Intelligent optimization method for formula parameters of temporary in-fracture bridging agent for fracturing

Abstract

The present application relates to the technical field of fracturing construction in oil exploitation, and is an intelligent optimization method for formula parameters of in-fracture temporary plugging agent for fracturing, which is performed according to the following steps: collecting data, filling in missing values, establishing a corresponding target block fracturing database, according to the corresponding target block fracturing database, combining a clustering algorithm of PCA principal component analysis, according to the clustering result, performing reservoir division, establishing a BP neural network regression model, according to the clustering result obtained by the clustering algorithm and the reservoir division, optimizing the formula parameters of in-fracture temporary plugging agent for fracturing, and determining the proportion of the amount of the temporary plugging agent. The present application comprehensively considers geological information of the target block and in-fracture temporary plugging fracturing construction condition information, uses pressure increase caused by in-fracture temporary plugging construction of a single well as an evaluation index, performs clustering division on reservoirs of the target block, establishes a regression model, and realizes targeted optimization of formula parameters of in-fracture temporary plugging agent.

Description

Technical Field

[0001] This invention relates to the field of fracturing construction technology in oil extraction, and is a method for intelligent optimization of the formulation parameters of temporary plugging agent for fracturing. Background Technology

[0002] The target block is a tight sandstone-conglomerate reservoir, primarily composed of sandstone and conglomerate, but it exhibits strong heterogeneity and complex geological conditions. However, the design parameters of the plugging agent formulation for the in-fracture temporary plugging operation in the target block at the oilfield site lacked differentiation and specificity, exhibiting a degree of arbitrariness and poor distinguishability. For example, in the two vertical wells of the BKQ block, despite significant differences in geological parameters such as minimum horizontal principal stress and fracturing fluid injection rate—parameters that significantly affect fracture morphology—the total amount and ratio of powdered and granular plugging agents used were extremely similar, failing to demonstrate specificity in the dosage and formulation of the plugging agent for in-fracture use. Consequently, the desired temporary plugging pressure increase was not achieved during the in-fracture temporary plugging operation.

[0003] Furthermore, current design practices for temporary plugging fracturing operations rely solely on simplistic charts depicting the total amount of plugging agent and the dynamic fracture width relationship when determining the plugging agent formulation parameters. This leads to significant overlap in the design results and allows for arbitrary alteration, further resulting in blind design of the plugging agent formulation parameters. These issues cause substantial economic waste during temporary plugging fracturing operations in most cases.

[0004] Based on the above issues, for heterogeneous reservoirs of tight sandstone and conglomerate, it is necessary to re-examine and analyze the geological information and in-fracture temporary plugging operation parameters of the target block. This will allow for the optimization of the in-fracture temporary plugging agent formulation parameters without increasing the cost of in-fracture temporary plugging fracturing and while achieving the desired temporary plugging pressure rise. Furthermore, it will be possible to predict the ideal temporary plugging pressure rise that can be achieved in the target block under the optimized formulation parameters. Summary of the Invention

[0005] This invention provides an intelligent optimization method for the formulation parameters of temporary plugging agent for fracturing, which overcomes the shortcomings of the prior art. It can optimize the formulation of the temporary plugging agent during fracturing operations, so as to optimize the ratio of different types of temporary plugging agents without wasting the total amount of temporary plugging agent used, and obtain better sealing strength in the fracture.

[0006] The technical solution of this invention is achieved through the following measures: A method for intelligent optimization of the formulation parameters of a temporary plugging agent for fracturing, comprising the following steps: First, collecting geological data, temporary plugging operation data, types of temporary plugging agents used under actual operation conditions, and pressure rise data of the target block, including but not limited to the geological data, temporary plugging operation data, and pressure rise data of the actual operation conditions; filling in missing values ​​in the acquired data to establish a corresponding target block fracturing database; Second, based on the corresponding target block fracturing database, using a clustering algorithm combined with PCA principal component analysis, and dividing the reservoir according to the clustering results; Third, establishing a BP neural network regression model for the geological data, temporary plugging operation data, and pressure rise data of the target block under actual operation conditions; Fourth, optimizing the formulation parameters of the temporary plugging agent for fracturing based on the clustering results and reservoir division obtained by the clustering algorithm, and determining the dosage ratio of the temporary plugging agent.

