Device and method for intelligently optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced liquid

Abstract

The invention relates to a bottle test device and method for intelligently optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced liquid, and the device comprises a test cabin module, a visual demulsification bottle test module, a camera, an image acquisition system and an intelligent terminal. The diversified chemical components are uniformly mixed in a stirring domain of the test cabin module, and a chemical flooding produced liquid sample with a certain component proportion is prepared; the visual demulsification bottle test module realizes visual recording and data storage of an oil-water interface evolution process of a chemical flooding produced liquid sample under the action of a high-frequency pulsed electric field; and the intelligent terminal processes the high-frequency pulse electric dehydration image, and constructs a chemical flooding produced liquid high-frequency pulse electric dehydration bottle test data set with different chemical component proportions. According to the method, qualitative description of the high-frequency pulse electric dehydration evolution process is changed into quantitative characterization, the produced liquid high-frequency pulse electric dehydration parameters are optimized through the artificial intelligence algorithm, and stable and efficient operation of the oilfield produced liquid treatment process is achieved.

Description

Technical Field:

[0001] The present invention relates to the operation technology of electro-dehydration treatment process for oilfield produced fluids, and particularly to a bottle test device and method for intelligently optimizing the high-frequency pulsed electro-dehydration parameters of chemically flooded produced fluids. Background Art:

[0002] The mechanism of enhancing oil recovery by chemical flooding technology lies in injecting multi-component chemicals with different properties into the formation. Among them, according to the action mechanism, it can be divided into polymer flooding technology for improving the oil-water mobility ratio to expand the swept volume, surfactant flooding technology for reducing the oil-water interfacial tension to form pore surface wetting reversal, alkali flooding technology for dissolving the rigid film to promote emulsification trapping and entrainment, and chemical composite flooding technology in which more than two components are proportioned to each other to synergistically improve the efficiency. However, in this oil production mode of enhancing oil recovery by chemical flooding, after the multi-component chemicals are continuously injected into the formation, they will return with the produced oil fluids one after another. Active components such as surfactants and polymers will polarize, aggregate, co-adsorb, etc. near the interfacial film of emulsion droplets, and then change the molecular configuration and mechanical structure of the adsorbed layer of the interfacial film, seriously aggravating the emulsification degree of the produced fluids. The stability of the interfacial film of its emulsion droplets is increasing day by day, which makes the pressure and challenges of oil-water separation of chemically flooded produced fluids continuously appear. It not only directly affects the stable and efficient operation of the oilfield produced fluid treatment process, but also restricts the construction of intelligent oilfields.

[0003] At present, for the oil-water separation link of the produced fluid, the commonly used process is high-frequency pulsed electro-dehydration. It conducts demulsification treatment by causing the interfacial film of emulsion droplets to vibrate periodically through a high-frequency transformed pulsed electric field. It has become a consensus that the pulsed electric field parameters during high-frequency pulsed electro-dehydration will affect the oil-water separation effect of chemically enhanced oil recovery (CEOR) produced fluid. The internal mechanism lies in that there is an adsorption layer with charge distribution characteristics on the interfacial film of emulsion droplets. And the alternating loads received by the adsorption layer in different pulsed electric fields are different, resulting in different degrees of fatigue, demulsification, and coalescence of the interfacial film of CEOR produced fluid. Especially for CEOR produced fluid containing multiple components, under the co-adsorption effect of these multiple components, the mechanical response of the interfacial film of emulsion droplets caused by the pulsed electric field is more complex and changeable, making it difficult for the high-frequency pulsed electro-dehydration process of CEOR produced fluid to operate stably and efficiently and to implement intelligent control. Therefore, in recent years, regarding the personalized determination of the operating parameter limits of the electro-dehydration treatment process for oilfield surface produced fluid, and the reliable optimization of the operating parameters of high-frequency pulsed electro-dehydration, and then realizing the efficient oil-water separation of CEOR produced fluid, have become the key points and difficulties in the project. However, the existing understandings are all based on large-volume sample simulation tests or on the empirical laws of production operation to optimize the operating parameters of pulsed electro-dehydration. Although this approach provides a basis for the design and management of the high-frequency pulsed electro-dehydration treatment process, it has not yet and is difficult to consider the differences in the emulsification characteristics such as the charge distribution and mechanical response of the interfacial film of emulsion droplets under different chemical component ratios. That is, there is a lack of an intelligent optimization device and method for the high-frequency pulsed electro-dehydration parameters of CEOR produced fluid that can adapt to changes in chemical component ratios, which directly affects the scientific design and efficiency improvement of the CEOR produced fluid treatment process, and also restricts the research and development of high-efficiency pulsed electro-dehydration equipment, as well as the mutual coordination and feedback between historical operation data and the overall operation efficiency under the background of intelligent oilfield construction. Therefore, this poses the scientific problem of scientifically and reliably optimizing the high-frequency pulsed electro-dehydration parameters of CEOR produced fluid based on neural network to intelligently predict and feedback changes in chemical component ratios and based on small-volume sample bottle test devices. It breaks through the limitations and problems in traditional large-volume sample simulation devices and test methods, such as considering only a single chemical component of the produced fluid, the optimization process of high-frequency pulsed electro-dehydration parameters lacking quantification and relying more on production experience, especially the unclear evolution process of high-frequency pulsed electro-dehydration of CEOR produced fluid and the unclear description of the details of the demulsification mechanism of the alternating load applied by the pulsed electric field. Therefore, it is particularly necessary to form an intelligent optimization bottle test device and method for the high-frequency pulsed electro-dehydration parameters of CEOR produced fluid. Summary of the Invention:

[0004] An object of the present invention is to provide a device for intelligently optimizing the high-frequency pulse electro-dehydration parameters of chemically flooded produced fluids. This device for intelligently optimizing the high-frequency pulse electro-dehydration parameters of chemically flooded produced fluids is used to solve the problem of visual characterization in the high-frequency pulse electro-dehydration process of chemically flooded produced fluids. In particular, considering the complex and variable chemical component ratios and resulting emulsification characteristics in chemically flooded produced fluids, it addresses the technical challenge of scientifically and reliably optimizing the high-frequency pulse electro-dehydration parameters of produced fluids in the context of empowering intelligent oilfield construction with artificial intelligence algorithms. Another object of the present invention is to provide an optimization method for this device for intelligently optimizing the high-frequency pulse electro-dehydration parameters of chemically flooded produced fluids.

[0005] The technical solution adopted by the present invention to solve its technical problems is as follows: This device for intelligently optimizing the high-frequency pulse electro-dehydration parameters of chemically flooded produced fluids includes a test chamber module, a visual demulsification bottle test module, a high-speed camera, an image acquisition system, a data storage system, and an intelligent terminal. Different multi-component chemicals are input into the parallel chemical syringes in the test chamber module. The multi-component chemicals include polymers, surfactants, and alkalis. The intelligent terminal controls the outlet valves of each chemical component syringe and sequentially inputs the multi-component chemicals into the stirring area through the input pipeline according to a certain chemical component ratio. A micro flowmeter is set between the input pipeline and the intelligent terminal to control the total input amount of the multi-component chemicals. A layer of annular water bath sandwich is arranged outside the main body of the stirring area to simulate the potential component factors and temperature environment in the emulsification process of chemically flooded produced fluids. The multi-component chemicals are mixed evenly in the stirring area to form a chemically flooded produced fluid sample with a certain component ratio.

[0006] The prepared chemically flooded produced fluid sample is transported to each bottle test chamber of the visual demulsification bottle test module. An electrode cover is set at the upper port of the bottle test chamber. A positive electrode conductor and a negative electrode conductor are respectively embedded on the left and right sides of the electrode cover. The positive electrode conductor and the negative electrode conductor extend downward and are respectively connected to a rectangular sheet positive electrode and a negative electrode placed opposite to each other to form a high-frequency pulse electric field. Each bottle test chamber is fixed in a metal reaction kettle. A water bath sandwich is arranged between the bottle test chamber and the metal reaction kettle. The gas cylinder pressurizing device controls the operating pressure in each bottle test chamber. The metal reaction kettle is provided with a glass viewing window. A high-speed camera is connected to the image acquisition system to record the oil-water interface image information at any time and transmit the information to the data storage system, realizing the visual recording and data storage of the evolution process of the oil-water interface of the chemically flooded produced fluid sample under the action of the high-frequency pulse electric field.

