Micro-fluidic chip pour point depressant intelligent screening and optimizing device and method integrating multi-parameter detection

Abstract

The invention relates to a micro-fluidic chip pour point depressant intelligent screening and optimizing device and method integrating multi-parameter detection, and the micro-fluidic chip pour point depressant intelligent screening and optimizing device integrating multi-parameter detection comprises an experiment module, a data acquisition module and a data analysis and control module, the experimental module comprises a power area, a conveying area, a temperature control area, an imaging area and a recovery area, the imaging area comprises a polarizing microscope and a micro-fluidic chip, the micro-fluidic chip comprises a sample introduction channel, a storage chamber, a micro-mixing chamber, a wax precipitation chamber and a temperature control system, and the micro-mixing chamber realizes efficient mixing of a pour point depressant and crude oil; the wax precipitation chamber simulates the complex flow state environment of a real oil pipeline, a wax crystal morphology observation area, a wax crystal motion behavior observation area and a wax crystal dynamic behavior observation area are arranged, the wax crystal flow state is observed, and wax crystal dynamic parameters are collected; through dynamic time sequence modeling of LSTM and strategy gradient optimization of PPO, efficient dynamic adjustment of a pour point depressant formula is realized.

Description

Technical Field

[0001] The present invention relates to a technology for preventing pour point depressant in long-distance pipeline transportation of waxy crude oil, specifically a device and method for intelligent screening and optimization of pour point depressants on a microfluidic chip with integrated multi-parameter detection. Background Art

[0002] Oil production and transportation face numerous complex and challenging technical challenges, with crude oil gelation being a particularly prominent issue. As oilfield production enters the middle and late stages, crude oil produced by a large number of wells exhibits high levels of wax and colloid asphaltene. At low temperatures, the wax molecules in this type of crude oil gradually precipitate and cross-link through van der Waals forces, forming a three-dimensional network of crystals. This leads to a sharp deterioration in the crude's rheological properties, an exponential increase in apparent viscosity, a significant increase in yield stress, and ultimately, overall solidification of the crude oil. This phenomenon poses a serious threat to the safety of oil production, pipeline gathering, transportation, storage, and transportation, and can even cause major production accidents such as pipeline blockages and surges in pump pressure.

[0003] Traditional strategies for dealing with crude oil gelation mainly rely on the use of pour point depressants, but their screening methods have significant flaws. The conventional bottle test method requires static mixing of crude oil samples with pour point depressants of different ratios in a constant temperature box, and the flow changes are detected by visual observation or rotational rheometer. This method not only consumes a large amount of crude oil (more than 500 mL is required for a single experiment), but also has an extremely cumbersome operation process: from accurately weighing reagents, controlling the mixing shear rate (usually 10 to 100 s -1 ), to maintaining a constant temperature environment (-20℃ to 50℃±1℃) to wait for the wax crystals to fully precipitate (24 to 72 hours), and finally completing the evaluation through manual interpretation of the cloud point temperature or pour point temperature. The entire cycle often takes 2 to 4 weeks, and the experimental results are significantly affected by the operator's experience (data deviation can reach 15% to 20%). More importantly, this method is completely unable to simulate the dynamic working conditions of the downhole-surface system: in real oil wells, crude oil experiences a temperature drop rate of 0.5-2℃ per minute when rising from deep in the formation (temperature 60-120℃) to the wellhead. When flowing in the pipeline, it is subjected to flow velocity gradients and shear rates, and is also affected by the interface effects of pipe wall roughness and material. The pour point depressants selected through static laboratory tests are often unable to adapt to dynamic shear fields, sudden temperature changes and complex interface environments in actual applications, resulting in a 30% to 50% decrease in pour point depressant efficiency, and even causing secondary problems such as demulsification and asphaltene precipitation. Summary of the Invention

[0004] The purpose of the present invention is to provide a microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection. This microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection is used to solve the problems in the prior art of using laboratory static testing to screen pour point depressants, such as long screening cycles, cumbersome methods, large deviations, and the inability of the screened pour point depressants to adapt to pipeline transportation environments. Another purpose of the present invention is to provide a screening and optimization method for this microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection.