[0007] The following are further optimizations and / or improvements to the above-mentioned technical solution: The geological data mentioned above include, but are not limited to, lithology, pressure coefficient, Young's modulus, minimum horizontal principal stress, Poisson's ratio, reservoir temperature, formation pressure, effective porosity, effective permeability, filtration loss coefficient, and underground fluid viscosity; the temporary plugging construction condition data include, but are not limited to, pre-filled fluid volume and discharge volume.

[0008] The aforementioned types of temporary plugging agents include granular temporary plugging agents with a particle size of <1mm, granular temporary plugging agents with a particle size of 1mm to 3mm, granular temporary plugging agents with a particle size of 1mm to 5mm, granular temporary plugging agents with a particle size of 1mm to 7mm, and powdered temporary plugging agents.

[0009] The missing value imputation method mentioned above includes: first, imputing missing values ​​using a non-machine learning method that is as realistic as possible, where the non-machine learning method that is as realistic as possible includes imputation using neighboring well information; then, in the absence of neighboring well information, imputing missing values ​​using machine learning methods; and finally, using machine learning feature engineering methods to perform preprocessing operations for outlier identification.

[0010] When imputing missing values, if the missing rate is no greater than 5%, the median is used for imputation; if the missing rate is greater than 5%, random interpolation is used for imputation.

[0011] When identifying outliers, a box plot is used. If the outlier is caused by an error in data input, it is corrected to a normal value. If the outlier is too outrageous, it is treated as a missing value. Other outliers are temporarily retained, and a database of fracturing interstitial plugging agents for the corresponding target blocks is established.

[0012] In the fourth step above, the specific process of optimizing the formulation parameters of the fracture plugging agent includes: First, combining the geological data and temporary plugging operation data of the target block, inputting them into a BP neural network regression model to predict the plugging strength and pressure, and determining the first dimension parameter of the fracture plugging agent: the total amount of fracture plugging agent used; then, under the condition of determining the first dimension parameter of the fracture plugging agent: the total amount of fracture plugging agent used, increasing the proportion of particulate plugging agent, and inputting it again into the BP neural network model to predict the plugging pressure and intensity, and determining the second dimension parameter of the fracture plugging agent: the ratio of the two specifications of the fracture plugging agent used.

[0013] This invention collects and processes geological information, temporary plugging operation information, and pressure rise information of temporary plugging wells in the target block, and establishes regression models for the pressure rise and formulation parameters of the temporary plugging agent in the fracture. This allows for the determination of the formulation parameters of the temporary plugging agent in the fracture while simultaneously considering the economic cost of temporary plugging and the expected ideal pressure rise. Attached Figure Description

[0014] Appendix Figure 1 This is a structural diagram of the multi-layer neural network model used in this invention.

[0015] Appendix Figure 2 This is a diagram showing the reservoir classification results of New Well 1 in Embodiment 8 of the present invention.

[0016] Appendix Figure 3 This is a pressure change curve of the temporary plugging agent in the new well 1 of Embodiment 8 of the present invention after being input into the model according to the design range requirements.

[0017] Appendix Figure 4 This is a diagram of the proportioning design scheme for the optimized total amount of temporary plugging agent in the new well 1 of Embodiment 8 of the present invention.

[0018] Appendix Figure 5 This is a diagram showing the reservoir classification results of New Well 2 in Embodiment 9 of the present invention.

[0019] Appendix Figure 6 This is a pressure change curve of the temporary plugging agent in the new well 2 of Embodiment 9 of the present invention after being input into the model according to the design range requirements.

[0020] Appendix Figure 7 This is a diagram of the proportioning design scheme for the optimized total amount of temporary plugging agent in New Well 2 of Embodiment 9 of the present invention. Detailed Implementation

[0021] The present invention is not limited to the following embodiments, and the specific implementation can be determined according to the technical solution of the present invention and the actual situation.

