A method for determining the main controlling factors of demulsification of naphthenic crude oil with high acid value

Abstract

The present invention provides a method for determining the main control factors for demulsification of naphthenic crude oil with high acid value. The method includes: obtaining demulsification data of demulsifying naphthenic crude oil with high acid value and different physical properties in different formations using a target demulsifier, and normalizing the demulsification data to the same interval; based on the normalized demulsification data, obtaining a multiple regression model for demulsification effect, a ridge regression model for demulsification effect, a lasso regression model for demulsification effect, and a stepwise regression model for demulsification effect, and respectively determining the weighted evaluation indexes of each model; based on the weighted evaluation indexes, determination coefficient, mean square error, root mean square error, and mean absolute error of each model, selecting the optimal model as the demulsification effect model; conducting an analysis of the contribution rates of various formation property parameters, crude oil property parameters, and engineering parameters to the demulsification effect parameters of the demulsification effect model, and determining the top N parameters with the largest contribution rates to the demulsification effect parameters as the main control factors in the process of demulsifying naphthenic crude oil with high acid value using the target demulsifier.

Description

Technical Field

[0001] The present invention belongs to the technical field of oil and gas development and storage and transportation, and particularly relates to a method for determining the main control factors for demulsification of high-acid-value naphthenic crude oil. Background Art

[0002] Crude oil is unprocessed petroleum, which is a mixture of various liquid hydrocarbons such as alkanes, cycloalkanes, aromatic hydrocarbons, and olefins. The acid value of crude oil (Total Acid Number, TAN) is an index quantitatively representing the content of acidic oxygen-containing compounds in crude oil. That is, the total amount of free acids contained in organic substances such as oils and fats, polyesters, and paraffins. Quantitatively speaking, it is the number of milligrams of potassium hydroxide required to neutralize the acidic substances in 1 g of crude oil sample, expressed as mgKOH / g. Crude oil with TAN less than 0.5 mgKOH / g is low-acid-value crude oil, crude oil with TAN greater than 0.5 mgKOH / g is acid-containing crude oil, and crude oil with TAN greater than 1.0 mgKOH / g is high-acid-value crude oil.

[0003] The most significant characteristic after crude oil emulsification is the increase in viscosity, which will cause many problems. For example, it will cause more fuel consumption for the water jacket furnace to adjust the furnace temperature; it will lead to an increase in power consumption of the pumping unit and the external transmission pump; it will cause the risk of broken rods; it requires frequent line flushing; it increases the pipeline friction, reduces the flow rate, causes an increase in the wellhead back pressure, and increases the risk of pipeline leakage; in refining, it will directly cause frequent operation fluctuations or rapid current increase in the electric desalting tank of the vacuum distillation unit of the refinery; during the gathering and transportation process, a large amount of water contained in the emulsified crude oil cannot be removed in time, which will cause problems such as pipeline corrosion.

[0004] The main characteristics of high-acid-value crude oil are high density, high acid value, high heavy metal content, high colloid content, and relatively high viscosity. With the deepening of oilfield development, the heavy components of crude oil will be more and more, and the acid value will also be higher and higher. When high-acid-value crude oil is discharged from the reservoir and mixed with external fluids and formation water contained in itself at high speed, natural emulsifiers in the crude oil such as asphaltenes, colloids, naphthenic acids, organic substances containing nitrogen and sulfur, wax crystals, and clay will promote the formation of a stable crude oil emulsion, that is, the crude oil is emulsified. Especially in recent years, most oilfields have entered the high water cut stage, and the application of various production technologies has made the crude oil mostly produced in the form of emulsion, and the emulsification of crude oil is becoming more and more serious. In the W / O emulsion system of crude oil, colloids, asphaltenes, etc. aggregate on the surface of the dispersed-phase water droplets to form a stable interfacial layer, thus preventing the condensation of water droplets.

[0005] At present, physical and chemical demulsifiers are generally used to solve the problem of crude oil emulsification. For high-acid-value crude oil, there are problems such as low oil-water separation efficiency in the settling tank, large dosage of demulsifier, and overloading of the electric dehydrator during the demulsification process. The main reason is that the density of high-acid-value crude oil is high (0.942 - 1.031 g / cm 3) The proportion of naphthenic crude oil with high viscosity (196.7 - 5400.8 mPa·s) and high gum content increases. Naphthenate is the main factor affecting the demulsification and dehydration of high-acid-value crude oil. In addition, the water content of high-acid-value crude oil increases, and the W / O type of the crude oil emulsion turns into the O / W type, making its chemical composition and emulsion structure more complex. It is difficult to demulsify and dehydrate by electric field method and chemical method, which increases the difficulty of crude oil demulsification and dehydration. It can be seen that demulsification is a very complex problem, which is related to the components and properties of crude oil, the type and stability factors of the emulsion, and also related to the molecular structure and properties of the demulsifier. Therefore, when selecting a demulsifier, the dehydration rate, dehydration speed, oil / water interface state, oil content in the dehydrated water, optimal dosage of the demulsifier, and low-temperature dehydration performance need to be considered comprehensively. At present, in view of the numerous and complex influencing factors, there is no effective method to effectively determine the main control factors for the demulsification of high-acid-value naphthenic crude oil. Summary of the Invention

[0006] The purpose of the present invention is to provide a technical solution that can determine the main control factors in the demulsification process of high-acid-value naphthenic crude oil using a certain demulsifier.