[0005] The present invention solves its technical problems by adopting the following technical solutions: the microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection includes an experimental module, a data acquisition module, and a data analysis and control module. The experimental module includes a power zone, a conveying zone, a temperature control zone, an imaging zone, and a recovery zone. The imaging zone includes a polarizing microscope and a microfluidic chip. The microfluidic chip includes an injection channel, a storage chamber, a micro-mixing chamber, a wax precipitation chamber, and a temperature control system. Multiple pour point depressant samples and a fluid to be treated are simultaneously connected through each injection channel. The inner wall of the micro-mixing chamber is coated with a super-hydrophobic anti-adsorption coating. Combined with a staggered fishbone structure and a cylindrical flow-around design, the fluid shear effect is used to achieve efficient mixing of the pour point depressant and crude oil. The wax precipitation chamber simulates the complex flow environment of a real oil pipeline, and is provided with a wax crystal morphology observation zone, a wax crystal movement behavior observation zone, and a wax crystal dynamic behavior observation zone. A combination of straight channels, annular channels, channels with inclined surfaces, and variable-diameter channels is used to observe the flow state of wax crystals under different environments and collect wax crystal dynamic parameters.

[0006] The data acquisition module collects the pressure of the fluid in the microfluidic chip, obtains the flow rate and shear force parameters of the fluid through the microelectromechanical system, and uses the microscope imaging system and image acquisition to capture images of the wax crystal precipitation process in the fluid in the wax precipitation chamber to obtain the wax crystal volume fraction, wax crystal number density, wax crystal standard deviation skewness and peak value; the data analysis and control module intelligently screens the pour point depressant and dynamically optimizes the pour point depressant ratio strategy based on the long short-term memory network LSTM and PPO reinforcement learning algorithm.

[0007] The above scheme for collecting dynamic parameters of wax crystals specifically includes: collecting morphological parameters of wax crystals, including the size, average size and number of wax crystals under the action of cooling, collecting kinematic behavior parameters of wax crystals, including the linear velocity, angular velocity and acceleration of wax crystals, and collecting dynamic behavior parameters of wax crystals, including the size, number, fractal dimension and roundness of wax crystals under the action of shear force.

[0008] In the above scheme, the power zone includes a constant flow pump and a mechanical device. The constant flow pump provides kinetic energy to enable the medium to enter the microfluidic chip evenly; the delivery zone includes a micro sampler and a delivery hose; the temperature control zone includes a Peltier temperature control table to stabilize the sample in the microfluidic chip at the set temperature range of the experiment; the recovery zone includes a waste liquid recovery hose and a waste liquid recovery barrel; under the action of the power zone, the sample in the micro sampler is transported to the imaging area through the delivery hose and finally flows into the recovery zone.

[0009] In the above scheme, the microfluidic chip is made of polydimethylsiloxane (PDMS) or glass. Its injection channel is set to different size specifications according to experimental requirements, with a width of 100μm-1000μm and a depth of 5μm-500μm. Each injection channel is precisely controlled to open and close by a microfluidic valve, and multiple pour point depressant samples and fluids to be treated are connected at the same time; the micro-mixing chamber is located in the central area of ​​the chip and has a volume of 1μl-100μl.

[0010] In the above scheme, the wax crystal morphology observation area is set up with an ordinary straight channel and a straight channel with baffles. The straight channel with baffles is a straight channel with several baffles set in it, and the distance between two adjacent baffles is different; the straight channel is used to observe the formation process of wax crystals in a natural state, and the straight channel with baffles promotes the aggregation and collision of wax crystals; through a high-resolution microscope imaging system and image analysis algorithm, the size, average size and number of wax crystals are obtained, and the effect of the pour point depressant on the wax crystal morphology is analyzed from a microscopic level.