[0022] The present invention will be further described below with reference to embodiments: Example 1: The intelligent optimization method for the formulation parameters of the intra-fracture plugging agent for fracturing is carried out according to the following steps: First, collect geological data, plugging operation data, types of intra-fracture plugging agents used under actual operation conditions, and pressure rise data of the target block, including but not limited to the target block geological data, plugging operation data, and pressure rise data of the actual operation conditions. Impute missing values ​​in the acquired data to establish a corresponding target block fracturing database. Second, based on the corresponding target block fracturing database, perform reservoir division using a clustering algorithm combined with PCA principal component analysis. Third, establish a BP neural network regression model for the geological data, plugging operation data, and pressure rise data of the target block under actual operation conditions. Fourth, optimize the formulation parameters of the intra-fracture plugging agent for fracturing based on the clustering results and reservoir division obtained by the clustering algorithm, and determine the dosage ratio of the plugging agent.

[0023] Example 2: As an optimization of the above examples, the geological data includes, but is not limited to, lithology, pressure coefficient, Young's modulus, minimum horizontal principal stress, Poisson's ratio, reservoir temperature, formation pressure, effective porosity, effective permeability, filtration loss coefficient, and underground fluid viscosity; the temporary plugging construction condition data includes, but is not limited to, pre-filled fluid volume and discharge volume.

[0024] In this invention, the geological information of the target block and the information on the temporary plugging construction conditions within the fracture are used to fit the pressure rise value of the temporary plugging within the fracture, which is obtained from actual temporary plugging fracturing construction within the fracture. This pressure rise value data is used as data preparation for the subsequent establishment of the regression model.

[0025] Meanwhile, to facilitate reservoir delineation, the collected geological information of the target blocks is used to perform clustering algorithms combined with PCA principal component analysis, which directly guides the optimization of the formulation parameters of the temporary plugging agent for fracturing.

[0026] Example 3: As an optimization of the above examples, the types of temporary plugging agents in the joint include particulate temporary plugging agents with a particle size of <1mm, particulate temporary plugging agents with a particle size of 1mm to 3mm, particulate temporary plugging agents with a particle size of 1mm to 5mm, particulate temporary plugging agents with a particle size of 1mm to 7mm, and powdered temporary plugging agents.

[0027] Example 4: As an optimization of the above embodiment, the missing value imputation method includes: first, imputing missing values ​​using a non-machine learning method that is as realistic as possible, wherein the non-machine learning method that is as realistic as possible includes imputing with neighboring well information; then, in the absence of neighboring well information, imputing missing values ​​using a machine learning method; and finally, performing preprocessing operations for outlier identification by comprehensively applying machine learning feature engineering methods.

[0028] Example 5: As an optimization of the above example, when imputing missing values, if the missing rate is no more than 5%, the median is used for imputation; if the missing rate is greater than 5%, random interpolation is used for imputation.

[0029] Example 6: As an optimization of the above example, when identifying outliers, a box plot is used. If the outlier is caused by a data input error, it is modified to a normal value. If the outlier error is too outrageous, it is treated as a missing value. Other outliers are temporarily retained, and a database of fracturing interstitial plugging agents for the corresponding target block is established.

[0030] Example 7: As an optimization of the above example, the specific process of optimizing the formulation parameters of the fracture plugging agent in the fourth step includes: First, combining the geological data and temporary plugging construction condition data of the target block, inputting them into the BP neural network regression model to predict the plugging strength pressure, and determining the first dimension parameter of the fracture plugging agent: the total amount of fracture plugging agent used; then, under the condition of determining the first dimension parameter of the fracture plugging agent: the total amount of fracture plugging agent used, increasing the proportion of particulate plugging agent, and inputting it again into the BP neural network model to predict the plugging pressure intensity, and determining the second dimension parameter of the fracture plugging agent: the ratio of the two specifications of the fracture plugging agent used.

[0031] When establishing a BP neural network regression model for the pressure rise during temporary plugging operations and the geological data and temporary plugging operation data of the target block, the geological data and temporary plugging operation data of the target block are used as input nodes of the neural network, and the pressure rise during temporary plugging operations within the fracture is used as the output node. The model is trained using oil well information collected in the target block. When the similarity between the test sample and the predicted value of the BP neural network model reaches 0.96 or higher, the model reaches the corresponding accuracy.