[0007] To achieve the above purpose, the present invention provides a method for determining the main control factors for the demulsification of high-acid-value naphthenic crude oil, which includes:

[0008] Obtain the demulsification data of demulsifying high-acid-value naphthenic crude oil with different physical properties in different formations using the target demulsifier; wherein, the demulsification data includes formation property parameter data, crude oil property parameter data, engineering parameter data, and demulsification effect parameter data;

[0009] Normalize the demulsification data to the same interval to obtain the normalized demulsification data;

[0010] Based on the normalized demulsification data, use multiple regression, ridge regression, lasso regression, and stepwise regression to fit the relationship between the demulsification effect parameters and the formation property parameters, crude oil property parameters, and engineering parameters respectively, to obtain a multiple regression model for demulsification effect, a ridge regression model for demulsification effect, a lasso regression model for demulsification effect, and a stepwise regression model for demulsification effect, and determine the weighted evaluation index of each model respectively; wherein, the weighted evaluation index = 0.5×MSE + 0.2×MAE + 0.2×RMSE + 0.1×(1 - coefficient of determination R 2 ), MSE refers to the mean square error, MAE refers to the mean absolute error, RMSE refers to the root mean square error, and R 2 refers to the coefficient of determination;

[0011] Based on the weighted evaluation indexes of each model, determine the model with the smallest weighted evaluation index; if there is only one model with the smallest weighted evaluation index, use this model with the smallest weighted evaluation index as the demulsification effect model; if there are two or more models with the smallest weighted evaluation index, select the optimal model from these models with the smallest weighted evaluation index as the demulsification effect model according to at least one of the coefficient of determination, mean square error, root mean square error, and mean absolute error;

[0012] Conduct an analysis of the contribution rates of various formation property parameters, crude oil property parameters, and engineering parameters to the demulsification effect parameters for the demulsification effect model, and determine the top N parameters with the largest contribution rates to the demulsification effect parameters as the main control factors in the demulsification process of high-acid-value naphthenic crude oil using the target demulsifier.

[0013] According to the preferred implementation manner of the above method, among them, the formation property parameters include: clay content, whole-rock mineral content; further, the whole-rock mineral content includes quartz content, potassium feldspar content, plagioclase content, calcite content, dolomite content, halite content, analcime content, barite content, anatase content.

[0014] According to the preferred implementation manner of the above method, among them, the crude oil property parameters include: crude oil density, crude oil viscosity, crude oil pour point, crude oil sulfur content, crude oil wax content, initial boiling point of crude oil, final boiling point of crude oil, wax precipitation point of crude oil, content of gum and asphalt in crude oil, acidity of crude oil, flash point of crude oil, initial water content of crude oil, water content of the oil layer after 4 hours of crude oil settlement.

[0015] According to the preferred implementation manner of the above method, among them, the engineering parameters include demulsifier dosage, dosing cycle, and temperature at the time of demulsifier injection.

[0016] According to the preferred implementation manner of the above method, among them, the demulsification effect parameter is selected as the dehydration rate or the water content of the crude oil after demulsification.

[0017] According to the preferred implementation manner of the above method, among them, the method further includes:

[0018] Before normalizing the demulsification data to the same interval, determine whether the demulsification data conforms to a normal distribution. If the demulsification data conforms to a normal distribution, perform the subsequent steps before normalizing the demulsification data to the same interval. If the demulsification data does not conform to a normal distribution, re-obtain the demulsification data;

[0019] Further, the method further includes: before normalizing the demulsification data to the same interval, determine the extreme values, mean value, variance, and distribution characteristics of the demulsification data;

[0020] Among them, determining whether the demulsification data conforms to the normal distribution can be, but is not limited to, through visual analysis; visual analysis can be, but is not limited to, achieved through visual charts, and the visual charts include at least one of Q-Q plots, frequency plots, box plots, and statistical tables; through the visual charts, it can be clearly and intuitively checked whether the selected variables follow the normal distribution;

[0021] The demulsification data that conforms to the normal distribution can effectively avoid the problem that the dependent variable is overly controlled by other variables to be studied during data analysis, thereby ensuring the accuracy and scientific nature of data analysis; the demulsification data conforming to the normal distribution can also ensure the universality of the samples and reduce the errors when selecting data manually.

[0022] According to the preferred implementation manner of the above method, among them, the method further includes:

[0023] Based on the normalized demulsification data, pairwise factor analysis is performed to respectively determine the relationships between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters;

[0024] Furthermore, performing pairwise factor analysis based on the normalized demulsification data includes:

[0025] Based on the normalized demulsification data, pairwise factor scatter plots between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters are respectively drawn, and then the pairwise factor mathematical relationships between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters are respectively fitted, and then Pearson correlation analysis, Spearman correlation analysis, grey relational analysis, or significance analysis is used to determine the relationships between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters.