[0007] The intelligent terminal performs grayscale processing on the high-frequency pulse electro-dehydration image to obtain the oil-water interface distribution state at the x coordinate of the X axis in the image at time t:

[0008]

[0009] Where, They are the thicknesses of the aqueous phase region, the oil phase region, and the oil-water transition region at the X-axis coordinate x in the image at time t, in m; It is the maximum value of the clustering samples in the aqueous phase region at the X-axis coordinate x in the image at time t, in m; It is the maximum value of the clustering samples in the oil-water transition region at the X-axis coordinate x in the image at time t, in m; h is the total height of the chemical flooding produced fluid sample in the bottle test chamber, in m;

[0010] Then, taking the thicknesses of the aqueous phase region, the oil phase region, and the oil-water transition region as prediction indicators, and selecting polymer concentration, surfactant concentration, alkali concentration, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time as characteristic variables, a high-frequency pulsed electric dehydration bottle test dataset of chemical flooding produced fluid with a certain chemical component ratio is constructed;

[0011] Change the chemical component ratio, construct a high-frequency pulsed electric dehydration bottle test dataset of chemical flooding produced fluid with another chemical component ratio, repeat the same method, and construct high-frequency pulsed electric dehydration bottle test datasets of chemical flooding produced fluid with different chemical component ratios.

[0012] In the above scheme, the stirring domain is a three-layer coaxial cylinder structure. The rectangular inner rotor connected to the lower side of the rotating motor is located at the center of the cylinder. A number of circular channels penetrating both sides are drilled on the surface of the rectangular inner rotor, so that the multi-component chemical components are mixed and emulsified with each other after flowing through the circular channels during the rotation of the rectangular inner rotor; the cylindrical outer rotor located outside the rectangular inner rotor is connected upward to another rotating motor, and a number of annular gaps are set in the upper and lower circular cross-sections of the cylindrical outer rotor to ensure that the multi-component chemical components outside the cylindrical outer rotor can flow into the inner side, and then through the reverse rotation of the rectangular inner rotor and the cylindrical outer rotor, the multi-component chemical components are evenly mixed to prepare a chemical flooding produced fluid sample with a certain emulsified droplet particle size; the water bath temperature control device pumps hot water at a certain temperature to the upper end inlet of the annular water bath sandwich, continuously exchanges heat with the chemical flooding produced fluid sample in the stirring domain, flows through the bottom outlet of the annular water bath sandwich, and returns to the water bath temperature control device, and the temperature sensor in the water bath temperature control device is connected to the intelligent terminal, and its water bath temperature is controlled by the set temperature of the intelligent terminal.

[0013] In the above scheme, four bottle test chambers are arranged in the metal reaction kettle, and there are emptying ports at the bottom of each bottle test chamber, which are connected to the emptying pipeline; a glass viewing window is arranged on the front of the metal reaction kettle, and the upper side of the glass viewing window corresponding to each bottle test chamber extends 30 mm above the positive electrode, and the lower side extends to the bottom of the bottle test chamber; the intelligent terminal is sequentially connected to the rectifying unit, the modulating unit, and the amplifying unit to form a high-frequency pulsed electric signal with a certain frequency, a certain voltage, and a certain duty cycle, and the positive electrode and the negative electrode of the output end of the amplifying unit are respectively connected to the positive electrode contact piece and the negative electrode contact piece on the upper side of the electrode cover by wires, and a high-frequency pulsed electric field is formed in the area where the positive electrode and the negative electrode in the bottle test chamber are facing each other.

[0014] In the above solution, the rectangular inner rotor and the cylindrical outer rotor are connected by rolling bearings. The inner diameter of the circular channel on the surface of the rectangular inner rotor is 5 mm, the width of the annular gap of the upper and lower circular cross-sections of the cylindrical outer rotor is 10 mm, and the interval between the rectangular inner rotor and the cylindrical outer rotor is 30 mm.

[0015] In the above solution, the metal reactor and the bottle test chamber are filled with melamine foam heat insulation material, and the aluminum outer shell of the bottle test chamber is connected to the grounding line to avoid mutual interference of high-frequency pulse electric fields between the bottle test chambers.

[0016] In the above solution, the length of the bottle test chamber is 30 mm, the width is 12 mm, the height is 10 mm, the positive electrode is 40 mm away from the top of the bottle test chamber, and the negative electrode is 40 mm away from the bottom of the bottle test chamber.

[0017] In the above solution, the widths of the annular water bath sandwich outside the stirring area and the water bath sandwich outside the bottle test chamber are both 10 mm.

[0018] In the above solution, the input voltage of the rectifying unit is 220 V, the output pulse electric signal voltage of the amplifying unit is continuously adjustable from 0 to 6000 V, and the output current is continuously adjustable from 0 to 100 mA.

[0019] In the above solution, the pixel of the high-speed camera is 1024×1024, and the shooting frequency is 16,000 frames per second.

[0020] A method for optimizing the high-frequency pulse electro-dehydration process parameters of chemically enhanced oil recovery produced liquid by using the intelligent optimization device for high-frequency pulse electro-dehydration parameters of chemically enhanced oil recovery produced liquid:

[0021] Step 1: Use the intelligent optimization device for high-frequency pulse electro-dehydration parameters of chemically enhanced oil recovery produced liquid to obtain a high-frequency pulse electro-dehydration bottle test data set of chemically enhanced oil recovery produced liquid with different chemical component ratios. In the data set, the thickness of the aqueous phase region, the thickness of the oil phase region, and the thickness of the oil-water transition region are prediction indicators, and the polymer concentration, surfactant concentration, alkali concentration, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time are characteristic variables;

[0022] Step 2: After integrating and normalizing the high-frequency pulse electro-dehydration bottle test data set of chemically enhanced oil recovery produced liquid with different chemical component ratios, divide it into a training set and a test set according to a sample ratio of 75% and 25%. Adopt the ELM neural network algorithm to construct a high-frequency pulse electro-dehydration performance prediction model of chemically enhanced oil recovery produced liquid based on the ELM neural network;

[0023] Step 3: Optimize the high-frequency pulse electro-dehydration process parameters of chemically enhanced oil recovery produced liquid:

[0024] During the high-frequency pulsed electro-dehydration process of chemically enhanced oil recovery (CEOR) produced fluids, the thickness of the oil phase region quantitatively characterizes the oil-water separation effect of the produced fluids, and the thickness of the oil-water transition region reflects the adaptability of the high-frequency pulsed electro-dehydration operating parameters to the demulsification and coalescence of emulsion droplets. Taking the thickness of the oil phase region and the thickness of the oil-water transition region as the objective functions, and the value ranges of the characteristic variables in the dataset as the solution domain, a multi-objective nonlinear optimization model is constructed:

[0025]

[0026] In the formula, ELM o (X) is the predicted value of the thickness of the oil phase region corresponding to a certain characteristic variable; ELM m (X) is the predicted value of the thickness of the oil-water transition region corresponding to a certain characteristic variable; X max is the maximum value of the characteristic variable; X min is the minimum value of the characteristic variable;

[0027] Based on the multi-objective particle swarm optimization algorithm, within the solution domain [X min , X max , the particle coordinates and particle velocities of 50 particles are randomly generated to form a population. Among them, the particle coordinates are the characteristic variables. Combining with the above-mentioned high-frequency pulsed electro-dehydration performance prediction model of CEOR produced fluids, the predicted values of the thickness of the oil phase region and the thickness of the oil-water transition region corresponding to each particle in the t-th generation population are respectively obtained, and they are transmitted to the pareto optimal solution set and the particle evolution solution set of the intelligent terminal for optimization. Continuously optimize the pareto optimal solution set and the particle evolution solution set within the range of high-frequency pulsed electro-dehydration operating parameters. When the number of iterations reaches the maximum iteration limit, the iterative optimization of the particles is completed. At this time, the characteristic variables in the pareto optimal solution set are the high-frequency pulsed electro-dehydration process parameters of CEOR produced fluids that meet the optimal dehydration performance.

[0028] Beneficial effects:

[0029] (1) The present invention is directed to the structural design of the high-frequency pulsed electric dehydration emulsification test chamber module for chemically flooded produced fluids. It fully considers the influence of multiple chemical components in chemically flooded produced fluids on the emulsification characteristics of the produced fluids, and the high-frequency pulsed electric dehydration performance is also controlled by the emulsification characteristics of the produced fluids and the pulsed electric field parameters. By connecting the outlet valves of each chemical component syringe through an intelligent terminal to precisely control the chemical component ratio, and continuously supplying heat based on the annular water bath sandwich layer, while fully reproducing the potential component factors and temperature environment of the emulsification process of chemically flooded produced fluids, a reverse-rotating rectangular inner rotor and a cylindrical outer rotor are introduced, so that the multiple chemical components continuously flow through the circular flow channel of the rectangular inner rotor and the annular gap between the circular outer rotor, and the multiple chemical components are uniformly blended and randomly dispersed at the oil-water interface from the mechanism level, thereby forming a chemically flooded produced fluid sample that is the same as the emulsification characteristics, component characteristics, temperature environment, and formation mechanism of the chemically flooded produced fluid, effectively avoiding the deviation of the optimized parameters of the high-frequency pulsed electric dehydration process caused by the difference in emulsification characteristics.

