Fracturing slickwater evaluation method

Abstract

The present invention discloses a method for evaluating fracturing slickwater, including: S1: obtaining a fracturing slickwater sample; S2: extracting the types and addition ratios of drag reducers, flowback aids, and stabilizers in the sample; S3: performing simulation calculations on the sample to obtain the viscosity, friction resistance, filtrate loss property, and sand-carrying capacity of the sample; S4: constructing a training set using the data obtained in S2 and S3, and training a decision tree model to obtain an evaluation model; S5: using the evaluation model to predict the viscosity, friction resistance, filtrate loss property, and sand-carrying capacity of the to-be-evaluated fracturing slickwater. The present invention can significantly shorten the evaluation cycle and reduce the evaluation cost.

Description

Technical Field

[0001] The present invention relates to fracturing slickwater. More specifically, the present invention relates to a method for evaluating fracturing slickwater. Background Art

[0002] In the field of coalbed methane extraction, the hydraulic fracturing technology is one of the core means to achieve efficient development of coalbed methane. As a working fluid widely used in hydraulic fracturing operations, the performance of fracturing slickwater plays a decisive role in the fracturing effect and gas production efficiency of coalbed methane. At present, the evaluation of fracturing slickwater for coalbed methane mainly adopts experimental testing methods. For example, the basic physical and chemical parameters of slickwater, such as viscosity and surface tension, are measured through laboratory experiments, and its drag reduction and sand-carrying performance are evaluated by means of simulation experiments. Obviously, the experimental method has significant defects. One is that the experimental process is cumbersome and time-consuming, requiring a large amount of manpower, material resources and funds. The other is that there is a large difference between the experimental environment and the actual working conditions of fracturing operations, resulting in a significant deviation between the evaluation results and the actual situation, and it is difficult to effectively guide on-site operations.

[0003] Therefore, it is necessary to design a technical solution that can overcome the above defects. Summary of the Invention

[0004] An object of the present invention is to provide a method for evaluating fracturing slickwater, which can greatly shorten the evaluation cycle and reduce the evaluation cost.

[0005] To achieve these objects and other advantages of the present invention, according to one aspect of the present invention, there is provided a method for evaluating fracturing slickwater, including: S1: obtaining a sample of fracturing slickwater; S2: extracting the types and addition ratios of drag reducers, flowback aids, and stabilizers in the sample; S3: performing simulation calculations on the sample to obtain the viscosity, friction, filtrate loss property, and sand-carrying capacity of the sample; S4: constructing a training set using the data obtained in S2 and S3, and training a decision tree model to obtain an evaluation model; S5: using the evaluation model to predict the viscosity, friction, filtrate loss property, and sand-carrying capacity of the fracturing slickwater to be evaluated; wherein, in the step of training the decision tree model: the maximum depth of the decision tree is determined in the following manner: setting the upper limit value of the maximum depth to 12-16; when performing node splitting operations each time, calculating the difference in information entropy before and after splitting, and the difference in information entropy is obtained by the formula "difference in information entropy = information entropy of the node before splitting - sum of the information entropies of the child nodes after splitting"; setting the initial threshold value of the difference in information entropy to 0.2-0.3, and during the construction of the decision tree, when the difference in information entropy after splitting a certain node is less than the current threshold value of the difference in information entropy, stop splitting the node; and every time 4-6 node splitting operations are completed, adjust the threshold value of the difference in information entropy according to the change in the number of nodes: if the number of nodes increases by greater than or equal to 20% in two consecutive adjustment cycles, then increase the threshold value of the difference in information entropy to 0.25; if the number of nodes increases by less than 20% in two consecutive adjustment cycles, then decrease the threshold value of the difference in information entropy to 0.15, so as to dynamically determine the maximum depth of the decision tree.

[0006] Further, in the step of training the decision tree model, it further includes: using the information gain ratio as a measurement criterion, and during the construction of the decision tree, each time select the feature with the largest information gain ratio for node splitting until the following stopping conditions are met: when the information gain ratios of all remaining selectable features are less than 0.01-0.02, stop splitting; at the same time, set the minimum number of samples for splitting, and when the number of samples on a node is less than this value, no longer split.