[0011] In the above scheme, the wax crystal movement behavior observation area includes an annular channel and an inclined surface channel. The inclined surface channel is a channel with inclined surfaces at different angles. The annular channel causes the wax crystals to produce circular motion, which is used to measure linear velocity and angular velocity; the inclined surface channel changes the direction and speed of wax crystal movement to obtain acceleration data; with the help of high-speed cameras and digital image correlation technology, the movement trajectory of wax crystals is tracked, the movement parameters are calculated, and the influence of pour point depressants on the movement state of wax crystals is studied.

[0012] In the above scheme, the observation area for the dynamic behavior of wax crystals uses a variable diameter channel and a strain gauge channel. The strain gauge channel is a variable diameter channel with strain gauges arranged in an internal array. The variable diameter channel changes the fluid velocity and applies different shear forces to the wax crystals. When crude oil containing wax crystals flows through the variable diameter channel, the flow rate of the fluid will change. The change in flow rate causes the fluid to apply different degrees of shear force on the wax crystals. In the contraction section, the flow rate increases and the shear force on the wax crystals increases; in the expansion section, the shear force is relatively reduced. Different pour point depressant concentrations change the internal structure and surface properties of the wax crystals, affecting their ability to resist shear force. The temperature, pressure of the crude oil, and the initial size and morphology of the wax crystals also affect the wax crystal fragmentation process. As the temperature increases, the wax crystals become more flexible and less likely to break. Pressure changes will change the interaction between the wax crystals and crude oil molecules, indirectly affecting the fragmentation process. The strain gauge channel is used to measure the resistance to the movement of the wax crystals. Then, using a microscope imaging system and a high-speed camera, combined with an image processing algorithm, the wax crystal diameter, fractal dimension, number, and roundness are obtained to analyze the effect of the pour point depressant concentration on the wax crystal fragmentation process.

[0013] The above-mentioned microfluidic chip integrated multi-parameter detection intelligent screening and optimization device screening and optimization method of pour point depressants are as follows: the pressure, viscosity and wax crystal dynamic parameters are collected in real time through the microfluidic chip, and the high-dimensional features within the 60-second time window are extracted through the long short-term memory network LSTM, and the evolution law of the wax crystal structure under the high-dimensional feature temperature gradient is constructed to construct a state vector representation system, input the PPO reinforcement learning algorithm, and make decisions in the three-dimensional continuous action space. The strategy optimization is driven by the weighted reward function of pour point depression, viscosity change rate, fractal dimension change rate and cost control. The weight ratio of pour point depression, viscosity change rate, fractal dimension change rate and cost control is 10:6:3:1, forming a closed-loop intelligent control mechanism of time series perception-decision optimization. Through the deep collaboration of the long short-term memory network LSTM and the PPO reinforcement learning algorithm, efficient pour point depressant screening and optimization from time series dynamic modeling to intelligent decision optimization is realized.

[0014] The above-mentioned microfluidic chip integrated multi-parameter detection intelligent screening and optimization device screening and optimization method of pour point depressant are specifically as follows:

[0015] Step 1: Model pre-fitting experiment: Before conducting the microfluidic chip experiment, a small sample conventional freezing point determination experiment is first performed to systematically obtain the freezing point drop ΔT of the oil sample after adding the freezing point depressant. g , establish the basic data of the correlation between freezing point depression and normalized index S;

[0016] In the second step, efficient mixing of pour point depressants is achieved in the microfluidic chip through multi-channel coordinated control: Pour point depressant samples of different concentrations are introduced through injection channels of corresponding inner diameters. After preliminary mixing with crude oil in the storage chamber, they enter the micro-mixing chamber. The flow rate of each injection channel is precisely controlled by the microfluidic valve group at 0.1-5μL / min to ensure that different samples participate in the reaction at a stable flow rate ratio. The microstructure of the micro-mixing chamber induces laminar flow disturbance, allowing the crude oil and pour point depressant to achieve full contact in a short time.