[0026] According to the preferred implementation manner of the above method, among them, normalizing the demulsification data to the same interval is performed using the Min - Max method or the Z - score method.

[0027] According to the preferred implementation manner of the above method, among them, during the process of normalizing the demulsification data to the same interval, the demulsification data is normalized to [0, 1].

[0028] According to the preferred implementation manner of the above method, among them, during the process of determining the multiple regression model of demulsification effect, the ridge regression model of demulsification effect, the lasso regression model of demulsification effect, and the stepwise regression model of demulsification effect, the normalized demulsification data is divided into a test set and a training set.

[0029] According to the preferred embodiment of the above method, when fitting the relationship between the demulsification effect parameters and the formation property parameters, crude oil property parameters, and engineering parameters by multiple regression, ridge regression, lasso regression, and stepwise regression respectively to obtain the multiple regression model of demulsification effect, ridge regression model of demulsification effect, lasso regression model of demulsification effect, and stepwise regression model of demulsification effect, multiple regression, ridge regression, lasso regression, and stepwise regression can be carried out in a conventional manner;

[0030] In a specific embodiment, the multiple regression model of demulsification effect is determined by the following formula:

[0031] y = Xβ + ε

[0032] Where: y is the response vector, X is the design matrix, β is the parameter vector, and ε is the error vector;

[0033] The least squares estimate of its parameters is:

[0034] β = (X T X) -1 X T y

[0035] Where: y is the response vector, X is the design matrix, and β is the parameter vector;

[0036] To minimize the difference between the model predicted value and the actual observed value, that is, the sum of squared residuals is used to determine the model parameters. The sum of squared residuals is defined as:

[0037] RSS(β) = Σ(y i - x i T β) 2

[0038] Where: y i is the response value corresponding to the i-th matrix element, x i is the i-th matrix element, β is the parameter vector, and RSS(β) is the sum of squared residuals of the parameter vector;

[0039] In a specific embodiment, the ridge regression model of demulsification effect is determined by the following method:

[0040] After giving different design matrices X and response vectors y, determine J(β) = Σ(y i - x i Tβ) 2 + λΣβ j 2 , when J(β) = Σ(y i - x i Tβ) 2 + λΣβ j 2When it is minimized, the demulsification effect ridge regression model can be obtained; where: λ is a non - negative regularization parameter that determines the strength of regularization; the second term λΣ|β j | is the regularization term, which penalizes the absolute value of the coefficients; when λ = 0, ridge regression degenerates to standard linear regression, and when λ increases, the influence of regularization increases and the estimated values of the coefficients shrink;

[0041] In a specific embodiment, the demulsification effect lasso regression model is determined as follows:

[0042] After giving different design matrices X and response vectors y, determine J(β)=Σ(y i -x i Tβ) 2 +λΣ|β j |, when J(β)=Σ(y i -x i Tβ) 2 +λΣ|β j | is minimized, the demulsification effect lasso regression model can be obtained; where: λ is a non - negative regularization parameter that determines the strength of regularization; the second term λΣ|β j | is the regularization term, which penalizes the absolute value of the coefficients;

[0043] In a specific embodiment, the demulsification effect stepwise regression model is determined by using the stepwise regression iteration method, which is divided into forward selection, backward elimination or bidirectional stepwise process; in forward selection, the model starts with no predictor variables and then adds predictor variables one by one, each time selecting the predictor variable that provides the greatest improvement; in backward elimination, the model starts with all predictor variables and then deletes predictor variables one by one, each time deleting the predictor variable that provides the least improvement to the model; bidirectional stepwise regression combines forward selection and backward elimination, considering both adding and deleting predictor variables; the decision at each step is based on the Akaike information criterion (AIC) or the Bayesian information criterion (BIC);

[0044] Among them, according to the Akaike information criterion, the smaller the AIC value of the model, the better it is. AIC = 2k - 2ln(L), where: k is the number of parameters in the model; L is the maximum likelihood of the model;

[0045] Among them, according to the Bayesian information criterion, the smaller the BIC value of the model, the better it is. BIC = kln(n)-2ln(L), where: k is the number of parameters in the model; L is the maximum likelihood of the model; n is the number of data points;

[0046] AIC takes into account the complexity of the model and its goodness of fit to the data, and attempts to strike a balance between selecting a model complex enough to fit the data well and selecting a simple model to avoid overfitting. Compared with AIC, BIC tends to select a simpler model. Generally, if the main goal of model selection is prediction, AIC may be more preferred. If the goal is to select a simple model that can still explain the data well, BIC may be more preferred.