[0030] (2) In the present invention, the intelligent terminal is sequentially connected to a rectifying unit, a modulating unit, and an amplifying unit, and is connected to the positive and negative electrodes of the rectangular sheet through the positive and negative conductors embedded in the electrode cover. This enables the visual demulsification bottle test module to not only independently and real-time control the pulsed electric field parameters in each bottle test chamber, but also construct a longitudinal high-frequency pulsed electric field with the same size scale and consistent mechanical effects in each bottle test chamber, which is beneficial to systematically exploring the influence of different pulsed electric field parameters on the high-frequency pulsed electric dehydration of chemically flooded produced fluids; at the same time, by connecting a high-speed camera, an image acquisition system, and a data storage system through the intelligent terminal, and relying on the glass viewing window on the front of the metal reaction kettle, the visualization reproduction of the demulsification mechanism of the pulsed electric field applying an alternating load to the oil-water interface during the high-frequency pulsed electric dehydration process of chemically flooded produced fluids with different emulsification characteristics can be realized, which can provide a theoretical basis and technical means for deeply revealing the interaction mechanism between multiple components, and can also enrich and expand the theory of oil-water emulsification formation and stability.

[0031] (3) The present invention fully considers the color difference of the images in the bottle test chamber during the high-frequency pulsed electric dehydration process of chemically flooded produced fluids. Starting from the correlation between the image pixel values and gray values, based on the K-means clustering algorithm, it can intelligently read the image information at any time, and calculate the thickness data of the water phase region, the oil-water transition region, and the oil phase region, so as to quantitatively characterize the electric dehydration performance corresponding to different high-frequency pulsed electric dehydration process parameters, and construct a high-frequency pulsed electric dehydration bottle test data set for chemically flooded produced fluids, providing scientific data support for predicting the high-frequency pulsed electric dehydration performance of chemically flooded produced fluids, making it possible to change the high-frequency pulsed electric dehydration process from traditional qualitative description to quantitative characterization.

[0032] (4) For the construction of the high-frequency pulsed electro-dehydration performance prediction model of chemical flooding produced fluid, the present invention not only focuses on the differences in the potential change mechanisms of different prediction indicators in the dataset, and separately constructs the ELM neural network prediction models for the thickness of the aqueous phase region, the thickness of the oil-water transition region, and the thickness of the oil phase region, but also takes into account the complex non-linear mapping relationship between the characteristic variables and the prediction indicators, and introduces the Sigmoid activation function to non-linearly transform the linear output values of the hidden layer, providing a non-linear characterization means for describing the potential process of the oil-water interface being damaged and coalesced under the alternating load of the pulsed electric field, thus effectively ensuring the reliability of the high-frequency pulsed electro-dehydration performance prediction of chemical flooding produced fluid. At the same time, according to the multi-objective particle swarm optimization algorithm, the high-frequency pulsed electro-dehydration process parameters of chemical flooding produced fluid with the optimal dehydration performance are searched within the solution domain, effectively expanding the applicable range of the high-frequency pulsed electro-dehydration performance prediction model of chemical flooding produced fluid, and making the determination of the high-frequency pulsed electro-dehydration process parameters transform from being based on production operation experience to quantitative optimization combined with the image data of the bottle test cabin, providing an example for the optimal design and operation management of the gathering and transportation processing system under the background of intelligent oilfield construction.

[0033] (5) On the basis of the integrated design of the high-frequency pulsed electro-dehydration emulsion test cabin module and the visual demulsification bottle test module for chemical flooding produced fluid, the present invention uses the K-means clustering algorithm to mine the potential multiple prediction indicators in the bottle test cabin images at any time, forms the bottle test dataset of the high-frequency pulsed electro-dehydration of chemical flooding produced fluid, constructs the corresponding high-frequency pulsed electro-dehydration performance prediction model, and combines the multi-objective particle swarm optimization algorithm to obtain the high-frequency pulsed electro-dehydration process parameters of chemical flooding produced fluid with the optimal dehydration performance. The device principle is clear and feasible, and the method is scientific and reliable, breaking through the limitations of the unclear evolution process of the high-frequency pulsed electro-dehydration of chemical flooding produced fluid and the unclear description of the details of the demulsification mechanism of the alternating load applied by the pulsed electric field in the traditional large-volume sample simulation device and test method, effectively providing a bottle test device and method for intelligent optimization of the high-frequency pulsed electro-dehydration parameters of chemical flooding produced fluid, with strong scientificity, operability and practicability, providing reference and basis for promoting the development of efficient separation technologies for complex produced fluids and the research, development and application of new oil-water separation equipment. Description of the Drawings:

[0034] Figure 1 is the structural schematic diagram of the device of the present invention;

[0035] Figure 2 is the principle schematic diagram of the method of the present invention.

[0036] 1 Multi-component chemical substances 2 Input pipeline 3 Chemical syringe 4 Output pipeline 5 Stirring area 6 Intelligent terminal 7 Micro flowmeter 8 Rectangular inner rotor 9 Cylindrical outer rotor 10 Rotating motor 11 Circular flow channel 12 Annular water bath sandwich layer 13 Water bath temperature control device 14 Temperature sensor 15 Chemical flooding produced fluid sample 16 Outlet pipeline 17 Electrode cover 18 Bottle test chamber 19 Metal reaction kettle 20 Venting pipeline 21 Positive electrode conductor 22 Negative electrode conductor 23 Positive electrode 24 Negative electrode 25 Rectifying unit 26 Modulating unit 27 Amplifying unit 28 Positive electrode contact piece 29 Negative electrode contact piece 30 Water bath sandwich layer 31 Gas cylinder pressurizing device 32 Glass viewing window 33 High-speed camera 34 Image acquisition system 35 Data storage system 36 Pixel value sequence 37 Gray value sequence 38 Thickness of aqueous phase region 39 Thickness of oil-water transition region 40 Thickness of oil phase region 41 Clustering category 42 Characteristic variable 43 Prediction index 44 Data set 45 Sample 46 Training set 47 Test set 48 Input layer 49 Hidden layer 50 Output layer 51 Neuron 52 Weight matrix and threshold matrix 53 Output weight matrix 54 Activation function 55 Solution domain 56 Particle coordinates 57 Particle velocity 58 Particle 59 Pareto optimal solution 60 Particle evolution solution set 61 Dynamic density distance 62 Correction parameter. Specific implementation method:

[0037] The present invention will be further described below with reference to the accompanying drawings:

[0038] As Figure 1As shown, in this bottle test device for intelligently optimizing the high-frequency pulsed electro-dehydration parameters of chemical flooding produced fluid, after the multi-component chemical 1 flows into the chemical injector 3 through the input pipeline 2, the multi-component chemical 1 is sequentially injected into the stirring area 5 through the output pipeline 4, where the chemical component ratio is regulated by the micro flowmeter 7 connected to the intelligent terminal 6. Inside the stirring area 5, there is a rectangular inner rotor 8 and a cylindrical outer rotor 9 connected by rolling bearings, and their reverse stirring is independently controlled by 2 rotating motors 10, so that during the stirring process, the multi-component chemical 1 continuously flows through the circular flow channels 11 arranged in an array on the surface of the rectangular inner rotor 8 and through on both sides. The multi-component chemical 1 is mixed and fully emulsified with each other. Annular gaps are provided at the upper and lower cross-sections of the cylindrical outer rotor 9 to ensure that the multi-component chemical 1 inside and outside the cylindrical outer rotor 9 can be exchanged with each other and evenly emulsified. An annular water bath sandwich 12 is provided on the outermost side of the stirring area 5 and is connected to the water bath temperature control device 13 on the right by a steel pipe. The intelligent terminal 6 controls the water bath temperature through a wire connected to the temperature sensor 14, and pumps the hot water with a certain amount of heat to the inlet above the stirring area 5 to enter the annular water bath sandwich 12. After exchanging heat with the multi-component chemical 1, it returns from the bottom outlet of the annular water bath sandwich 12 to the water bath temperature control device 13, providing temperature conditions for the emulsification process of the multi-component chemical 1, forming a chemical flooding produced fluid sample 15 with a certain chemical component ratio, and then constructing a high-frequency pulsed electro-dehydration emulsification test chamber module for chemical flooding produced fluid.