[0007] Further, in the step of training the decision tree model, it also includes: dividing the collected sample data according to the ratio of 70% training set, 15% validation set, and 15% test set; using the training set data to train the decision tree. During the training process, for each branch formed after the node split, continuously calculate the information gain ratio of each feature under this branch, and select the feature with the largest information gain ratio to continue splitting the node; after each training, evaluate performance indicators such as the mean squared error and accuracy of the model on the validation set, and adjust the hyperparameters according to the evaluation results, and retrain the model until the performance on the validation set no longer improves or reaches the preset number of training rounds; after the training is completed, use the test set data to evaluate the mean squared error between the predicted values and the true values of the viscosity, frictional resistance, filtrate loss property, and sand-carrying capacity of the model, and calculate the coefficient of determination to measure the goodness of fit of the model; if the performance of the model on the test set does not meet the expectation, start pruning the sub-branches from the leaf nodes and replace them with leaf nodes, and by comparing the mean squared error and coefficient of determination indicators of the model on the validation set before and after pruning, select the model with a higher comprehensive score S i as the final model.

[0008] Further, the comprehensive score S i is calculated by the following method:

[0009] Set the weight of the mean squared error to be 0.4 - 0.5, set the weight of the coefficient of determination to be 0.5 - 0.6, and satisfy ;

[0010] For each model i, obtain its mean squared error MSE i and coefficient of determination on the validation set respectively;

[0011] Through the formula

[0012] calculate the comprehensive score S of each model i .

[0013] Further, step S2 includes: analyzing the components of the drag reducer and the drainage aid in the sample by HPLC. According to the known drag reducer and drainage aid standards, by comparing the retention time and peak area, determine the types of the drag reducer and the drainage aid, and calculate their contents; separate the stabilizer in the sample by GC-MS, compare the mass spectrometry with the standard spectral library to determine the chemical structure and type of the stabilizer, and then quantify its content by the external standard method; among them, when performing HPLC analysis, the following steps are also included: first prepare an initial mixed solvent according to the volume ratio of water to methanol of 3:2, then add N,N-dimethylformamide accounting for 1% of the total volume of the initial mixed solvent, and dimethyl sulfoxide accounting for 1% of the total volume of the initial mixed solvent to obtain the final mixed solvent; add the sample to the final mixed solvent, and the addition ratio is 1 g of the sample: 30-50 ml of the final mixed solvent, accelerate dissolution by stirring, and finally use the dissolved solution for HPLC analysis.

[0014] Further, the simulation calculation of viscosity includes: according to the type and addition ratio obtained in step S2, establish a virtual model of fracturing slickwater using a non-Newtonian fluid viscosity calculation model in the fluid dynamics simulation software ANSYS Fluent, set the temperature parameter to 70-80 °C and the pressure to 30-35 MPa, and obtain the virtual viscosity value A of the fracturing slickwater sample through iterative calculation; measure the measured viscosity value of the sample using a rotational viscometer, repeat the measurement three times, and take the average measured viscosity value B; compare the virtual viscosity value with the average measured viscosity value. If the difference: (A - B) / B > ±5%, then change to another non-Newtonian fluid viscosity calculation model, and again obtain the virtual viscosity value of the fracturing slickwater sample through iterative calculation. Compare the newly obtained virtual viscosity value with the average measured viscosity value. If the difference: (A - B) / B is not greater than ±5%, then select the non-Newtonian fluid viscosity calculation model of the changed type as the annual calculation model. If the differences of all types of non-Newtonian fluid viscosity calculation models are > ±5%, then select the one with the smallest absolute value of the difference; the non-Newtonian fluid viscosity calculation models include the power-law model, the Herschel-Bulkley model, the Bingham plastic model, and the Carreau model; and the iterative calculation is set with an upper limit of 1000 times.

[0015] Further, in S3, the method for simulating and calculating the friction of the sample includes: according to the actual fracturing operation, use CFD software to establish a pipeline geometric model, divide the grid, set the fluid properties and the flow channel model, define the boundary conditions, solve the Navier-Stokes equation to simulate the flow of fracturing slickwater in the pipeline, obtain the pressure values at the inlet and outlet of the pipeline, and calculate the friction using the pressure difference method.

[0016] Furthermore, in the S3, the method for simulating and calculating the filtration loss of the sample includes: according to the actual fracturing operation, using CFD software to construct a porous medium model, determining the interaction parameters between fracturing slick water and rock, setting initial conditions and boundary conditions, combining Darcy's law, solving the continuity equation, simulating the seepage process of fracturing slick water in the porous medium, setting a monitoring surface in the porous medium model, calculating the filtration rate of the fracturing slick water flow through the surface, and then calculating the filtration rate and filtration amount.

[0017] Furthermore, in S3, the method for simulating and calculating the sand carrying capacity of the sample includes: using Fluent software to construct a geometric model, setting parameters of fracturing slick water and proppant, dividing the grid, setting boundary conditions, and using the Eulerian-Eulerian model to simulate the two-phase flow of fracturing slick water and proppant. After the simulation, obtaining the proppant velocity data at the preset monitoring point and calculating the sedimentation velocity.