[0017] Step 3, data acquisition: The data acquisition process of the microfluidic chip realizes comprehensive parameter measurement through multi-sensor fusion technology. The temperature control system simulates the crude oil cooling and gelation process at a rate of 0.1℃ / min, and the micro-piezoelectric sensors integrated on both sides monitor the channel pressure fluctuations in real time. The micro-flow rate sensor and shear force sensor manufactured based on MEMS technology synchronously obtain fluid dynamic parameters, and the Hagen-Poiseuille equation is combined to calculate the fluid viscosity change. The microscopic imaging system continuously shoots the dynamics of the wax precipitation chamber at a rate of 60 frames per second, capturing the morphological evolution, motion trajectory and fragmentation behavior of wax crystals during precipitation, collecting the dynamic parameters of wax crystals, giving different weights to the dynamic parameters of wax crystals, and constructing the normalized parameter S. Based on the freezing point measured in step 1, its quantitative relationship with the normalized parameter is determined. After all image data are three-dimensionally reconstructed and voxelized, the wax crystal volume fraction, number density and spatial distribution parameters are extracted by feature statistical algorithm. The wax crystal volume fraction is 0.1%-15%, the number density is 10 2 -10 4 Pieces / mm 3 , spatial distribution parameters: skewness 0.5-1.2, kurtosis 2.8-4.5, providing a complete data set for quantitative analysis of the effect of pour point depressants on the three-dimensional growth characteristics of wax crystals;

[0018] Step 4: Data analysis and formula optimization: Normalize the fluid mechanics parameters and wax crystal time series feature data. The fluid mechanics parameters include pressure, viscosity, and flow rate. The wax crystal time series features include morphological parameters, kinematic parameters, dynamic parameters, and voxelized parameters. Arrange the sensor data by time to form sequence data of time steps. Use 60 time cloths as the input units of LSTM. Next, extract the time series features of LSTM. Train the wax crystal evolution data with different pour point depressant ratios by designing 128-dimensional vectors of the input layer and output layer. The input layer includes pressure, viscosity, wax crystal size, fractal dimension, and roundness. The time series feature vectors output from TM are imported into the PPO algorithm to construct the state space. The pour point depression, viscosity reduction rate, wax crystal fractal dimension change rate and cost penalty are selected for reward function design. The pour point depressant concentration optimization scheme and reward value are obtained by calculation. The relevant parameters are then input into the long short-term memory network LSTM to construct a pour point depressant effect evaluation model. Finally, the PPO algorithm is used to perform formula optimization search. The pour point depressant content is used as the gene code. New formula combinations are continuously generated through selection, crossover and mutation. The fitness of each combination is calculated according to the evaluation model. After multiple generations of evolutionary iterations, it gradually converges to the optimal or near-optimal pour point depressant formula.

[0019] The normalized index data in step 4 of the above scheme is used to determine the quantitative relationship between the normalized parameter and the freezing point through experimental measurement. The specific formula is as follows:

[0020] S=a×(a i D avg +a j N+a k σ)+b×(b i V+b j A+b k ω)+c×(c i D+c j F+c k N ad +c m C)

[0021] ΔT g =AS p +B

[0022] Where: a+b+c=1a i +a j +a k =1b i +b j +b k =1c i +c j +c k +c l =1

[0023] Where: S is the normalized index, where a n 、b n 、c n is the weight of each parameter; ΔT g is the freezing point depression, ℃, where A, B, and P are obtained by fitting experimental data; D avg is the average size of wax crystals, μm; N is the number of wax crystals per unit volume; σ is the dispersion of wax crystal size; V is the linear velocity of wax crystals, mm / s; A is the acceleration of wax crystals, mm / s 2 ; ω is the angular velocity of the wax crystal, rad / s; D is the wax crystal crushing diameter, μm; F is the fractal dimension of the wax crystal; N ad is the increase in the number of wax crystals; C is the attenuation of the roundness of the wax crystals.