[0039] The chemically enhanced oil recovery produced liquid sample 15 after sufficient emulsification in the stirring area 5 flows into the bottle test chamber 18 from the steel outlet pipe 16 on the right side of the annular water bath sandwich 12 through the electrode cover 17, and a steel drain pipe 20 is provided at the bottom of the bottle test chamber 18 wrapped by the metal reaction kettle 19, which is convenient for sampling detection of the chemically enhanced oil recovery produced liquid sample 15 and internal cleaning of the bottle test chamber 18; on the left and right sides of the electrode cover 17 above the bottle test chamber 18, a positive electrode conductor 21 and a negative electrode conductor 22 are respectively embedded and extend downward to the rectangular sheet positive electrode 23 and negative electrode 24 in the bottle test chamber 18. At the same time, the intelligent terminal 6 is sequentially connected to the rectifying unit 25, modulating unit 26, and amplifying unit 27 through wires to the right, and the positive and negative terminal interfaces of the amplifying unit 27 are connected to the distributed positive electrode contact piece 28 and negative electrode contact piece 29 on the upper side of the electrode cover 17 by wires, forming a longitudinal high-frequency pulsed electric field in the area directly opposite the positive electrode 23 and negative electrode 24, and a water bath sandwich 30 is provided between the bottle test chamber 18 and the metal reaction kettle 19 and is connected to the water bath temperature control device 13 through a steel pipe, and the metal reaction kettle 19 is connected to the gas cylinder pressurizing device 31 on the left by a small-diameter high-pressure steel pipe to realize high-frequency pulsed electric dehydration of the chemically enhanced oil recovery produced liquid sample 15 under certain operating temperature and certain operating pressure conditions. In addition, a glass viewing window 32 is provided on the front of the metal reaction kettle 19, and a high-speed camera 33 on the right side of the metal reaction kettle 19 is used to record high-frequency pulsed electric dehydration images at different times, and is sequentially connected to an image acquisition system 34 and a data storage system 35 through a data optical cable for image processing and storage, and then a visual demulsification bottle test module for high-frequency pulsed electric dehydration of the chemically enhanced oil recovery produced liquid is constructed.

[0040] The pipes between the chemical injector, stirring area, water bath temperature control device, and bottle test chamber are 316L corrosion-resistant and high-pressure-resistant thin pipes. The pipe joints are connected by thread fitting and sealed with fluoroplastics sealing tape.

[0041] Figure 2It is a schematic diagram of the principle of the method of the present invention, which provides an optimization process for the high-frequency pulsed electro-dehydration parameters of chemical flooding produced fluid. After obtaining the high-frequency pulsed electro-dehydration image at any moment, the pixel value sequence 36 corresponding to any X-axis coordinate is extracted. The red, green, and blue components in the pixel value sequence 36 are converted into a gray value sequence 37 by the weighted average method, and the gray value sequence 37 is divided into three clustering categories 41 representing the thickness 38 of the aqueous phase region, the thickness 39 of the oil-water transition region, and the thickness 40 of the oil phase region according to the Kmeans algorithm. Furthermore, with the three chemical component concentrations, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time as 10 characteristic variables 42, and the thickness 38 of the aqueous phase region, the thickness 39 of the oil-water transition region, and the thickness 40 of the oil phase region extracted from the three clustering categories 41 as 3 prediction indicators 43, a high-frequency pulsed electro-dehydration bottle test data set 44 of chemical flooding produced fluid with different chemical component ratios is constructed. At the same time, after integrating and normalizing the samples 45 in the data sets 44 with multiple chemical component ratios, they are divided into a training set 46 and a test set 47 according to the sample 45 ratio of 75% and 25%, and an ELM neural network topology structure including an input layer 48, a hidden layer 49, and an output layer 50 is constructed. Among them, the number of neurons 51 in the input layer 48 is the same as the number of characteristic variables 42 in the data set 44, which is 10. The number of neurons 51 in the hidden layer 49 is the same as the number of samples 45 in the data set 44. The output layer 50 only sets a single neuron 51 to represent a certain prediction indicator 43. When the characteristic variables 42 of a certain sample 45 in the training set 46 enter the input layer 48, a linear transformation is performed on the characteristic variables 42 by randomly generating a weight matrix and a threshold matrix 52. According to the Moore–Penrose generalized inverse matrix theory, an output weight matrix 53 is obtained. Furthermore, in combination with the activation function 54 and the output weight matrix 53, the predicted values of the thickness 38 of the aqueous phase region, the thickness 39 of the oil-water transition region, and the thickness 40 of the oil phase region corresponding to any characteristic variable 42 are respectively obtained, and a high-frequency pulsed electro-dehydration performance prediction model of chemical flooding produced fluid based on the ELM neural network is constructed, and the prediction accuracy of the model is evaluated by the mean square error of the prediction results of the test set 47. In addition, based on the multi-objective particle swarm algorithm, for the solution domain 55 of the 10 characteristic variables 42 in the data set 44, 50 particles 58 with different particle coordinates 56 and particle velocities 57 are randomly generated to form a population. Among them, the particle coordinates 56 represent the values of the 10 characteristic variables 42, and the particle velocity 57 describes the change trend of the characteristic variables 42. Then, the predicted values of the oil region thickness 40 and the oil-water transition region thickness 39 corresponding to each particle 58 are calculated by the prediction model, and the prediction data is saved to the pareto optimal solution set 59 and the particle evolution solution set 60.Meanwhile, after screening the dominance relationship, the Pareto optimal solution set 59 retains the 50 particles 58 corresponding to the characteristic variables 42 with the largest dynamic density distance 61, and extracts the population-optimal characteristic variables and particle-optimal characteristic variables with the largest dynamic density distance 61 from the Pareto optimal solution set 59 and the particle evolutionary solution set 60 respectively, calculates and obtains the correction parameters 62 of each particle 58 within the solution domain 55, and thus continuously corrects and iterates each particle 58 within the population to the maximum iteration number limit, that is, the optimization process of the particle 58 is completed. At this time, the characteristic variables 42 in the Pareto optimal solution set 59 are the high-frequency pulse electro-dehydration process parameters of the chemical flooding produced liquid that meet the optimal dehydration performance.

[0042] An optimization method for the high-frequency pulse electro-dehydration process parameters of chemical flooding produced liquid using this intelligent bottle test device for optimizing the high-frequency pulse electro-dehydration parameters of chemical flooding produced liquid:

[0043] (1) Since the chemical injection components are injected into the formation in a segmented slug manner and will return out to varying degrees with the produced liquid, there will be multiple chemical components 1 such as polymers, surfactants, and alkalis in the produced liquid. First, the multiple chemical components 1 such as oil, water, polymer, surfactant, and alkali are respectively injected into the corresponding chemical injectors 3 through the input pipeline 2. The outlet valve of the chemical component injector 3 is controlled by the intelligent terminal 6, and the multiple chemical components 1 are sequentially input into the stirring domain 5 in a certain chemical component ratio through the output pipeline 4. A micro flowmeter 7 is set between the input pipeline 2 and the intelligent terminal 6 to accurately control the total input amount of the multiple chemical components 1.

[0044] The main body of the stirring domain 5 can be divided into a three-layer coaxial cylinder structure. The rectangular inner rotor and the cylindrical outer rotor are connected by rolling bearings. The inner diameter of the circular channels on the surface of the rectangular inner rotor is 5 mm. The width of the annular gap between the upper and lower circular cross-sections of the cylindrical outer rotor is 10 mm. The interval between the rectangular inner rotor and the cylindrical outer rotor is 30 mm. The rectangular inner rotor 8 connected to the lower side of the rotating motor 10 is located at the center of the cylinder. A number of circular flow channels 11 that penetrate both sides are drilled on the surface of the rectangular inner rotor 8, so that when the rectangular inner rotor 8 rotates, the multi-component chemical substances 1 are mixed and emulsified with each other after flowing through the circular flow channels 11. The cylindrical outer rotor 9 located outside the rectangular inner rotor 8 is connected upward to another rotating motor 10. A number of annular gaps are provided in the upper and lower circular cross-sections of the cylindrical outer rotor 9 to ensure that the multi-component chemical substances 1 outside the cylindrical outer rotor 9 can flow into the inner side, and then through the reverse rotation of the rectangular inner rotor 8 and the cylindrical outer rotor 9, the multi-component chemical substances 1 are mixed evenly to form a chemical flooding produced liquid sample 15 with a certain emulsified droplet particle size. In addition, considering that the temperature conditions will affect the stirring and emulsification process of the chemical flooding produced liquid sample 15, an annular water bath sandwich layer 12 is provided outside the main body of the stirring domain 5. Hot water at a certain temperature is pumped to the upper end inlet of the annular water bath sandwich layer 12 through the water bath temperature control device 13. After continuously exchanging heat with the chemical flooding produced liquid sample 15 in the stirring domain 5, it flows through the bottom outlet of the annular water bath sandwich layer 12 and returns to the water bath temperature control device 13. Moreover, the temperature sensor 14 in the water bath temperature control device 13 is connected to the intelligent terminal 6, and its water bath temperature is controlled by the set temperature of the intelligent terminal 6.