[0018] Furthermore, it also includes: using real data to verify the viscosity, friction, filtration loss and sand carrying capacity obtained by simulation calculation, if the verification fails, re-simulation calculation; wherein, the Pearson correlation coefficient and consistency limit of the simulation calculation data and the real data are calculated, if the Pearson correlation coefficient and the consistency limit are both within the preset range, it is judged that the verification is passed.

[0019] The present invention has at least the following beneficial effects:

[0020] The present invention uses simulation calculation to obtain key performance indicators such as fracturing slick water viscosity, friction, filtration loss and sand carrying capacity. Combined with the types and addition ratios of drag reducers, drainage aids and stabilizers in fracturing slick water, an evaluation model is constructed with the help of a decision tree model. The performance of the slick water to be evaluated can be quickly predicted, the evaluation cycle can be greatly shortened, the evaluation cost can be reduced, and the deep relationship between the proportion of additives and performance can be explored, which is helpful to optimize the additive formula.

[0021] Other advantages, objectives and features of the present invention will be embodied in part through the following description, and in part will be understood by those skilled in the art through study and practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a flow chart of an embodiment of the present application. DETAILED DESCRIPTION

[0023] The present invention is further described in detail below in conjunction with the accompanying drawings so that those skilled in the art can implement the invention with reference to the description.

[0024] It should be understood that terms such as "having", "including", and "comprising" used in the embodiments of the present application do not exclude the presence or addition of one or more other elements or their combinations. All directional indications (such as up, down, left, right, front, back...) in the embodiments of the present application are only used to explain the relative positional relationship and movement conditions between components in a specific posture. If the specific posture changes, the directional indications will also change accordingly. When an element is referred to as being "fixed to" or "disposed on" another element, it can be directly on the other element or there may be an intermediate element at the same time. When an element is referred to as being "connected" to another element, it can be directly connected to the other element or indirectly connected to the other element through an intermediate element. The descriptions involving "first", "second", etc. in the embodiments of the present application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one such feature.

[0025] It should be noted that the technical solutions between the various embodiments of the present application can be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions results in contradictions or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection required by the present application.

[0026] As Figure 1 shown, the embodiments of the present application provide a method for evaluating fracturing slickwater, including:

[0027] S1: Obtain a fracturing slickwater sample;

[0028] Specifically, collect or purchase 50 groups of different fracturing slickwater samples, record information such as the source and formulation composition of each sample to ensure that the samples have broad representativeness and can reflect various situations in actual applications.

[0029] S2: Extract the types and addition ratios of drag reducers, flowback aids, and stabilizers in the sample;

[0030] If the formulation is not clear, use HPLC to analyze the components of the drag reducer and flowback aid in the sample. According to the known drag reducer and flowback aid standards, by comparing the retention time and peak area, determine the types of the drag reducer and flowback aid and calculate their contents; use GC-MS to separate the stabilizer in the sample, compare the mass spectrometry diagram with the standard spectral library to determine the chemical structure and type of the stabilizer, and then quantify its content by the external standard method;

[0031] Among them, when performing HPLC analysis, the following steps are also included:

[0032] First, an initial mixed solvent is prepared according to the volume ratio of water to methanol of 3:2. Then, N,N-dimethylformamide accounting for 1% of the total volume of the initial mixed solvent and dimethyl sulfoxide accounting for 1% of the total volume of the initial mixed solvent are added to obtain the final mixed solvent;

[0033] The sample is added to the final mixed solvent, and the addition ratio is 1 g of the sample: 30 - 50 ml of the final mixed solvent. Dissolution is accelerated by stirring, and finally the dissolved solution is used for HPLC analysis;

[0034] Generally, drag reducers include polyacrylamide drag reducers, guar gum drag reducers, and surfactant drag reducers. Flowback aids include fluorocarbon surfactant flowback aids, polyether-type flowback aids, and alkylphenol polyoxyethylene ether flowback aids. Stabilizers include organic phosphonate stabilizers, polymer stabilizers, and borate stabilizers.