[0047] According to a preferred embodiment of the above method, the contribution rate is determined by the following formula:

[0048]

[0049] In the formula: CR i is the contribution rate of the i-th parameter among the formation property parameters, crude oil property parameters, and engineering parameters in the model; a i is the coefficient of the i-th parameter among the formation property parameters, crude oil property parameters, and engineering parameters in the model; |a j | is the absolute value of the coefficient of the j-th parameter among the formation property parameters, crude oil property parameters, and engineering parameters in the model; is the sum of the absolute values of the coefficients of the formation property parameters, crude oil property parameters, and engineering parameters in the model.

[0050] Based on a comprehensive analysis of formation property parameters, crude oil property parameters, engineering parameters, and demulsification effect parameters, the technical solution provided by the present invention establishes a mathematical model for the demulsification effect of high-acid crude oil under the comprehensive action of multi-dimensional factors, and clarifies the main controlling factors affecting the demulsification effect of crude oil. The technical solution provided by the present invention can more accurately determine the factors affecting the demulsification effect when using a target demulsifier for high-acid crude oil demulsification, which is of great significance for guiding the use of the target demulsifier. The technical solution provided by the present invention takes into account the influence of multiple factors on the demulsification effect, allows finding the optimal solution in the continuous parameter space, can more comprehensively consider multi-objective optimization, and improves the accuracy of determining the main controlling factors of the crude oil demulsification effect. Specific embodiments

[0051] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer and more understandable, the embodiments of the present invention will be further described in detail below. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0052] In the description of this specification, the terms "including", "comprising", "having", "containing", etc. are all open-ended terms, meaning including but not limited to. The descriptions referring to terms such as "one embodiment", "one specific embodiment", "some embodiments", "for example", etc. mean that the specific features, structures or characteristics described in connection with that embodiment or example are included in at least one embodiment or example of this application. In this specification, the schematic descriptions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures or characteristics described can be combined in a suitable manner in any one or more embodiments or examples. The order of steps involved in each embodiment is used to schematically illustrate the implementation of this application, and the order of steps is not limited and can be adjusted appropriately as needed.

[0053] Embodiment 1:

[0054] This embodiment provides a method for determining the main control factors for demulsification of high-acid-value naphthenic crude oil, and the method includes:

[0055] 1. Obtain demulsification data for demulsifying high-acid-value naphthenic crude oil with different physical properties in different formations using the target demulsifier through literature research data and laboratory experiment data; wherein, the demulsification data includes formation property parameter data, crude oil property parameter data, engineering parameter data, and demulsification effect parameter data;

[0056] Among them, the formation property parameters include: clay content X20, whole-rock mineral content; the whole-rock mineral content includes quartz content X11, potassium feldspar content X12, plagioclase content X13, calcite content X14, dolomite content X15, halite content X16, analcime content X17, barite content X18, anatase content X19;

[0057] Among them, the crude oil property parameters include: crude oil density X1 at 20 °C, crude oil viscosity X2 at 50 °C, crude oil pour point X3, crude oil sulfur content X4, crude oil wax content X5, initial boiling point X6 of crude oil, final boiling point X7 of crude oil, wax precipitation point X8 of crude oil, content of gum and asphalt in crude oil X9, crude oil acidity X10, flash point X24 of crude oil, initial water content X25 of crude oil, water content X26 of the oil layer after 4 hours of crude oil sedimentation;

[0058] Among them, the engineering parameters include demulsifier dosage X23, dosing cycle X22, temperature X21 when the demulsifier is injected;

[0059] Among them, the demulsification effect parameter selects the water content Y after crude oil demulsification.

[0060] The results are shown in Table 1.

[0061] Table 1

[0062]

[0063]

[0064] 2. Visualize and analyze the obtained demulsification data through a frequency diagram to determine the extreme values, mean, variance, and distribution characteristics of the demulsification data. If the demulsification data conforms to a normal distribution, proceed to the subsequent step 3; if the demulsification data does not conform to a normal distribution, re-obtain the demulsification data.

[0065] In this embodiment, the demulsification data conforms to a normal distribution, and the selected samples are universal, scientific, and rigorous.

[0066] 3. Normalize the demulsification data to the same interval [0, 1] using the Min - Max method to obtain the normalized demulsification data.

[0067] The clay content X20, quartz content X11, potassium feldspar content X12, plagioclase content X13, calcite content X14, dolomite content X15, halite content X16, analcime content X17, barite content X18, anatase content X19, crude oil density at 20°C X1, crude oil viscosity at 50°C X2, crude oil pour point X3, crude oil sulfur content X4, crude oil wax content X5, initial boiling point of crude oil X6, final boiling point of crude oil X7, wax precipitation point of crude oil X8, content of gum and asphalt in crude oil X9, acidity of crude oil X10, flash point of crude oil X24, initial water content of crude oil X25, water content of the oil layer after 4 - hour sedimentation of crude oil X26, demulsifier dosage X23, dosing cycle X22, temperature at the time of demulsifier injection X21, and water content of crude oil after demulsification Y are respectively normalized to obtain the normalized clay content XX20, normalized quartz content XX11, normalized potassium feldspar content XX12, normalized plagioclase content XX13, normalized calcite content XX14, normalized dolomite content XX15, normalized halite content XX16, normalized analcime content XX17, normalized barite content XX18, normalized anatase content XX19, normalized crude oil density at 20°C XX1, normalized crude oil viscosity at 50°C XX2, normalized crude oil pour point XX3, normalized crude oil sulfur content XX4, normalized crude oil wax content XX5, normalized initial boiling point of crude oil XX6, normalized final boiling point of crude oil XX7, normalized wax precipitation point of crude oil XX8, normalized content of gum and asphalt in crude oil XX9, normalized acidity of crude oil XX10, normalized flash point of crude oil XX24, normalized initial water content of crude oil XX25, normalized water content of the oil layer after 4 - hour sedimentation of crude oil XX26, normalized demulsifier dosage XX23, normalized dosing cycle XX22, normalized temperature at the time of demulsifier injection XX21, and normalized water content of crude oil after demulsification YY.