[0045] Thus, the structural design of the high-frequency pulsed electric dehydration and emulsification test chamber module for chemical flooding produced liquid is completed.

[0046] (2) After forming the chemical flooding produced liquid sample 15 with a certain chemical component ratio, it flows into the bottle test chamber 18 directly below the electrode cover 17 through the outlet pipe 16 on the right side of the stirring domain 5. Among them, the 4 bottle test chambers 18 fixed in the metal reaction kettle 19 constitute the loading space for the chemical flooding produced liquid sample 15. There are emptying ports at the bottom of each bottle test chamber 18, which are connected to the emptying pipe 20. The positive electrode conductor 21 and the negative electrode conductor 22 are respectively embedded on the left and right sides of the electrode cover 17 and extend downward to connect to the rectangular sheet positive electrode 23 and the negative electrode 24 placed opposite to each other.

[0047] Thus, the intelligent terminal 6 is sequentially connected to the rectification unit 25, the modulation unit 26, and the amplification unit 27 to form a high-frequency pulsed electrical signal with a certain frequency, a certain voltage, and a certain duty cycle. The positive and negative poles of the output end of the amplification unit 27 are respectively connected to the positive electrode contact piece 28 and the negative electrode contact piece 29 on the upper side of the electrode cover 17 by wires. Thus, a high-frequency pulsed electric field is formed in the area where the positive electrode 23 and the negative electrode 24 in the bottle test chamber 18 face each other. In addition, the operating pressure of each bottle test chamber 18 is controlled by the gas cylinder pressurizing device 31, and a water bath sandwich layer 30 is arranged between the bottle test chamber 18 and the metal reaction kettle 19. The water bath temperature control device 13 and the high-frequency pulsed electric dehydration and emulsification test chamber module share each other. At the same time, considering the visual reproduction of the demulsification mechanism of applying alternating loads by the pulsed electric field, a glass viewing window 32 is arranged on the front of the metal reaction kettle 19. The upper side of the glass viewing window 32 corresponding to each bottle test chamber 18 extends 30 mm above the positive electrode 23, and the lower side extends to the bottom of the bottle test chamber 18. A high-speed camera 33 is connected to the image acquisition system 34 to record the image information of the oil-water interface at any time in real time, and the information is transmitted to the data storage system 35 to realize the visual recording and data storage of the evolution process of the oil-water interface of the chemical flooding produced liquid sample 15 under the action of the high-frequency pulsed electric field. The input voltage of the rectification unit is 220V, the output pulsed electric signal voltage of the amplification unit is continuously adjustable from 0 to 6000V, and the output current is continuously adjustable from 0 to 100mA. The pixel of the high-speed camera is 1024×1024, and the shooting frequency is 16,000 frames per second.

[0048] The size specification of the bottle test chamber is 30mm×12mm×10mm. The positive electrode is 40mm away from the top of the bottle test chamber, and the negative electrode is 40mm away from the bottom of the bottle test chamber. The width of the annular water bath sandwich layer outside the stirring area and the water bath sandwich layer outside the bottle test chamber is both 10mm. The space between the metal reaction kettle and the bottle test chamber is filled with melamine foam heat insulation material, and the aluminum outer shell of the bottle test chamber is connected to the grounding line to avoid mutual interference of the high-frequency pulsed electric fields between the bottle test chambers.

[0049] Thus, the structural design of the high-frequency pulsed electric dehydration and visual demulsification bottle test module for chemical flooding produced liquid is completed.

[0050] (3) Generate a chemical flooding produced liquid sample 15 with a certain chemical component ratio according to step (1), and flow it into the bottle test chamber 18 in step (2) through the outlet pipe 16 on the right side of the stirring area 5. The intelligent terminal 6 records the concentration and water content data of the set multi-component chemical component 1. At the same time, according to the set operating pressure and operating temperature, the intelligent terminal 6 automatically turns on the gas cylinder pressurizing device 31 and the water bath temperature control device 13 to control the operating temperature and operating pressure of the metal reaction kettle 19. After the temperature and pressure in the bottle test chamber 18 are stable, the intelligent terminal 6 forms a high-frequency pulsed electric field with a certain frequency, a certain voltage, and a certain duty cycle between the positive electrode 23 and the negative electrode 24.

[0051] During the high-frequency pulsed electro-dehydration process, the high-speed camera 33 outputs image information to the image acquisition system 34 at a frequency of 8000 frames per second and stores it in the data storage system 35 in chronological order. The intelligent terminal 6 scans the X-axis and Y-axis with a step size of 0.5 mm, sequentially reads the pixel value sequence 36 at the coordinate (x, y) in the image at a certain moment, and performs gray-scale processing on the image based on the weighted average method as follows:

[0052] Gr t Gr(x,y) = 0.299R(x,y)+0.587G(x,y)+0.114B(x,y) (1) t Gr(x,y) = 0.299R(x,y)+0.587G(x,y)+0.114B(x,y) (1) t Gr(x,y) = 0.299R(x,y)+0.587G(x,y)+0.114B(x,y) (1) t Gr(x,y) = 0.299R(x,y)+0.587G(x,y)+0.114B(x,y) (1)

[0053] In the formula, Gr(x,y) is the gray value at the coordinate (x, y) in the image at time t; R(x,y) is the red component of the pixel value at the coordinate (x, y) in the image at time t; G(x,y) is the green component of the pixel value at the coordinate (x, y) in the image at time t; B(x,y) is the blue component of the pixel value at the coordinate (x, y) in the image at time t. t Gr(x,y) is the gray value at the coordinate (x, y) in the image at time t; R(x,y) is the red component of the pixel value at the coordinate (x, y) in the image at time t; G(x,y) is the green component of the pixel value at the coordinate (x, y) in the image at time t; B(x,y) is the blue component of the pixel value at the coordinate (x, y) in the image at time t. t Gr(x,y) is the gray value at the coordinate (x, y) in the image at time t; R(x,y) is the red component of the pixel value at the coordinate (x, y) in the image at time t; G(x,y) is the green component of the pixel value at the coordinate (x, y) in the image at time t; B(x,y) is the blue component of the pixel value at the coordinate (x, y) in the image at time t. t Gr(x,y) is the gray value at the coordinate (x, y) in the image at time t; R(x,y) is the red component of the pixel value at the coordinate (x, y) in the image at time t; G(x,y) is the green component of the pixel value at the coordinate (x, y) in the image at time t; B(x,y) is the blue component of the pixel value at the coordinate (x, y) in the image at time t. t Gr(x,y) is the gray value at the coordinate (x, y) in the image at time t; R(x,y) is the red component of the pixel value at the coordinate (x, y) in the image at time t; G(x,y) is the green component of the pixel value at the coordinate (x, y) in the image at time t; B(x,y) is the blue component of the pixel value at the coordinate (x, y) in the image at time t.

[0054] Furthermore, according to the gray value sequence 37 corresponding to a certain X-axis coordinate x, considering that after the chemical flooding produced liquid sample 15 undergoes high-frequency pulsed electro-dehydration, an aqueous phase region, an oil-water transition region, and an oil phase region will be formed from bottom to top, and their color distribution gradually becomes darker, and color mutation will occur at the boundary of adjacent regions. Then, the high-frequency pulsed electro-dehydration performance can be quantitatively characterized by the thickness 38 of the aqueous phase region, the thickness 39 of the oil-water transition region, and the thickness 40 of the oil phase region. Taking a certain X-axis coordinate x as a one-dimensional feature and the gray value sequence 37 obtained by scanning the Y-axis with a step size of 0.5 mm as the clustering sample values, the K-means algorithm is used to divide 3 clustering categories 41. Among them, 3 clustering sample values are randomly selected as the initial clustering centers, and the difference between different clustering samples and the clustering centers is characterized by the Manhattan distance as follows:

[0055]

[0056] In the formula, is the Manhattan distance between the i-th clustering sample value and the m-th clustering center in the gray value sequence at time t; P is the m-th clustering center corresponding to the gray value sequence at time t; Gr m,t is the gray value sequence at time t corresponding to the m-th clustering center; Gr t Gr(x,y i ) is the gray value sequence at time t The i-th sample value.