[0024] Beneficial effects:

[0025] 1. This invention achieves efficient dynamic adjustment of pour point depressant formulations through LSTM dynamic time series modeling and PPO policy gradient optimization. Its core advantages are: LSTM captures the dynamic evolution of wax crystal behavior, addressing the "time blind spot" of traditional static experiments. It can accurately predict the optimal time point for crude oil to react with pour point depressants, reducing time costs; the PPO algorithm utilizes these patterns to adjust the formulation in real time, avoiding the policy lag caused by traditional methods that ignore the time dimension.

[0026] 2. The present invention is a device and method for rapid screening of pour point depressants and intelligent formula optimization of microfluidic chips with integrated multi-parameter detection. First, it demonstrates an experimental device for observing wax crystal growth under the action of pour point depressants and cooling based on a microfluidic and microscopic system, as well as the specific structure of the microfluidic chip; second, it demonstrates a characteristic statistical method based on wax crystal morphology, kinematics, and dynamic behavior parameters and voxels, which serves as relevant data for a pour point depressant effect evaluation model constructed with a long-short-term memory network and a PPO algorithm, thereby achieving the purpose of screening and optimizing pour point depressants with different contents.

[0027] 3. The present invention conducts research on the deterioration of fluidity during the transportation of waxy crude oil. During long-distance pipeline transportation, the drop in ambient temperature causes wax crystals in the crude oil to precipitate, aggregate, and form a three-dimensional network structure, significantly reducing fluid fluidity. To solve this problem, the present invention has developed a new pour point depressant screening and optimization system based on a microfluidic chip. By simulating the actual working conditions of the pipeline, the control effect of pour point depressants of different concentrations on the microstructure of wax crystals is quantitatively analyzed. The microfluidic chip and its supporting method can accurately evaluate the improvement effect of pour point depressants on key parameters such as the pour point and viscosity of crude oil, providing a theoretical basis and technical support for optimizing the flow safety assurance plan for waxy crude oil.

[0028] 4. The present invention uses a voxel-based characteristic statistical method to quantitatively calculate the wax crystal volume fraction, wax crystal number density, wax crystal standard deviation skewness and peak value, and analyze the distribution characteristics of wax crystals in three-dimensional space, providing an important basis for studying the formation and growth of wax crystals during the cooling process of crude oil under the action of different concentrations of pour point depressants. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is the technical roadmap for the intelligent screening and optimization device and method of pour point depressants on a microfluidic chip with integrated multi-parameter detection.

[0030] Figure 2 This is a schematic diagram of the microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection of the present invention.

[0031] Figure 3 Schematic diagram of the microfluidic chip of the present invention.

[0032] In the figure: 1 micro-mixing chamber, 2 morphological observation area, 3 kinematic and dynamic observation area, 4 variable diameter channel, 5 injection channel, 6 injection port, 7 recovery port, 8 microfluidic chip, 9 constant flow pump, 10 polarizing microscope, 11 waste liquid recovery barrel. DETAILED DESCRIPTION

[0033] The present invention will be further described below with reference to the accompanying drawings:

[0034] See Figure 1-Figure 3 This microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection includes an experimental module, a data acquisition module, and a data analysis and control module. The modules work together to form a complete experimental closed loop.