[0032] Example 8: The intelligent optimization method for the formulation parameters of the fracture plugging agent for fracturing is carried out according to the following steps: First, collect the fracturing data of 149 vertical wells with fracture plugging in the BKQ block, which has heterogeneous reservoir properties. This data includes, but is not limited to, lithology, pressure coefficient, Young's modulus, minimum horizontal principal stress, Poisson's ratio, reservoir temperature, formation pressure, effective porosity, effective permeability, filtration coefficient, underground fluid viscosity, pre-filled fluid volume, discharge volume, and the fracture plugging agent used. The fracture plugging agent types include four types: particulate plugging agent with particle size <1mm, particle size 1mm to 3mm, particle size 1mm to 5mm, and particle size 1mm to 7mm, and their respective dosage information. Collect the pressure rise of the fracture plugging strength during the fracture plugging construction process and establish a corresponding target block fracturing database. Second, by combining the clustering algorithm of PCA principal component analysis, it can be seen that the horizontal wells in this block are divided into two categories, A and B (clustering results are shown in Figure 1). Figure 2As shown), the main contributing factors for Class A wells are low stress difference, low Poisson's ratio, and low Young's modulus, which will form wider fractures. The main contributing factors for Class B wells are high stress difference, high Poisson's ratio, and high Young's modulus, which will form narrower fractures. It is determined that for Class A wells, which are prone to forming wider fractures, the proportion of particulate temporary plugging agent should be increased, and for Class B wells, which are prone to forming narrower fractures, the proportion of powdered temporary plugging agent should be increased. The third step is to establish a BP neural network regression model of the pressure rise during in-fracture temporary plugging under geological data, temporary plugging construction condition data, and actual construction conditions. The fourth step is to optimize the formulation parameters of the in-fracture temporary plugging agent for a new well 1 based on the clustering results and reservoir division obtained by the clustering algorithm. First, the clustering algorithm obtained by PCA principal component analysis is used to determine that it is a Class A oil well (e.g., Figure 2 As shown), it was determined that the proportion of particulate temporary plugging agent should be increased; in the fifth step, the total amount of temporary plugging agent designed for the new well 1 (360kg) was used as the upper limit, and half of the total amount (180kg) was used as the lower limit to determine the input value range of the total amount of temporary plugging agent. This range was then combined with geological information and information on the temporary plugging construction conditions within the fracture and input into the BP neural network regression model to predict the plugging strength and pressure. Initially, 360kg was used as the initial value, decreasing in steps of 20kg to 180kg, and the changes in the plugging strength and pressure were observed (e.g., ...). Figure 3 As shown in the figure), observing the curve, it was found that when the total amount of temporary plugging agent was reduced to 300 kg, the plugging strength pressure value decreased (as shown in the figure). Figure 3 As shown in the first "inflection point" in the diagram, the first dimension parameter of the fracture plugging agent for fracturing is determined and optimized—the total amount of fracture plugging agent used for fracturing (at this point, the total amount of plugging agent used is reduced from the designed 360kg to 300kg).

[0033] In this embodiment 8, since the new well 1 was previously determined to be a Class A well (e.g., ... Figure 2 As shown), under the condition that the total dosage is determined to be 300 kg, the proportion of particulate temporary plugging agent is increased, and the mixture is input into the BP neural network model again to predict the plugging pressure intensity (e.g. Figure 4 As shown), it was found that when the ratio of particulate temporary plugger to powdered temporary plugger was 3:1, the plugging strength pressure experienced a jump (as shown). Figure 4 As shown in the second "inflection point" in the diagram, the second dimension parameter of the fracture plugging agent for fracturing has been determined and optimized—the ratio of the two specifications of the fracture plugging agent for fracturing.

[0034] In summary, the initial design for the new well 1, with a total dosage of 360 kg and a ratio of granular to powdered plugging agent of 1:1, corresponding to a predicted pressure value of 9 MPa for the sealing strength within the fracture, has been optimized to a total dosage of 300 kg and a ratio of granular to powdered plugging agent of 3:1. At this point, the sealing pressure value within the fracture is 12 MPa, demonstrating a good optimization effect.