[0057] After obtaining the Manhattan distances between all clustering samples and the clustering centers in the gray value sequence 37 according to Equation (2), the clustering samples are assigned to the clustering category 41 with the smallest Manhattan distance, and the mean values of all clustering samples in each clustering category 41 are calculated. If the accuracy condition is met, it indicates that the iteration of the clustering centers is completed; otherwise, after replacing the clustering centers with the mean values of the clustering samples in each clustering category 41, the Manhattan distances in Equation (2) are recalculated and the clustering centers are iterated until the accuracy condition is met. After the iteration of the clustering centers is completed, the clustering category 41 with the smallest clustering center is identified as the water phase region clustering, the clustering category 41 with the largest clustering center is identified as the oil phase region clustering, and the rest are the oil-water transition region clustering. The intelligent terminal 6 respectively records the maximum values of the clustering samples in the water region clustering and the oil-water transition region clustering, so as to obtain the water phase region thickness 38, the oil-water transition region thickness 39, and the oil phase region thickness 40 at the X-axis coordinate x in the image at time t as follows:

[0058]

[0059] In the formula, are respectively the water phase region thickness, the oil phase region thickness, and the oil-water transition region thickness at the X-axis coordinate x in the image at time t, m; is the maximum value of the water phase region clustering samples at the X-axis coordinate x in the image at time t, m; is the maximum value of the oil-water transition region clustering samples at the X-axis coordinate x in the image at time t, m; h is the total height of the chemical flooding produced liquid sample in the bottle test chamber, m.

[0060] Meanwhile, the X-axis of the image at time t is scanned with a step size of 0.5 mm, and the water phase region thickness 38, the oil-water transition region thickness 39, and the oil phase region thickness 40 corresponding to different X-axis coordinates are sequentially obtained. The mean values of the thicknesses of each region are respectively taken as the oil-water interface distribution state of the image at time t. Then, taking the water phase region thickness 38, the oil-water transition region thickness 39, and the oil phase region thickness 40 as the prediction indexes 43, and selecting the polymer concentration, surfactant concentration, alkali concentration, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time as the characteristic variables 42, a chemical flooding produced liquid high-frequency pulsed electric dehydration bottle test data set 44 for a certain chemical component ratio is constructed.

[0061] The accuracy requirement of the K-means algorithm is that the difference between the mean value of the clustering samples and the clustering center is less than 10 -6 , the weight matrix and threshold matrix of the hidden layer, the initial particle coordinates and particle velocities, and the random numbers r1 and r2 in the particle correction parameters all follow a normal distribution, and the maximum value w max of the inertia coefficient is 0.9, and the minimum value w min of the inertia coefficient is 0.4, and the acceleration factors c1 and c2 are 1.5.

[0062] Thus, the construction of the high-frequency pulsed electric dehydration bottle test data set for chemical flooding produced liquid is completed.

[0063] Repeating this step, a high-frequency pulsed electric dehydration bottle test data set for chemical flooding produced liquid with another chemical component ratio can be constructed.

[0064] (4) For the high-frequency pulsed electric dehydration bottle test data sets 44 of chemical flooding produced liquid with different chemical component ratios in step (3), after integrating and normalizing the data sets 44 with different chemical component ratios in the 4 bottle test cabins 18, they are divided into a training set 46 and a test set 47 according to the ratio of 75% and 25%. The normalization of the characteristic variables and prediction indicators adopts the Min-Max normalization method. Considering that the time series nature of the oil-water interface state monitoring will result in a large number of samples 45 in the data set 44, the ELM neural network algorithm with few training parameters and high learning efficiency is used for predicting the high-frequency pulsed electric dehydration performance of chemical flooding produced liquid. Among them, the ELM neural network topology is the same as that of the single hidden layer feedforward neural network (SLFN), and both are composed of an input layer 48, a hidden layer 49, and an output layer 50. For predicting the thickness 38 of the water phase region in the high-frequency pulsed electric dehydration process of chemical flooding produced liquid, the number of neurons 51 in the input layer 48 is set to be the same as the number of characteristics of the training set 46, which is 10, the number of neurons 51 in the hidden layer 49 is set to be the same as the number of samples 45 in the training set 46, which is m, and only a single neuron 51 is set in the output layer 50 to represent the prediction indicator 43.

[0065] Furthermore, the characteristic variable 42 of the i-th sample 45 in the training set 46 is transmitted from the input layer 48 to the hidden layer 49. After the neurons 51 in the hidden layer 49 randomly generate the weight matrix and threshold matrix 52, a linear transformation is performed on the characteristic variable 42 as follows:

[0066]

[0067] where h ij is the linear output value when the data of the i-th sample in the training set enters the j-th neuron in the hidden layer; ω kj is the corresponding value of the k-th input layer characteristic and the j-th hidden layer neuron in the hidden layer weight matrix; x ik is the corresponding value of the k-th characteristic in the characteristic variable of the i-th sample; θ j is the corresponding value of the j-th hidden layer neuron in the hidden layer threshold matrix.

[0068] Since there is a complex non-linear mapping relationship between the characteristic variable 42 and the prediction indicator 43 in the high-frequency pulsed electric dehydration process of chemical flooding produced liquid, a Sigmoid activation function 54 is set for the linear output value of the hidden layer 49, and there is:

[0069]

[0070] In the formula, H ij is the output value when the feature variable of the i-th sample in the training set enters the j-th neuron in the hidden layer.

[0071] According to the output weight matrix 53 of the hidden layer 49, the predicted value of the water zone thickness 38 corresponding to the i-th sample 45 in the training set 46 is obtained as:

[0072]

[0073] In the formula, β j is the corresponding value of the j-th hidden layer neuron in the output weight matrix.

[0074] Meanwhile, according to formulas (4) to (6), the water phase zone thickness 38 of the m groups of samples 45 in the training set 46 is predicted in sequence, and combined with the true value of the water zone thickness 38 recorded by the intelligent terminal 6 in the training set 46, the loss function is represented in matrix form as:

[0075] loss = ||P - Hβ|| (7)

[0076] In the formula, P is the target matrix composed of the true values of the water phase zone thickness of each sample in the training set; H is the output matrix of the hidden layer; β is the output weight matrix.

[0077] Thus, according to the Moore–Penrose generalized inverse matrix theory, the output weight matrix 53 when the loss function in formula (7) reaches the global minimum is obtained as:

[0078] β = (H T H) -1 H T -P (8)

[0079] In addition, by obtaining the predicted value of the water phase zone thickness 38 of the samples 45 in the test set 47, the prediction accuracy of this model is characterized by the mean square error as:

[0080]

[0081] In the formula, δ j is the predicted value of the water phase zone thickness of the i-th sample in the test set; b is the number of samples in the test set.

[0082] Similarly, based on the prediction indexes 43 of the oil-water transition zone thickness 39 and the oil phase zone thickness 40 in the training set 46 and the test set 47, the changes of the oil-water transition zone thickness 39 and the oil phase zone thickness 40 in the high-frequency pulse electro-dehydration process of chemical flooding produced liquid can be predicted according to formulas (4) to (8), and the corresponding prediction accuracy is evaluated by the mean square error in formula (9).

[0083] Thus, the construction of the prediction model for the high-frequency pulsed electro-dehydration performance of chemical flooding produced fluids based on the ELM neural network is completed.

[0084] Repeating this step, the prediction model for the high-frequency pulsed electro-dehydration performance of chemical flooding produced fluids in another dataset can be constructed.

[0085] (V) During the high-frequency pulsed electro-dehydration of chemical flooding produced fluids, the thickness of the oil phase region 40 quantitatively characterizes the oil-water separation effect, and the thickness of the oil-water transition region 39 reflects the adaptability of the high-frequency pulsed electro-dehydration operating parameters to the demulsification and coalescence of emulsion droplets. Therefore, taking the thickness of the oil phase region 40 and the thickness of the oil-water transition region 39 as the objective functions, and the value range of the characteristic variables 42 in the dataset 44 as the solution domain 55, a multi-objective nonlinear optimization model is constructed:

[0086]

[0087] In the formula, ELM o (X) is the predicted value of the thickness of the oil phase region corresponding to a certain characteristic variable; ELM m (X) is the predicted value of the thickness of the oil-water transition region corresponding to a certain characteristic variable; X max is the maximum value of the characteristic variable; X min is the minimum value of the characteristic variable.