[0068] 4. Based on the demulsification data after normalization, pairwise scatter plots are respectively drawn between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters. Then, the pairwise mathematical relationships between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters are respectively fitted. Next, Pearson correlation analysis is used to determine the relationships between various formation property parameters, various crude oil property parameters, and various engineering parameters and the demulsification effect parameters.

[0069] The single-factor fitting relationship of X1 is: YY = -33.429632 + 38.382090XX1

[0070] The single-factor fitting relationship of X2 is: YY = 0.871526 + 0.004006XX2

[0071] The single-factor fitting relationship of X3 is: YY = 0.273977 - 0.22560XX3

[0072] The single-factor fitting relationship of X4 is: YY = 6.769214 - 4.933839XX4

[0073] The single-factor fitting relationship of X5 is: YY = 3.538332 - 2.66681XX5

[0074] The single-factor fitting relationship of X6 is: YY = -2.113277 + 3.8628XX6

[0075] The single-factor fitting relationship of X7 is: YY = -1.285159 + 0.40455XX7

[0076] The single-factor fitting relationship of X8 is: YY = 4.152554 - 6.2065XX8

[0077] The single-factor fitting relationship of X9 is: YY = 0.233866 + 0.089764XX9

[0078] The single-factor fitting relationship of X10 is: YY = -0.146426 + 0.587107XX10

[0079] The single-factor fitting relationship of X11 is: YY = 0.733333 + 0.000000XX11

[0080] The single-factor fitting relationship of X12 is: YY = 0.733333 + 0.000000XX12

[0081] The single-factor fitting relationship of X13 is: YY = 0.733333 + 0.000000XX13

[0082] The single-factor fitting relationship of X14 is: YY = 17.1000 - 16.250000XX14

[0083] The single-factor fitting relationship of X15 is: YY = 0.733333 + 0.000000XX15

[0084] The single-factor fitting relationship of X16 is: YY = -15.81100 + 16.250000XX16

[0085] The single-factor fitting relationship of X17 is: YY = 0.733333 + 0.000000XX17

[0086] The single-factor fitting relationship of X18 is: YY = 0.733333 + 0.000000XX18

[0087] The single-factor fitting relationship of X19 is: YY = 0.733333 + 0.000000XX19

[0088] The single-factor fitting relationship of X20 is: YY = 0.733333 + 0.000000XX20

[0089] The single-factor fitting relationship of X21 is: YY = 0.733333 - 0.000000XX21

[0090] The single-factor fitting relationship of X22 is: YY = 0.733333 - 0.000000XX22

[0091] The single-factor fitting relationship of X23 is: YY = 0.733333 - 0.000000XX23

[0092] The single-factor fitting relationship of X24 is: YY = 0.733333 + 0.000000XX24

[0093] The single-factor fitting relationship of X25 is: YY = 0.733333 - 0.000000XX25

[0094] The single-factor fitting relationship of X26 is: YY = 0.733333 + 0.000000XX26

[0095] 5. Divide the normalized demulsification data into a test set and a training set. In this embodiment, the normalized data of groups 1, 2, and 3 in Table 1 are the test set, and the normalized data of groups 4, 5, and 6 are the training set. Based on the test set and training set data, use multiple regression, ridge regression, lasso regression, and stepwise regression to fit the relationship between the demulsification effect parameters and the formation property parameters, crude oil property parameters, and engineering parameters respectively, to obtain a multiple regression model for demulsification effect, a ridge regression model for demulsification effect, a lasso regression model for demulsification effect, and a stepwise regression model for demulsification effect, and determine the weighted evaluation index of each model respectively; among them, the weighted evaluation index = 0.5×MSE + 0.2×MAE + 0.2×RMSE + 0.1×(1 - coefficient of determination R 2 ), MSE refers to the mean square error, MAE refers to the mean absolute error, RMSE refers to the root mean square error, R 2 refers to the coefficient of determination;

[0096] Among them, the multiple regression model for demulsification effect is determined by the following formula:

[0097] y = Xβ + ε

[0098] In the formula: y is the response vector, X is the design matrix, β is the parameter vector, and ε is the error vector;