[0035] (1) The experimental module includes a power zone, a delivery zone, a temperature control zone, an imaging zone and a recovery zone. The power zone includes a constant flow pump 9 and a mechanical device. The constant flow pump provides kinetic energy so that the medium can evenly enter the interior of the microfluidic chip. The delivery zone includes a micro-sampler and a delivery hose. The main equipment in the temperature control zone is a Peltier temperature control table to ensure that the sample in the microfluidic chip is stable in the experimental set temperature range. The main equipment in the imaging zone includes a polarizing microscope 10 and a microfluidic chip 8. The recovery zone includes a waste liquid recovery hose and a waste liquid recovery barrel 11. The sample flows in from the injection port 6. Under the action of the power zone, the sample in the micro-sampler is transported to the imaging zone through the delivery hose, flows out from the recovery port 7, and finally flows into the recovery zone. The core of the experimental module is a customized microfluidic chip made of polydimethylsiloxane (PDMS) or glass, taking into account both chemical stability and biocompatibility. The internal design of the chip includes multi-level functional units including an injection channel 5, a storage chamber, a micro-mixing chamber 1, a wax precipitation chamber and a temperature control system. The sample inlet channel 5 can be configured to various sizes depending on experimental requirements, with widths ranging from 100μm to 1000μm and depths from 5μm to 500μm. Each inlet channel is precisely controlled by microfluidic valves, allowing simultaneous access to multiple pour point depressant samples and the fluid being processed. The micromixing chamber, located in the center of the chip, has a volume of 1μl to 100μl. Its inner wall is coated with a superhydrophobic, anti-adsorption coating. Combined with a staggered herringbone structure and cylindrical flow-around design, it utilizes fluid shear effects to achieve efficient mixing of pour point depressant and crude oil. The wax precipitation chamber, a key component of the microfluidic chip's pour point depressant screening and formulation optimization device, serves as an important platform for studying the properties of wax crystals under various conditions. It primarily simulates the complex flow conditions of a real oil pipeline and features three observation areas. A combination of different channels—a straight channel, an annular channel, a channel with an inclined surface, and a variable-diameter channel—is used to observe the flow behavior of wax crystals under different environments. In addition, the system can also collect morphological parameters such as size, average size, and number of wax crystals under cooling; kinematic parameters such as linear velocity, angular velocity, and acceleration; and dynamic parameters such as size, number, fractal dimension, and roundness of wax crystals under shear stress. The wax analysis chamber comprehensively evaluates the effectiveness of the pour point depressant through multi-indicator observation. The temperature control system integrates a microscopic temperature control device and high-precision temperature sensors to precisely control the internal temperature of the chip.

[0036] (2) Data acquisition module, which is used to collect the pressure of the fluid in the microfluidic chip, obtain the flow rate and shear force parameters of the fluid through microelectromechanical system (MEMS) technology, and use the microscope imaging system and image acquisition to capture images of the wax crystal precipitation process in the fluid in the wax precipitation chamber. The voxel-based feature statistical method is used to obtain data such as the wax crystal volume fraction, wax crystal number density, wax crystal standard deviation skewness and peak value.

[0037] (3) Data analysis and control module, which has a built-in intelligent algorithm model, inputs fluid mechanics parameters such as pressure and viscosity and wax crystal temporal characteristics into the long short-term memory network (LSTM) to extract high-order temporal representations; then, based on algorithms such as PPO reinforcement learning, the representation is used as the state space to dynamically optimize the pour point depressant ratio strategy.

[0038] The microfluidic chip and its accessories enable observation of the morphological, kinematic, and dynamic behaviors of wax crystals. In the morphological observation module, a combination of straight and annular channels simulates the laminar and turbulent flow of crude oil in a pipeline.

[0039] (1) Morphological observation can observe the formation, aggregation, collision and other behaviors of wax crystals during movement, among which the main observations are the size, average size, number, size and other data of wax crystals. When waxy crude oil flows through the observation area, a high-resolution high-speed camera combined with fluorescent labeling technology can clearly capture the dynamic process of wax crystals from nucleation to aggregation. The image processing system uses an edge detection algorithm to automatically identify the outline of wax crystals, accurately calculates the long and short axis dimensions of wax crystals through pixel calibration, and counts the number of wax crystals per unit volume based on connected domain analysis. The system specially sets the curvature radius gradient of the annular channel to study the influence of bending effect on the morphological evolution of wax crystals.