[0088] Based on the multi-objective particle swarm optimization algorithm (MOPSO), 50 particle coordinates 56 and particle velocities 57 representing the characteristic variables 42 are randomly generated within the solution domain 55. Combining with the prediction model for the high-frequency pulsed electro-dehydration performance of chemical flooding produced fluids based on the ELM neural network in step (IV), the predicted values of the thickness of the oil phase region 40 and the thickness of the oil-water transition region 39 corresponding to each particle 58 in the t-th generation population are obtained respectively, and are transmitted to the pareto optimal solution set 59 and the particle evolution solution set 60 of the intelligent terminal 6. Among them, all the iterative data of different particles 58 are recorded in the t-th generation particle evolution solution set 60. And in the t-th generation pareto optimal solution set 59, if the thickness of the oil phase region 40 and the thickness of the oil-water transition region 39 of the i-th particle 58 satisfy the dominance relationship of formula (11), the information of this particle 58 will be removed from the pareto optimal solution set 59.

[0089]

[0090] In the formula, is the characteristic variable corresponding to the i-th particle in the t-th generation population; is the characteristic variable corresponding to the j-th particle in the t-th generation pareto optimal solution set; Ω is the solution domain of the characteristic variable.

[0091] In addition, to avoid repeated retrieval of the optimal operating parameters of high-frequency pulsed electro-dehydration by each particle 58 in the population, the dynamic density distance 61 is used as the fitness index, and the optimization ability of each particle 58 in the t-th generation pareto optimal solution set 59 in the intelligent terminal 6 is evaluated according to Equation (12). The characteristic variables 42 corresponding to the 50 particles 58 with the largest dynamic density distance 61 are selected as the optimized t-th generation pareto optimal solution set 59.

[0092]

[0093] In the formula, are the characteristic variables corresponding to the 2 particles closest to in the t-th generation pareto optimal solution set.

[0094] Meanwhile, with the maximum dynamic density distance 61 as the optimization criterion, the t-th generation population optimal characteristic variable and the i-th particle optimal characteristic variable are respectively extracted from the optimized t-th generation pareto optimal solution set 59 and the t-th generation particle evolutionary solution set 60, so as to obtain the correction parameter 62 of the i-th particle 58 in the t + 1-th generation population as:

[0095]

[0096] In the formula, is the inertia coefficient of the i-th particle in the t + 1-th generation population; w max is the maximum value of the inertia coefficient set by the intelligent terminal; w min is the minimum value of the inertia coefficient set by the intelligent terminal; Q t is the t-th generation population optimal characteristic variable; is the d-th characteristic value of the t-th generation population optimal characteristic variable; is the t-th generation population, the i-th particle optimal characteristic variable; is the d-th characteristic value of the t-th generation population, the i-th particle corresponding characteristic variable; is the particle velocity of the t-th generation population, the i-th particle; is the corrected particle velocity of the t + 1-th generation population, the i-th particle; is the corrected particle coordinate of the t + 1-th generation population, the i-th particle, that is, the characteristic variable; c1, c2 are acceleration factors; r1, r2 are random numbers within the range of (0, 1).

[0097] Therefore, by repeating the iterative steps of formulas (11) to (12), the Pareto optimal solution set 59 and the particle evolutionary solution set 60 are continuously optimized within the operating parameters of high-frequency pulsed electro-dehydration, and the correction parameter 62 is obtained by iterating each particle 58 in the population through formula (13). When the number of iterations reaches the maximum iteration limit, the iterative optimization process of particle 58 is completed. At this time, the characteristic variable 42 in the Pareto optimal solution set 59 is the high-frequency pulsed electro-dehydration process parameter of the chemical flooding produced liquid that satisfies the optimal dehydration performance.

[0098] Thus, the optimization design of the high-frequency pulsed electro-dehydration process parameters for the chemical flooding produced liquid is completed.

[0099] By repeating steps (III) and (IV), a high-frequency pulsed electro-dehydration bottle test data set for the chemical flooding produced liquid of another high-frequency pulsed electro-dehydration process can be formed, and a corresponding high-frequency pulsed electro-dehydration performance prediction model (including the prediction of the thickness of the aqueous phase region, the thickness of the oil-water transition region, and the thickness of the oil phase region) can be constructed. Combining with step (V), a multi-objective nonlinear optimization model applicable to the prediction results of the high-frequency pulsed electro-dehydration performance of another case can be established, further realizing the optimization design of the high-frequency pulsed electro-dehydration process parameters for the chemical flooding produced liquid.

[0100] The present invention integrates the emulsification test chamber module and the demulsification bottle test module, introduces a glass viewing window, a high-speed camera, and an image acquisition system, and visually reproduces the demulsification mechanism of the pulsed electric field applying an alternating load to the oil-water interface; starting from the correlation between the image pixel value and the gray value, based on the K-means clustering algorithm and the ELM neural network algorithm, a high-frequency pulsed electro-dehydration bottle test data set for the chemical flooding produced liquid and an electro-dehydration performance prediction model are constructed, which is also the key to optimizing the high-frequency pulsed electro-dehydration process parameters; then, the dynamic density distance is introduced to characterize the particle swarm optimization performance, and combined with the Pareto optimal solution set and the particle evolutionary solution set, the high-frequency pulsed electro-dehydration process parameters of the chemical flooding produced liquid that satisfy the optimal dehydration performance are determined. Therefore, a reliable optimization device and a scientific method are provided for the optimization design of the high-frequency pulsed electro-dehydration process parameters of the chemical flooding produced liquid with any multi-component chemical composition ratio and any emulsification characteristics, and the visual demulsification mechanism of the pulsed electric field applying an alternating load to the oil-water interface is fully considered, changing the determination of the high-frequency pulsed electro-dehydration process parameters from relying on production operation experience to quantitative optimization combined with the image data of the bottle test chamber. At the same time, a data-driven neural network algorithm is introduced, providing an example for the optimization design and operation management of the gathering and transportation processing system in the context of intelligent oilfield construction, effectively promoting the development of efficient demulsification technologies for complex produced liquids, the research and development, and the popularization and application of new oil-water separation equipment.

[0101] The present invention fully considers the influence of multiple chemical components in chemical flooding produced fluids on the emulsification characteristics, introduces a rectangular inner rotor and a cylindrical outer rotor to form a stirring domain, and adjusts the pulse frequency, pulse width, and pulse voltage through an intelligent terminal, which can break through the fixed sample emulsification characteristics and electric field parameter structure in the traditional high-frequency pulsed electric dehydration bottle test device; at the same time, combined with a glass viewing window, a high-speed camera, and an image acquisition system, it visually reproduces the demulsification mechanism of the pulsed electric field applying an alternating load to the oil-water interface, providing a high-frequency pulsed electric dehydration bottle test device for chemical flooding produced fluids that is intelligently optimized, parameter-adjustable, and process-visible; the method of optimizing the high-frequency pulsed electric dehydration parameters of chemical flooding produced fluids by intelligent algorithms is highly scientific, operable, and practical. By constructing quantitative characterization indexes for the pulsed electric dehydration process, combining intelligent algorithms such as the ELM neural network and multi-objective particle swarm optimization, it efficiently predicts the optimal values of high-frequency pulsed electric dehydration process parameters, providing technical support for the optimal design and operation management of the surface gathering and transportation processing system under the background of the construction of a green and intelligent oilfield.