[0035] S3: Perform simulation calculations on the sample to obtain the viscosity, frictional resistance, fluid loss property, and sand-carrying capacity of the sample;

[0036] S31: One way to calculate viscosity is;

[0037] According to the type and addition ratio obtained in step S2, a virtual model of fracturing slickwater is established using a non-Newtonian fluid viscosity calculation model in the fluid dynamics simulation software ANSYS Fluent. Set the temperature parameter to 70 - 80 °C and the pressure to 30 - 35 MPa, and obtain the virtual viscosity value A of the fracturing slickwater sample through iterative calculation;

[0038] Measure the measured viscosity value of the sample using a rotational viscometer, repeat the measurement three times, and take the average measured viscosity value B;

[0039] Compare the virtual viscosity value with the average measured viscosity value. If the difference: (A - B) / B > ±5%, then change to another non-Newtonian fluid viscosity calculation model, and again obtain the virtual viscosity value of the fracturing slickwater sample through iterative calculation. Compare the newly obtained virtual viscosity value with the average measured viscosity value. If the difference: (A - B) / B is not greater than ±5%, then select the non-Newtonian fluid viscosity calculation model of the type changed as the annual calculation model. If the differences for all types of non-Newtonian fluid viscosity calculation models are > ±5%, then select the one with the smallest absolute value of the difference;

[0040] The non-Newtonian fluid viscosity calculation models include the power-law model, Herschel - Bulkley model, Bingham plastic model, and Carreau model;

[0041] And the iterative calculation is set with an upper limit of 1000 times.

[0042] Another way to calculate viscosity is:

[0043] With the help of Materials Studio molecular modeling software, according to the chemical composition and molecular composition of the samples, construct the molecular structures and simulation boxes of each component for each group of samples; obtain the temperature T (unit: K) and pressure P (unit: Pa) during the actual fracturing operation, and set the simulation conditions accordingly; use the COMPASS force field to describe the intermolecular interactions, which can accurately reflect the physical and chemical properties of molecules;

[0044] Use the molecular dynamics method for simulation. During the simulation process, record the position information of molecules at certain time intervals, and obtain the displacements of molecules by analyzing this position information. Conduct statistical analysis on the displacement data of a large number of molecules, and calculate the diffusion coefficient D using the Einstein diffusion formula;

[0045] ;

[0046] where N is the number of molecules, is the position vector of the i-th molecule at time t, and the molecular radius R is determined by referring to relevant literature; finally, calculate the viscosity η according to the Einstein relation where k B is the Boltzmann constant.

[0047] S32: Friction calculation;

[0048] Using ANSYS Fluent software, establish a pipeline geometric model based on parameters such as the diameter d, length L, and material of the pipeline during the actual fracturing operation. Perform mesh division on the model, and appropriately refine the mesh at key parts such as the pipeline wall, elbows, and diameter changes to improve the calculation accuracy.

[0049] Set the fluid properties. The dynamic viscosity μ is determined according to the viscosity value obtained from the previous calculation, and the density is obtained through laboratory measurement. Judge whether the fluid is in a turbulent or laminar state according to the actual situation, and select a suitable flow channel model. If the Reynolds number (v is the average fluid velocity), then it is judged as turbulent, and a suitable turbulent model is selected; if it is less than 2000, then it is laminar, and a laminar model is selected.

[0050] Define the boundary conditions. Set the inlet as the velocity boundary condition, and the velocity v is determined according to the actual injection velocity of the fracturing pump; set the outlet as the pressure boundary condition, and the pressure value P out refers to the formation pressure of the coalbed methane field.

[0051] Solve the following Navier - Stokes equations;

[0052] ;

[0053] where ρ is the fluid density, is the local rate of change of fluid velocity, is the convective acceleration, is the fluid velocity vector (unit: m / s), P is the pressure (unit: Pa), is the body force (unit: N / m³), and the body force mainly considers gravity, is the pressure gradient force, is the viscous force, and μ is the dynamic viscosity.

[0054] The inlet pressure P1 and outlet pressure P2 of the pipeline are obtained through simulation calculations, and the frictional resistance △P = P1 - P2.

[0055] S33: Filtration loss calculation;

[0056] Based on the geological exploration data of the coalbed methane field, such as information on the porosity Φ, permeability K, and rock type of the coal seam, a porous media model is constructed using CFD software. Through the interfacial tension measurement experiment and contact angle measurement experiment, the interaction parameters between the fracturing slickwater and the coal seam rock are determined, such as the interfacial tension σ and contact angle θ.

[0057] Set the initial conditions, the initial pressure P0 and initial flow velocity v0 of the fracturing slickwater, referring to the initial settings of the actual fracturing operation. Set the boundary conditions, and the pressure Pb and flow rate Qb at the model boundary are set according to the actual situation of the coalbed methane field.