[0099] The least squares estimate of its parameters is:

[0100] β = (X T X) -1 X T y

[0101] In the formula: y is the response vector, X is the design matrix, and β is the parameter vector;

[0102] To minimize the difference between the model predicted value and the actual observed value, that is, determine the model parameters by the sum of squared residuals, and the sum of squared residuals is defined as:

[0103] RSS(β) = Σ(y i - x i T β) 2

[0104] In the formula: y i is the response value corresponding to the i-th matrix element, x i is the i-th matrix element, and RSS(β) is the sum of squared residuals of the parameter vector;

[0105] Among them, the ridge regression model for demulsification effect is determined by the following method:

[0106] After giving different design matrices X and response vectors y, determine J(β) = Σ(y i - x iT β) 2 + λΣβ j 2 When J(β) = Σ(y i - x i T β) 2 + λΣβ j 2 is at its minimum, the demulsification effect ridge regression model can be obtained; where: λ is a non - negative regularization parameter that determines the strength of regularization; the second term λΣ|β j | is the regularization term, which penalizes the absolute value of the coefficients; when λ = 0, ridge regression degenerates to standard linear regression, and as λ increases, the influence of regularization increases and the estimated values of the coefficients shrink;

[0107] Among them, the demulsification effect lasso regression model is determined as follows:

[0108] After giving different design matrices X and response vectors y, determine J(β) = Σ(y i - x i T β) 2 + λΣ|β j |, when J(β) = Σ(y i - x i T β) 2 + λΣ|β j | is at its minimum, the demulsification effect lasso regression model can be obtained; where: λ is a non - negative regularization parameter that determines the strength of regularization; the second term λΣ|β j | is the regularization term, which penalizes the absolute value of the coefficients;

[0109] Among them, the demulsification effect stepwise regression model is determined by the stepwise regression iteration method, which is divided into forward selection, backward elimination or bidirectional stepwise process; in forward selection, the model starts without predictor variables and then adds predictor variables one by one, each time selecting the predictor variable that provides the greatest improvement; in backward elimination, the model starts with all predictor variables and then deletes predictor variables one by one, each time deleting the predictor variable that provides the least improvement to the model; bidirectional stepwise regression combines forward selection and backward elimination, considering both adding and deleting predictor variables; the decision at each step is based on the Akaike Information Criterion (AIC);

[0110] Among them, according to the Akaike Information Criterion, the smaller the AIC value of the model, the better it is. AIC = 2k - 2ln(L), where: k is the number of parameters in the model; L is the maximum likelihood of the model.

[0111] In this embodiment, the demulsification effect multiple regression model is:

[0112] YY = 0.0151744734XX2 - 0.2156646151XX6 + 0.3739089082XX7 + 0.4641529961XX8 - 0.1368554593XX9 - 0.2746455669XX16

[0113] Among them, among the 26 parameters, only the coefficients of XX2, XX6, XX7, XX8, XX9, and XX16 are not zero, indicating that the effective parameters of this method are the above parameters.

[0114] The mean square error MSE value of the multiple regression model is 0.0474, the root mean square error RMSE value is 0.0382, the mean absolute error MAE value is 0.0211, the coefficient of determination value is 0.0178, and the weighted evaluation index is 1.337756e-01.

[0115] In this embodiment, the demulsification effect ridge regression model is:

[0116] YY = 0.7499725398 - 0.0000256675XX1 + 0.0055725857XX2 + 0.0001873785XX3 - 0.0000237000XX4 - 0.0000376257XX5 - 0.0021261409XX6 - 0.0026686883XX7 - 0.0005629154XX8 - 0.0008027071XX9 - 0.0001653029XX10 - 0.0000685273XX14 - 0.0026774931XX16 - 0.0021040198XX21 - 0.0000121037XX22 - 0.0068934094XX23 - 0.0006286024XX24 - 0.0008883881XX25 - 0.0004608081XX26

[0117] The mean square error MSE value of the ridge regression model is 1.2022e-28, the root mean square error RMSE value is 1.6180e-15, the mean absolute error MAE value is 1.2130e-15, the coefficient of determination value is 0.0309, and the weighted evaluation index is 9.691358e-02.

[0118] In this embodiment, the lasso regression model for the demulsification effect is: YY = 0.7499725398 - 0.0000229077XX1 + 0.0050346262XX2 + 0.0001514526XX3 - 0.0000219202XX4 - 0.0000376149XX5 - 0.0019821761XX6 - 0.0024304759XX7 - 0.0005484089XX8 - 0.0007513521XX9 - 0.0001510126XX10 - 0.0000626418XX14 - 0.0024788497XX16 - 0.0019570564XX21 - 0.0000098460XX22 - 0.0063549588XX23 - 0.0005815853XX24 - 0.0008270538XX25 - 0.0004260289XX26

[0119] The mean square error MSE value of the lasso regression model is 0.0458, the root mean square error RMSE value is 0.0375, the mean absolute error MAE value is 0.0221, the coefficient of determination value is 0.0183, and the weighted evaluation index is 1.330155e-01.