Claims

1. An intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid, characterized in that: The intelligent device for optimizing the high-frequency pulse electric dehydration parameters of chemical flooding produced fluid includes a test chamber module, a visual demulsification bottle test module, a high-speed camera, an image acquisition system, a data storage system, and an intelligent terminal. The stirring domain in the test chamber module simulates the potential component factors and temperature environment of the emulsification process of chemical flooding produced fluid, and the multiple chemical components are evenly mixed in the stirring domain to prepare a set of chemical flooding produced fluid samples with a certain distribution ratio. The prepared chemical drive produced fluid sample is transported to each bottle test chamber of the visual demulsification bottle test module. An electrode cover is arranged at the upper port of the bottle test chamber. A positive electrode conductor and a negative electrode conductor are respectively embedded in the left and right sides of the electrode cover. The positive electrode conductor and the negative electrode conductor extend downward and are respectively connected to the rectangular sheet-shaped positive electrode and negative electrode placed opposite to each other, so as to form a high-frequency pulse electric field. Each bottle test chamber is fixed in a metal reactor. A water bath interlayer is arranged between the bottle test chamber and the metal reactor. A gas cylinder pressurizing device controls the operating pressure in each bottle test chamber. A glass visual window is arranged in the metal reactor. A high-speed camera is connected to an image acquisition system to record the image information of the oil-water interface at any time in real time, and transmit the information to a data storage system, so as to realize the visual recording and data storage of the evolution process of the oil-water interface of the chemical drive produced fluid sample under the action of a high-frequency pulse electric field. The intelligent terminal performs grayscale processing on the high-frequency pulse electric dehydration image to obtain the oil-water interface distribution state when the X-axis coordinate in the image at time t is x: In the formula, They are the thickness of the water phase region, the thickness of the oil phase region, and the thickness of the oil-water transition region when the X-axis coordinate is x in the image at time t, m; is the maximum value of the clustered samples in the water phase region when the X-axis coordinate in the image at time t is x, m; is the maximum value of clustered samples in the oil-water transition zone when the X-axis coordinate in the image at time t is x, m; h is the total height of the chemical flooding produced fluid sample in the bottle test chamber, m; Then, taking the thickness of the water phase zone, the thickness of the oil phase zone, and the thickness of the oil-water transition zone as prediction indicators, the polymer concentration, surfactant concentration, alkali concentration, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time were selected as characteristic variables to construct a high-frequency pulse electric dehydration bottle test data set for chemical flooding produced fluid with a certain chemical component ratio.

2. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 1, characterized in that: A method for preparing a chemical flooding produced fluid sample with a certain distribution ratio in a test chamber module: different multi-component chemical components are input into chemical injectors connected in parallel in the test chamber module, and the multi-component components include polymers, surfactants, and alkalis. An intelligent terminal controls the outlet valves of the injectors of each chemical component, and sequentially inputs the multi-component chemical components into a stirring domain through an input pipeline according to a certain chemical component distribution ratio. A micro flow meter is arranged between the input pipeline and the intelligent terminal to control the total input amount of the multi-component chemical components. An annular water bath interlayer is arranged outside the main body of the stirring domain to simulate the potential component factors and temperature environment of the emulsification process of the chemical flooding produced fluid. The multi-component chemical components are evenly mixed in the stirring domain to form a chemical flooding produced fluid sample with a certain distribution ratio.

3. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 2 is characterized in that: The stirring domain is a three-layer coaxial cylindrical structure. The rectangular inner rotor connected to the lower side of the rotating motor is located at the center of the cylinder. A plurality of circular flow channels penetrating on both sides are drilled on the surface of the rectangular inner rotor, so that the multiple chemical components flow through the circular flow channels during the rotation of the rectangular inner rotor and mix and emulsify with each other; the cylindrical outer rotor located outside the rectangular inner rotor is upwardly connected to another rotating motor, and a plurality of annular gaps are arranged in the upper and lower circular sections of the cylindrical outer rotor to ensure that the multiple chemical components on the outside of the cylindrical outer rotor can flow into the inside, and then the multiple chemical components are mixed evenly through the reverse rotation of the rectangular inner rotor and the cylindrical outer rotor to prepare a chemical flooding produced liquid sample with a certain emulsified droplet size; the water bath temperature control device pumps hot water of a certain temperature to the upper end inlet of the annular water bath interlayer, and after continuously exchanging heat with the chemical flooding produced liquid sample in the stirring domain, it flows through the bottom outlet of the annular water bath interlayer and flows back to the water bath temperature control device, and the temperature sensor in the water bath temperature control device is connected to the intelligent terminal, and its water bath temperature is controlled by the set temperature of the intelligent terminal.

4. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 3 is characterized in that: Four bottle test chambers are arranged in the metal reactor, and a vent is provided at the bottom of each bottle test chamber, which is connected to the vent pipe; a glass visual window is arranged on the front of the metal reactor, and the upper side of the glass visual window corresponding to each bottle test chamber extends to 30 mm above the positive electrode, and the lower side extends to the bottom of the bottle test chamber; the intelligent terminal is connected to the rectification unit, the modulation unit and the amplification unit in sequence to form a high-frequency pulse electrical signal of a certain frequency, a certain voltage and a certain duty cycle, and the positive and negative electrodes of the output end of the amplification unit are respectively connected to the positive electrode contact piece and the negative electrode contact piece on the upper side of the electrode cover with wires, so as to form a high-frequency pulse electric field in the area where the positive electrode and the negative electrode in the bottle test chamber are facing each other.

5. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 4 is characterized in that: The rectangular inner rotor is connected to the cylindrical outer rotor through a rolling bearing. The inner diameter of the circular channel on the surface of the rectangular inner rotor is 5 mm. The width of the annular gap of the upper and lower circular sections of the cylindrical outer rotor is 10 mm. The interval between the rectangular inner rotor and the cylindrical outer rotor is 30 mm.

6. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 5, characterized in that: The metal reactor and the bottle test chamber are filled with melamine foam insulation material, and the aluminum shells of the bottle test chambers are all connected to the grounding circuit to avoid mutual interference of high-frequency pulse electric fields between the bottle test chambers.

7. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 6, characterized in that: The bottle test chamber has a length of 30 mm, a width of 12 mm, and a height of 10 mm. The positive electrode is 40 mm away from the top of the bottle test chamber, and the negative electrode is 40 mm away from the bottom of the bottle test chamber.

8. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 7, characterized in that: The width of the annular water bath interlayer outside the stirring area and the water bath interlayer outside the bottle test chamber are both 10 mm.

9. The intelligent device for optimizing high-frequency pulse electric dehydration parameters of chemical flooding produced fluid according to claim 8, characterized in that: The input voltage of the rectifier unit is 220V, the output voltage of the pulse electric signal of the amplifying unit is continuously adjustable from 0 to 6000V, and the output current is continuously adjustable from 0 to 100mA.

10. A method for intelligently optimizing the parameters of high-frequency pulse electric dehydration of chemical flooding produced fluid, characterized in that The steps include: Step 1, using the intelligent optimization device for high-frequency pulse electric dehydration parameters of chemical flooding produced fluid to obtain a bottle test data set of high-frequency pulse electric dehydration of chemical flooding produced fluid with different chemical component ratios, wherein the thickness of the water phase region, the thickness of the oil phase region, and the thickness of the oil-water transition region are used as prediction indicators, and the polymer concentration, surfactant concentration, alkali concentration, water content, pulse frequency, pulse voltage, duty cycle, operating temperature, operating pressure, and operating time are used as characteristic variables; Step 2: After integrating and normalizing the data sets of high-frequency pulse electric dehydration bottle test of chemical flooding produced fluid with different chemical component ratios, the training set and the test set are divided into 75% and 25% samples respectively. The ELM neural network algorithm is used to construct a prediction model of high-frequency pulse electric dehydration performance of chemical flooding produced fluid based on the ELM neural network. Step 3: Optimize the process parameters of high-frequency pulse electric dehydration of chemical flooding produced fluid: During the high-frequency pulse dehydration process of chemical flooding produced fluid, the thickness of the oil phase zone quantitatively characterizes the oil-water separation effect of the produced fluid, and the thickness of the oil-water transition zone reflects the adaptability of the high-frequency pulse dehydration operating parameters to the demulsification and aggregation of emulsion droplets. Taking the thickness of the oil phase zone and the thickness of the oil-water transition zone as the objective function and the value range of the characteristic variables in the data set as the solution domain, a multi-objective nonlinear optimization model is constructed: Where, ELM o (X) is the predicted value of the oil phase thickness corresponding to a certain characteristic variable; ELM m (X) is the predicted value of the oil-water transition zone thickness corresponding to a certain characteristic variable; X max is the maximum value of the characteristic variable; X min is the minimum value of the characteristic variable; Based on the multi-objective particle swarm algorithm, in the solution domain [X min ,X max ] randomly generate particle coordinates and particle velocities of 50 particles to form a population, wherein the particle coordinates are the characteristic variables, and combined with the high-frequency pulse electric dehydration performance prediction model for chemical drive produced fluid, the predicted values ​​of the oil phase thickness and the oil-water transition zone thickness corresponding to each particle of the t-th generation population are obtained respectively, and transmitted to the Pareto optimal solution set and the particle evolution solution set of the intelligent terminal for optimization, and the Pareto optimal solution set and the particle evolution solution set are continuously optimized within the range of high-frequency pulse electric dehydration operation parameters. When the number of iterations reaches the maximum number of iterations, the iterative optimization of the particles is completed. At this time, the characteristic variables in the Pareto optimal solution set are the high-frequency pulse electric dehydration process parameters of chemical drive produced fluid that meet the optimal dehydration performance.

Images