshold is eliminated, and the preprocessing and fluid identification process is performed again until all fluids are effectively distinguished; The data to be identified is input into the resampled igneous reservoir fluid identification pre-training model to generate fluid characteristics, and then input into the igneous fluid prediction model to obtain the final fluid prediction result, thus completing the igneous reservoir fluid prediction.

Images