[0058] Combined with the following Darcy's law;

[0059] , where Q is the flow rate, K is the permeability, A is the cross-sectional area, is the pressure gradient, and μ is the fluid dynamic viscosity;

[0060] Discretize the following continuity equation by the finite difference method or finite element method;

[0061] , where, is the rate of change of fluid density with time t, is the divergence of the mass flux, is the velocity vector of the fluid, and ▽ is the Hamiltonian operator;

[0062] Set a monitoring surface in the porous media model, and the filtration rate is obtained by monitoring the flow rate Q in real time through the flow sensor at the monitoring surface. The filtration volume V is calculated by numerical integration Calculated using the trapezoidal integration method or Simpson's integration method for numerical calculation, where t is the simulation time.

[0063] S34: Sand-carrying capacity calculation;

[0064] Use Fluent software to construct a geometric model including the wellbore and fractures, considering the wellbore diameter D well , length L well and actual factors such as the width ω, height h, and extension direction of the fractures. Set the detailed parameters of the fracturing slickwater and proppant. The parameters of the fracturing slickwater are determined according to the previous calculation and measurement results, and the particle size d of the proppant p is determined by sieve analysis, and the density ρ p is measured by a densitometer, and the shape factor Ψ is determined by microscopic observation and image analysis.

[0065] Perform mesh division on the model, and appropriately refine the mesh in areas where the proppant movement and fluid flow are intense. Set the boundary conditions. Set the injection velocities v in and concentrations C in of the fracturing slickwater and proppant at the inlet. The injection velocity is determined according to the actual injection capacity of the fracturing pump, and the concentration is determined according to the design requirements in actual operations; set the pressure condition at the outlet, and the pressure value P out refers to the formation pressure of the coalbed methane field.

[0066] Adopt the Eulerian-Eulerian model to simulate the two-phase flow of the fracturing slickwater and proppant. This model considers the interactions such as drag force and lift force between the two phases. During the simulation process, solve the continuity equations and momentum equations of the two phases. After the simulation is completed, obtain the velocity data of the proppant at the preset monitoring points, and mainly focus on the velocity v z in the vertical direction, and the settling velocity v s is v z .

[0067] S35: Verification;

[0068] Compare the viscosity, friction, filtration loss, and sand-carrying capacity data of each group of samples obtained from the simulation calculation with the real data measured in the laboratory. Calculate the Pearson correlation coefficient r, and the formula is: ,

[0069] where x i is the simulation calculation data, y i is the real data, and are the average values of the simulation calculation data and the real data respectively, and n is the number of samples.

[0070] At the same time, calculate the mean value of the difference between the simulation value and the real value and the standard deviation , and determine the consistency limit as If both the Pearson correlation coefficient and the consistency limit are within the preset range (e.g., the Pearson correlation coefficient is greater than 0.8 and the consistency limit is within ±10%), it is determined that the simulation calculation verification of the sample passes; if the verification fails, the simulation calculation of the sample is performed again, and the simulation parameters or model settings are adjusted until the verification passes.

[0071] In this step, by using advanced molecular modeling software and CFD software, and comprehensively considering actual factors such as temperature, pressure, and geological structure of the coalbed methane field, the real state of fracturing slickwater in the well can be accurately simulated, and accurate data on viscosity, friction, filtration loss, and sand-carrying capacity can be obtained; compared with traditional laboratory experimental methods, simulation calculations can complete the performance evaluation of a large number of samples in a relatively short time, greatly shortening the evaluation cycle, and as long as the simulation parameters and conditions remain the same, the calculations can be repeated, ensuring the stability and reliability of the results.

[0072] By calculating the Pearson correlation coefficient and the consistency limit, and quantitatively comparing the simulation data with the real data, the degree of fit between the simulation results and the actual situation can be accurately judged. If the verification passes, it indicates that the models, parameter settings, etc. adopted in the simulation process are reasonable and effective, and the data such as viscosity, friction, filtration loss, and sand-carrying capacity obtained can truly reflect the performance of fracturing slickwater in actual coalbed methane exploitation. Accurate and reliable simulation data can enable engineers to more accurately evaluate the performance of different formula fracturing slickwaters, so as to train a more accurate evaluation model and reduce the exploitation cost; if the verification fails, the problems existing in the simulation calculation can be timely discovered, and by re-performing the simulation calculation and adjusting the parameters or model settings, the simulation process can be gradually optimized to make the simulation results closer to the actual situation and improve the performance and practicality of the entire evaluation system.

[0073] S4: Construct a training set using the data obtained in S2 and S3, and train a decision tree model to obtain an evaluation model.