[0035] Example 9: The intelligent optimization method for the formulation parameters of the fracture plugging agent for fracturing is carried out according to the following steps: First, collect the fracturing data of 149 vertical wells with fracture plugging in the BKQ block, which has heterogeneous reservoir properties. This data includes, but is not limited to, lithology, pressure coefficient, Young's modulus, minimum horizontal principal stress, Poisson's ratio, reservoir temperature, formation pressure, effective porosity, effective permeability, filtration coefficient, underground fluid viscosity, pre-flush volume, discharge volume, and the fracture plugging agent used. The fracture plugging agent types include four types: particulate plugging agent with particle size <1mm, particle size 1mm to 3mm, particle size 1mm to 5mm, and particle size 1mm to 7mm, and their respective dosage information. Collect the pressure rise of the fracture plugging strength during the fracture plugging construction process and establish a corresponding target block fracturing database. Second, by combining the clustering algorithm of PCA principal component analysis, it can be seen that the horizontal wells in this block are divided into two categories, A and B (clustering results are shown in Figure 1). Figure 5 As shown), the main contributing factors for Class A wells are low stress difference, low Poisson's ratio, and low Young's modulus, which will form wider fractures. The main contributing factors for Class B wells are high stress difference, high Poisson's ratio, and high Young's modulus, which will form narrower fractures. It is determined that for Class A wells, which are prone to forming wider fractures, the proportion of particulate temporary plugging agent should be increased, and for Class B wells, which are prone to forming narrower fractures, the proportion of powdered temporary plugging agent should be increased. The third step is to establish a BP neural network regression model of the pressure rise during intra-fracture temporary plugging construction under geological data, temporary plugging construction condition data, and actual construction conditions. The fourth step is to optimize the formulation parameters of the intra-fracture temporary plugging agent for a certain new well 2 based on the clustering results and reservoir division obtained by the clustering algorithm. First, combined with the clustering algorithm obtained by PCA principal component analysis, it is known that it is a Class B oil well (e.g. Figure 5 As shown), it was determined that the proportion of particulate temporary plugging agent should be increased; in the fifth step, the total amount of temporary plugging agent designed for the new well 2 (360kg) was used as the upper limit, and half of the total amount (180kg) was used as the lower limit to determine the input value range of the total amount of temporary plugging agent. This range was then combined with geological information and information on the temporary plugging construction conditions within the fracture and input into the BP neural network regression model to predict the plugging strength and pressure. Initially, 360kg was used as the initial value, decreasing in steps of 20kg to 180kg, and the changes in the plugging strength and pressure were observed (e.g., ...). Figure 6 As shown in the figure), observing the curve, it was found that when the total amount of temporary plugging agent was reduced to 260 kg, the plugging strength pressure value decreased (as shown in the figure). Figure 6 As shown in the first "inflection point" in the diagram, the first dimension parameter of the fracture plugging agent for fracturing is determined and optimized—the total amount of fracture plugging agent used for fracturing (at this point, the total amount of plugging agent used is reduced from the designed 360kg to 300kg).

[0036] In this embodiment 9, since the new well 2 was previously determined to be a Class B well (e.g., ... Figure 5As shown), under the condition that the total dosage is determined to be 260 kg, the proportion of powdered temporary plugging agent is increased and then input into the neural network model again to predict the plugging pressure intensity (e.g. Figure 7 It was found that when the ratio of powder plugging agent to particulate temporary plugging agent was 2:1, the sealing strength pressure experienced a jump (e.g., Figure 7 As shown in the second "inflection point" in the diagram, the second dimension parameter of the fracture plugging agent for fracturing has been determined and optimized—the ratio of the two specifications of the fracture plugging agent for fracturing.

[0037] In summary, the initial design for the total amount of material used in well 2 was 360 kg, with a ratio of granular to powdered temporary plugging agent of 1:1, corresponding to a predicted pressure value of 13 MPa for sealing within the fracture. This has been optimized to a total amount of 260 kg, with a ratio of powdered to granular temporary plugging agent of 2:1. At this point, the sealing pressure value within the fracture is 16 MPa, demonstrating a good optimization effect.