[0120] In this embodiment, the calculation result of the stepwise regression method for the demulsification effect is:

[0121] YY = -89.805477 + 0.203777XX7 - 0.148123XX8 + 0.129656XX9 - 9.892106XX10 + 92.455576XX16

[0122] Through calculation, the contribution coefficient of the final boiling point X7 is 0.001982; the contribution coefficient of the wax appearance point X8 is 0.001440; the contribution coefficient of the content of crude oil gum and asphalt X9 is 0.001261; the contribution coefficient of the acidity X10 is 0.096199; the contribution coefficient of the salt content X16 is 0.899118.

[0123] The evaluation parameter variance MSE value of the stepwise regression model is 4.0384e-13, the root mean square error RMSE value is 4.0384e-13, the mean absolute error MAE value is 3.6790e-13, the coefficient of determination value is 0.0309, and the weighted evaluation index is 0.09691358.

[0124] 6. Determine the model with the smallest weighted evaluation index based on the weighted evaluation indices of each model; if there is only one model with the smallest weighted evaluation index, then use this model with the smallest weighted evaluation index as the demulsification effect model; if there are two or more models with the smallest weighted evaluation index, then select the optimal model from these models with the smallest weighted evaluation index as the demulsification effect model according to the coefficient of determination, mean square error, root mean square error, and mean absolute error.

[0125] The smaller the weighted evaluation index, the better the model fitting degree. According to the above results, the weighted evaluation indices of the ridge regression model and the stepwise regression model are equal and the smallest; compare the coefficient of determination R 2 of the ridge regression model and the stepwise regression model with the smallest weighted evaluation index, the coefficient of determination R 2 of the ridge regression model and the stepwise regression model with the smallest weighted evaluation index are equal; therefore, it is necessary to compare other evaluation indices of the ridge regression model and the stepwise regression model. Analysis shows that the mean square error, root mean square error, and mean absolute error of the ridge regression model are all smaller than the corresponding indices of the stepwise regression model, that is, it indicates that the ridge regression model is the best model. Therefore, the ridge regression model is selected as the demulsification effect model.

[0126] 7. Conduct an analysis of the contribution rates of various formation property parameters, crude oil property parameters, and engineering parameters to the demulsification effect parameters for the demulsification effect model, and determine the top 6 parameters with the largest contribution rates to the demulsification effect parameters as the main control factors during the demulsification process of high-acid-value naphthenic crude oil using the target demulsifier.

[0127] Among them, the contribution rate is determined by the following formula:

[0128]

[0129] In the formula: CR i is the contribution rate of the i-th parameter among various formation property parameters, crude oil property parameters, and engineering parameters in the model; a i is the coefficient of the i-th parameter among various formation property parameters, crude oil property parameters, and engineering parameters in the model; |a j | is the absolute value of the coefficient of the j-th parameter among various formation property parameters, crude oil property parameters, and engineering parameters in the model; is the sum of the absolute values of the coefficients of various formation property parameters, crude oil property parameters, and engineering parameters in the model.

[0130] In the demulsification effect model, the contribution coefficient of the crude oil density at 20°C is 0.0010; the contribution coefficient of the crude oil viscosity at 50°C is 0.2151; the contribution coefficient of the crude oil pour point is 0.0072; the contribution coefficient of the sulfur content in the crude oil is 0.0009; the contribution coefficient of the wax content in the crude oil is 0.0015; the contribution coefficient of the initial boiling point of the crude oil is 0.0821; the contribution coefficient of the final boiling point of the crude oil is 0.1030; the contribution coefficient of the wax precipitation point of the crude oil is 0.0217; the contribution coefficient of the gum + asphalt content is 0.0310; the contribution coefficient of the acidity of the crude oil is 0.0064; the contribution coefficient of the calcite content is 0.0026; the contribution coefficient of the halite content is 0.1034; the contribution coefficient of the temperature at the time of demulsifier injection is 0.0812; the contribution coefficient of the chemical addition cycle is 0.0005; the contribution coefficient of the demulsifier dosage is 0.2661; the contribution coefficient of the flash point of the crude oil is 0.0243; the contribution coefficient of the initial water content of the crude oil is 0.0343; the contribution coefficient of the water content of the oil after 4h sedimentation time is 0.0178.

[0131] The main control factors are arranged in descending order as follows:

[0132] The contribution coefficient of the demulsifier dosage is 0.2661; the contribution coefficient of the crude oil viscosity at 50°C is 0.2151; the contribution coefficient of the halite content is 0.1034; the contribution coefficient of the final boiling point of the crude oil is 0.1030; the contribution coefficient of the initial boiling point of the crude oil is 0.0821; the contribution coefficient of the temperature at the time of demulsifier injection is 0.0812.

[0133] Analyzing the above results, it can be seen that the demulsifier dosage and the temperature at the time of demulsifier injection are the most important controllable engineering factors, and other factors are the most important uncontrollable factors; among them, the temperature at the time of demulsifier injection has a certain influence on the viscosity. Therefore, for a given treatment object, its own properties remain unchanged within a certain range, and the demulsification problem can be optimally solved by controlling the demulsifier dosage and the temperature at the time of demulsifier injection.