[0074] Encode the additive types in S2, such as using different numbers to represent different types of drag reducers, flowback aids, and stabilizers. Calculate the sum feature M+N+P of the addition ratio, the product feature M×N×P of the addition ratio, the sum logarithmic feature ln(M+N+P) of the addition ratio, and the product logarithmic feature ln(M×N×P) of the addition ratio.

[0075] Integrate these data features with the viscosity, friction, filtration loss, and sand-carrying capacity data of each group of samples obtained from the S3 simulation calculation to form a complete data set.

[0076] In the steps of training a decision tree model: The maximum depth of the decision tree is determined as follows: Set the upper limit of the maximum depth to 12 - 16; When performing node splitting operations each time, calculate the difference in information entropy before and after splitting, and the difference in information entropy is obtained through the formula "Difference in information entropy = Information entropy of the node before splitting - Sum of information entropies of the child nodes after splitting"; Set the initial difference threshold of information entropy to 0.2 - 0.3. During the construction of the decision tree, when the difference in information entropy after splitting a certain node is less than the current difference threshold of information entropy, stop splitting the node; And every time 4 - 6 node splitting operations are completed, adjust the difference threshold of information entropy according to the change in the number of nodes: If the number of nodes increases by greater than or equal to 20% in two consecutive adjustment cycles, then increase the difference threshold of information entropy to 0.25; If the number of nodes increases by less than 20% in two consecutive adjustment cycles, then reduce the difference threshold of information entropy to 0.15, thereby dynamically determining the maximum depth of the decision tree; Through the above method of dynamically determining the maximum depth of the decision tree, while avoiding overfitting, the learning ability of the decision tree model can be fully utilized, enabling it to better adapt to different data sets and improving the generalization performance and prediction accuracy of the model.

[0077] The calculation formula for information entropy is , where H(X) is the information entropy of the random variable X (data features and viscosity, frictional resistance, filtration loss, and sand - carrying capacity data of each group of samples obtained from simulation calculations), n is the number of all possible values of the random variable X, and p(x i ) is the probability that the random variable X takes the value x i .

[0078] In the steps of training a decision tree model, it also includes: Using the information gain ratio as the measurement standard, during the construction of the decision tree, each time select the feature with the largest information gain ratio for node splitting until the following stopping conditions are met: When the information gain ratios of all remaining selectable features are less than 0.01 - 0.02, stop splitting; At the same time, set the minimum number of samples for splitting, and when the number of samples on a node is less than this value, no longer split; The calculation formula is as follows:

[0079]

[0080]

[0081]

[0082]

[0083] Among them, D represents the data set, Di is the sample subset corresponding to the i - th value of the feature A, and |D| and |Di| are the number of samples;

[0084] is the information gain, is the information gain ratio, is the entropy of feature A itself;

[0085] In the step of training the decision tree model, it also includes: dividing the collected sample data according to the ratio of 70% training set, 15% validation set, and 15% test set; using the training set data to train the decision tree. During the training process, for each branch formed after the node splitting, continuously calculate the information gain ratio of each feature under this branch, and select the feature with the largest information gain ratio to continue splitting the node; after each training, evaluate the performance indicators such as the mean square error and accuracy of the model on the validation set, and adjust the hyperparameters according to the evaluation results, and retrain the model until the performance on the validation set no longer improves or reaches the preset number of training rounds; after the training is completed, use the test set data to evaluate the mean square error between the predicted values and the true values of the viscosity, frictional resistance, filtration loss, and sand-carrying capacity of the model, and calculate the coefficient of determination to measure the goodness of fit of the model; if the performance of the model on the test set does not meet the expectations, start pruning the sub-branches from the leaf nodes and replace them with leaf nodes. By comparing the mean square error and coefficient of determination indicators of the model on the validation set before and after pruning, select the model with a higher comprehensive score S i as the final model;

[0086] The comprehensive score S i is calculated by the following method:

[0087] Set the weight of the mean square error to be 0.4 - 0.5, set the weight of the coefficient of determination to be 0.5 - 0.6, and satisfy ;

[0088] For each model i, obtain its mean square error MSE i and the coefficient of determination on the validation set respectively;

[0089] Through the formula

[0090] calculate the comprehensive score S of each model i ;

[0091] S5: Use the evaluation model to predict the viscosity, frictional resistance, filtration loss, and sand-carrying capacity of the fracturing slickwater to be evaluated.