[0038] In summary, this invention comprehensively considers, but is not limited to, geological information of the target block and information on the operation conditions of intra-fracture temporary plugging fracturing. It uses the pressure rise caused by intra-fracture temporary plugging in a single well as the evaluation index to cluster the oil reservoirs of the target block, establish a regression model, and achieve targeted optimization of the formulation parameters of the intra-fracture temporary plugging agent.

[0039] The above technical features constitute the embodiments of the present invention, which have strong adaptability and implementation effect. Unnecessary technical features can be added or removed according to actual needs to meet the needs of different situations.

Claims

1. A method for intelligent optimization of the formulation parameters of a temporary plugging agent for fracturing, characterized in that... Follow these steps: The first step involves collecting, but is not limited to, geological data of the target block, data on temporary plugging operations, the type of temporary plugging agent used in the actual operation, and pressure rise data. Missing values ​​are filled into the acquired data to establish a corresponding fracturing database for the target block. Geological data includes, but is not limited to, lithology, pressure coefficient, Young's modulus, minimum horizontal principal stress, Poisson's ratio, reservoir temperature, formation pressure, effective porosity, effective permeability, filtration coefficient, and underground fluid viscosity. Data on temporary plugging operations includes, but is not limited to, pre-flush fluid volume and discharge rate. The second step is to divide the reservoir based on the corresponding target block fracturing database and the clustering algorithm of PCA principal component analysis. The horizontal wells in the target block are divided into two categories, A and B. The main contributing factors of category A are low stress difference, low Poisson's ratio and low Young's modulus, while the main contributing factors of category B are high stress difference, high Poisson's ratio and high Young's modulus. The third step is to establish a BP neural network regression model for the geological data, temporary plugging construction conditions data, and pressure rise data under actual construction conditions of the target block. The geological data and temporary plugging construction conditions data of the target block are used as input nodes of the neural network, and the pressure rise during temporary plugging construction within the joint is used as the output node. The fourth step is to optimize the formulation parameters of the temporary plugging agent for fracturing based on the clustering results and reservoir division obtained by the clustering algorithm, and determine the dosage ratio of the temporary plugging agent.

2. The intelligent optimization method for the formulation parameters of the temporary plugging agent for fracturing according to claim 1, characterized in that... Temporary plugging agents for joints include granular plugging agents with a particle size of <1mm, granular plugging agents with a particle size of 1mm to 3mm, granular plugging agents with a particle size of 1mm to 5mm, granular plugging agents with a particle size of 1mm to 7mm, and powdered plugging agents.

3. The intelligent optimization method for the formulation parameters of the temporary plugging agent for fracturing according to claim 1 or 2, characterized in that... Missing value imputation methods include: first, imputing missing values ​​using non-machine learning methods that are as realistic as possible, including imputation using neighboring well information; then, imputing missing values ​​using machine learning methods when there is no neighboring well information; and finally, using machine learning feature engineering methods for outlier identification preprocessing.

4. The intelligent optimization method for the formulation parameters of the temporary plugging agent for fracturing according to claim 3, characterized in that... When imputing missing values, if the missing rate is no greater than 5%, the median is used for imputation; if the missing rate is greater than 5%, random interpolation is used for imputation.

5. The intelligent optimization method for the formulation parameters of the temporary plugging agent for fracturing according to claim 1, 2, or 4, characterized in that... The fourth step, the optimization of the formulation parameters of the temporary plugging agent for fracturing, includes the following specific processes: First, the geological data and temporary plugging construction data of the target block are combined and input into the BP neural network regression model to predict the plugging strength and pressure, and the first dimension parameter of the in-fracture temporary plugging agent for fracturing, the total amount of in-fracture temporary plugging agent, is determined and optimized. Then, under the condition of determining the total amount of the fracture plugging agent, which is the first-dimensional parameter of the fracture plugging agent, and combined with the clustering algorithm based on PCA principal component analysis, if the horizontal wells in the target block are of type A, the proportion of particulate plugging agent is increased; if the horizontal wells in the target block are of type B, the proportion of powder plugging agent is increased. The data is then input into the BP neural network model to predict the sealing pressure intensity, thereby determining and optimizing the ratio of the two specifications of the fracture plugging agent, which is the second-dimensional parameter of the fracture plugging agent.

Images