[0134] The specific embodiments described above further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific embodiments of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for determining the main controlling factors of demulsification of high acid value naphthenic crude oil, the method comprising: Obtaining demulsification data of demulsifying high acid value cycloalkyl crude oil with different physical properties in different formations using a target demulsifier; wherein the demulsification data includes formation property parameter data, crude oil property parameter data, engineering parameter data and demulsification effect parameter data; Normalizing the demulsification data to the same interval to obtain normalized demulsification data; Based on the normalized demulsification data, multiple regression, ridge regression, lasso regression and stepwise regression were used to fit the relationship between demulsification effect parameters and formation property parameters, crude oil property parameters and engineering parameters, and the demulsification effect multiple regression model, demulsification effect ridge regression model, demulsification effect lasso regression model and demulsification effect stepwise regression model were obtained, and the weighted evaluation index of each model was determined respectively; among which, the weighted evaluation index = 0.5×MSE+0.2×MAE+0.2×RMSE+0.1×(1-determination coefficient R 2 ), MSE refers to mean square error, MAE refers to mean absolute error, RMSE refers to root mean square error, R 2 refers to the coefficient of determination; Based on the weighted evaluation index of each model, determine the model with the smallest weighted evaluation index; if there is only one model with the smallest weighted evaluation index, then use the model with the smallest weighted evaluation index as the demulsification effect model; if there are two or more models with the smallest weighted evaluation index, then select the best model from the models with the smallest weighted evaluation index as the demulsification effect model based on at least one of the determination coefficient, mean square error, root mean square difference, and mean absolute error; The contribution rate of formation property parameters, crude oil property parameters and engineering parameters to the demulsification effect parameters is analyzed for the demulsification effect model, and the top N parameters with the largest contribution rate to the demulsification effect parameters are determined to be the main controlling factors in the demulsification process of high acid value cycloalkyl crude oil using the target demulsifier.

2. The method according to claim 1, wherein: Stratigraphic property parameters include: clay content and whole-rock mineral content.

3. The method according to claim 2, wherein: The whole rock mineral content includes quartz content, potassium feldspar content, plagioclase content, calcite content, dolomite content, halite content, analcime content, barite content, and anatase content.

4. The method according to claim 1, wherein: Crude oil property parameters include: crude oil density, crude oil viscosity, crude oil pour point, crude oil sulfur content, crude oil wax content, crude oil initial distillation point, crude oil final distillation point, crude oil wax precipitation point, crude oil colloid and asphalt content, crude oil acidity, crude oil flash point, crude oil initial water content, and oil layer water content after crude oil sedimentation for 4 hours.

5. The method according to claim 1, wherein: Engineering parameters include demulsifier dosage, dosing cycle, and demulsifier injection temperature.

6. The method according to claim 1, wherein: The demulsification effect parameter is the dehydration rate or the water content of the crude oil after demulsification.

7. The method according to claim 1, wherein: The method further includes: Before normalizing the demulsification data to the same interval, determine whether the demulsification data conforms to the normal distribution. If the demulsification data conforms to the normal distribution, perform the subsequent steps before normalizing the demulsification data to the same interval. If the demulsification data does not conform to the normal distribution, reacquire the demulsification data.

8. The method according to claim 1, wherein: The method further includes: Based on the normalized demulsification data, pairwise factor analysis was performed to determine the relationships between the formation property parameters, crude oil property parameters, engineering parameters and demulsification effect parameters.

9. The method according to claim 8, wherein: The two-factor analysis based on the normalized demulsification data includes: Based on the normalized demulsification data, pairwise factor scatter plots between each formation property parameter, each crude oil property parameter, each engineering parameter and demulsification effect parameter are drawn respectively, and then the pairwise factor mathematical relationships between each formation property parameter, each crude oil property parameter, each engineering parameter and demulsification effect parameter are fitted respectively, and then Pearson correlation analysis, Spearman correlation analysis, grey correlation analysis or significance analysis are used to determine the relationship between each formation property parameter, each crude oil property parameter, each engineering parameter and demulsification effect parameter.

10. The method according to claim 1, wherein: Normalizing the demulsification data to the same interval using the Min-Max method or the Z-score method; and / or In the process of normalizing the demulsification data to the same interval, the demulsification data are normalized to [0, 1].

11. The method according to claim 1, wherein: The contribution rate is determined by the following formula: Where: CR i is the contribution rate of the ith parameter among the formation property parameters, crude oil property parameters and engineering parameters in the model; a i is the coefficient of the ith parameter among the formation property parameters, crude oil property parameters and engineering parameters in the model; |a j | is the absolute value of the coefficient of the jth parameter among the formation property parameters, crude oil property parameters and engineering parameters in the model; It is the sum of the absolute values ​​of the coefficients of the formation property parameters, crude oil property parameters and engineering parameters in the model.

Images