[0092] Obtain a new sample of the fracturing slickwater to be evaluated, extract the additive type and addition ratio according to the method of S2, perform data feature extraction, input the extracted data features into the trained evaluation model, and the model predicts the viscosity, frictional resistance, filtration loss, and sand-carrying capacity of the fracturing slickwater to be evaluated according to the learned rules. The prediction results provide a scientific basis for selecting a suitable fracturing slickwater for the actual fracturing operation in the coalbed methane field, which helps to optimize the fracturing plan and improve the coalbed methane extraction efficiency;

[0093] In this step, by integrating data features such as additive type and addition ratio with the performance indexes of slickwater obtained from simulation calculations, the decision tree model can learn the complex non-linear relationship between the two. After training with a large number of sample data and evaluation and optimization of the test set, the model can accurately predict the key performance indexes such as the viscosity, friction, filtrate loss, and sand-carrying capacity of the fracturing slickwater to be evaluated, providing a scientific basis for the design and optimization of the fracturing plan. Traditional evaluation methods for fracturing slickwater often require a large number of experiments and tests, consuming a lot of time and resources. However, by using the trained decision tree model, only the additive information of the sample to be evaluated needs to be input, and the performance prediction results can be obtained quickly, greatly improving the evaluation efficiency and reducing the evaluation cost. By extracting features such as the sum, product, and logarithm of the additive ratio, deeper relationships between the additives and the performance of the slickwater are explored, enriching the input information of the model and enabling the model to more comprehensively and accurately grasp the influence of various factors on the performance of the slickwater.

[0094] The following uses the above embodiments to predict and experimentally determine the viscosity, friction, filtrate loss, and sand-carrying capacity of Samples 1, 2, and 3. The results are shown in the following table.

[0095]

[0096] Although the embodiments of the present invention have been disclosed as above, they are not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the field, additional modifications can be easily made. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the examples shown and described herein.

Claims

1. A method for evaluating fracturing slick water, characterized in that: include: S1: Obtain fracturing slickwater samples; S2: extracting the types and addition ratios of drag reducers, drainage aids and stabilizers in the sample; S3: performing simulation calculation on the sample to obtain the viscosity, friction, filtration loss and sand carrying capacity of the sample; S4: Use the data obtained in S2 and S3 to construct a training set, and train the decision tree model to obtain an evaluation model; S5: using the evaluation model to predict the viscosity, friction, filtration loss and sand carrying capacity of the fracturing slick water to be evaluated; Among them, in the step of training the decision tree model: The maximum depth of the decision tree is determined by the following method: setting the upper limit of the maximum depth to 12-16; calculating the information entropy difference before and after the split each time a node split operation is performed, and the information entropy difference is obtained by the formula "information entropy difference = information entropy of the node before split - sum of information entropy of the child nodes after split"; The initial information entropy difference threshold is set to 0.2-0.

3. During the decision tree construction process, when the information entropy difference after a node split is less than the current information entropy difference threshold, stop splitting the node; and after every 4-6 node splits, adjust the information entropy difference threshold according to the change in the number of nodes: if the number of nodes increases by more than or equal to 20% in two consecutive adjustment cycles, increase the information entropy difference threshold to 0.25; if the number of nodes increases by less than 20% in two consecutive adjustment cycles, reduce the information entropy difference threshold to 0.15, so as to dynamically determine the maximum depth of the decision tree; The steps of training a decision tree model also include: The collected sample data is divided into 70% training set, 15% validation set, and 15% test set; Use the training set data to train the decision tree. During the training process, for each branch formed after the node splits, continuously calculate the information gain ratio of each feature under the branch, and select the feature with the largest information gain ratio to continue splitting the node; After each training, the model's performance indicators such as mean square error and accuracy are evaluated on the validation set, and the hyperparameters are adjusted based on the evaluation results. The model is retrained until the performance on the validation set no longer improves or the preset number of training rounds is reached; After training, the test set data was used to evaluate the mean square error between the model's predicted values ​​and the true values ​​for viscosity, friction, filtration loss, and sand carrying capacity, and the coefficient of determination was calculated to measure the goodness of fit of the model; If the model performance on the test set does not meet expectations, start pruning the sub-branch from the leaf node and replace it with the leaf node. By comparing the mean square error and determination coefficient of the model on the validation set before and after pruning, select the comprehensive score S i The higher model is taken as the final model.

2. The fracturing slick water evaluation method according to claim 1, characterized in that: The steps of training a decision tree model also include: Using information gain ratio as the metric, during the decision tree construction process, the feature with the largest information gain ratio is selected each time for node splitting until the following stop condition is met: when the information gain ratios of all remaining selectable features are less than 0.01-0.02, the splitting stops; At the same time, set the minimum number of sample splits. When the number of samples on a node is less than this value, it will no longer split.

3. The fracturing slick water evaluation method according to claim 1, characterized in that: Comprehensive score i It is calculated as follows: Setting the weight of mean square error The weight of the determination coefficient is set to 0.4-0.5 is 0.5-0.6 and satisfies ; For each model i, obtain its mean square error MSE on the validation set i and the coefficient of determination ; By formula Calculate the comprehensive score S for each model i .

4. The fracturing slick water evaluation method according to claim 1, characterized in that: Step S2 includes: HPLC was used to analyze the components of drag reducers and drainage aids in the samples. Based on known standards of drag reducers and drainage aids, the types of drag reducers and drainage aids were determined by comparing retention time and peak area, and their contents were calculated. GC-MS was used to separate the stabilizer in the sample, and the chemical structure and type of the stabilizer were determined by comparing the mass spectrum with the standard library, and then the content was quantified by the external standard method. Wherein, when carrying out HPLC analysis, also comprise the following steps: First, an initial mixed solvent is prepared according to a volume ratio of water to methanol of 3:2, and then 1% of N,N-dimethylformamide and 1% of dimethyl sulfoxide are added to the total volume of the initial mixed solvent to obtain a final mixed solvent; Add the sample to the final mixed solvent in a ratio of 1 g sample to 30-50 ml final mixed solvent, accelerate dissolution by stirring, and finally use the dissolved solution for HPLC analysis.

5. The method for evaluating fracturing slick water according to claim 1, wherein: The viscosity simulation calculation includes: According to the type and addition ratio obtained in step S2, a non-Newtonian fluid viscosity calculation model in the fluid mechanics simulation software ANSYS Fluent is used to establish a virtual model of fracturing slick water, and the temperature parameters are set to 70-80°C and the pressure is set to 30-35MPa. The virtual viscosity value A of the fracturing slick water sample is obtained through iterative calculation; Use a rotational viscometer to measure the viscosity of the sample, repeat the measurement three times, and take the average value B of the measured viscosity; Compare the virtual viscosity value with the average measured viscosity. If the difference (AB) / B>±5%, change the non-Newtonian fluid viscosity calculation model, and iterate again to get the virtual viscosity value of the fracturing slick water sample. If the difference (AB) / B between the newly obtained virtual viscosity value and the average measured viscosity is not greater than ±5%, select the changed non-Newtonian fluid viscosity calculation model as the annual calculation model. If the difference of all non-Newtonian fluid viscosity calculation models is greater than ±5%, select the one with the smallest absolute difference. The non-Newtonian fluid viscosity calculation model includes a power law model, a Herschel-Bulkley model, a Bingham plasticity model, and a Carreau model; The upper limit of the number of iterations is set to 1000.

6. The fracturing slick water evaluation method according to claim 1, characterized in that: In S3, the method of simulating and calculating the friction resistance of the sample includes: According to the actual fracturing operation, CFD software is used to establish the pipeline geometry model, divide the grid, set the fluid properties and flow channel model, define the boundary conditions, solve the Navier-Stokes equation to simulate the flow of fracturing slick water in the pipeline, obtain the pressure values ​​at the pipeline inlet and outlet, and calculate the friction resistance using the pressure difference method.

7. The fracturing slick water evaluation method according to claim 1, characterized in that: In S3, the method of simulating and calculating the sample filtration loss includes: According to the actual fracturing operation, the porous media model is constructed using CFD software, the interaction parameters between fracturing slick water and rock are determined, the initial conditions and boundary conditions are set, and the continuity equation is solved in combination with Darcy's law to simulate the seepage process of fracturing slick water in porous media. A monitoring surface is set in the porous media model, and the flow rate of fracturing slick water passing through the surface is calculated to obtain the filtration rate, and then the filtration rate and filtration amount are calculated.

8. The method for evaluating fracturing slick water according to claim 1, wherein: In S3, the method of simulating and calculating the sand carrying capacity of the sample includes: Fluent software was used to construct the geometric model, set the parameters of fracturing slick water and proppant, divide the grid, set the boundary conditions, and use the Eulerian-Eulerian model to simulate the two-phase flow of fracturing slick water and proppant. After the simulation, the proppant velocity data at the preset monitoring point was obtained and the sedimentation velocity was calculated.

9. The method for evaluating fracturing slick water according to claim 1, wherein: Also includes: Use real data to verify the viscosity, friction, filtration and sand carrying capacity obtained by simulation. If the verification fails, re-simulate the calculation; Among them, the Pearson correlation coefficient and consistency limit of the simulated data and the real data are calculated. If the Pearson correlation coefficient and the consistency limit are both within the preset range, the verification is judged to be passed.

Images