[0040] (2) Kinematic behavior analysis was carried out in observation area 2, which consists of three groups of functional channels: the linear velocity of the wax crystal in the straight channel is directly reflected by calculating the displacement of the wax crystal in the flow direction per unit time, which directly reflects the speed characteristics of the translational motion of the wax crystal in the microfluidic channel; the 30° inclined channel forms a non-uniform force field through geometric structure design. The acceleration generated when the wax crystal flows through can be obtained by capturing the displacement difference between consecutive frames of microscopic video. This acceleration data can not only characterize the response sensitivity of the wax crystal to the external force field, but also derive its inertial mass parameter; the annular channel is equipped with a rotary encoder to record the angular velocity of the wax crystal in real time when it flows around. This parameter quantitatively describes the intensity of its rotation behavior by tracking the rotation period of the wax crystal around the central axis of the channel, which is of key significance for evaluating the orientation control of the wax crystal in the curved channel. A specially designed strain gauge sensing channel (sensitivity coefficient 2.0) is embedded in the side wall of the channel to convert the mechanical signal into an electrical parameter. When wax crystals flow through, the 0.1-5μN force generated will cause the strain gauge resistance to change. After being converted into an electrical signal, combined with the motion trajectory of the wax crystals recorded by the high-speed imaging system, its mass distribution, moment of inertia and other dynamic parameters can be obtained simultaneously.

[0041] (3) The dynamic analysis module of the microfluidic system comprehensively evaluates the effect of pour point depressants on the structural stability of wax crystals through the synergistic effect of the variable diameter channel 4 and the multi-stage strain gauge array. In the variable diameter channel group, when the channel diameter is reduced from the initial 1000μm to the 20μm variable diameter channel through four steps (each step is reduced by 20%), when the wax crystals pass through the variable diameter channel, the shear force will act on the wax crystals and destroy the wax crystal structure. At the same time, through the embedded strain gauge, the strain gauge sensing channel adopts a sandwich structure integrated by MEMS technology: the bottom layer is a 50μm thick PDMS substrate, and the middle layer is a gold thin film strain gauge (sensitivity coefficient 2.0), which can monitor the pressure fluctuations on the channel wall in real time. After the pressure change is converted into an electrical signal, it can indirectly reflect the structural strength of the wax crystal to resist the shear force. Quantitative assessment of dynamic properties is based on four core parameters: diameter reflects the degree of wax crystal fragmentation. Intact wax crystals have a diameter of approximately 50-200 μm, which drops to 8-20 μm after fragmentation. Fractal dimension characterizes structural complexity and irregularity, and changes during the fragmentation process. Before fragmentation, the fractal dimension of wax crystals is relatively low and their structure is relatively regular. As fragmentation progresses, the shape of the wax crystals becomes more complex, and the fractal dimension increases. For example, the fractal dimension of unfragmented wax crystals ranges from 1.2-1.5, rising to 1.8-2.3 after fragmentation. Changes in number reflect the scale of fragmentation; a single fragmentation event can increase the number of wax crystals per unit volume by 3-5 times. Roundness measures the regularity of the shape, and changes during the fragmentation process. Unfragmented wax crystals have a relatively high roundness, while fragmented wax crystals experience a decrease in roundness due to their irregular shapes. Changes in roundness reflect changes in the shape of the wax crystals during the fragmentation process. Intact wax crystals have a roundness value of 0.8-0.9, which drops to 0.4-0.6 after fragmentation.

[0042] (4) Finally, the microfluidic system uses a multi-dimensional weighted integration method to comprehensively evaluate the performance of the pour point depressant and realizes quantitative analysis by constructing a hierarchical index system. In the morphological index layer (weight a), the average size of wax crystals is given the highest weight (a) because it directly affects the flow resistance of crude oil. i ), the inhibitory effect of the pour point depressant on wax crystal growth is reflected by statistically analyzing the median of diameter distribution (e.g., from 120 μm to 80 μm); the number of wax crystals per unit volume (weight a j ) and size dispersion (weight a k ) together characterize the uniformity of the system. For example, high-quality pour point depressants can control the increase in quantity within 50%. The kinematic parameter layer (weight b) focuses on the linear velocity (b i ) and acceleration (b j ), for example: when the pour point depressant is effective, the wax crystal linear velocity can be increased from 1.2 mm / s to 2.5 mm / s; the angular velocity index (b k) Pay special attention to the rotation retardation effect of wax crystals in the annular channel. High-performance pour point depressants can reduce the angular velocity by more than 40%. The dynamic characteristic layer (weight c) is based on the wax crystal crushing diameter (c i ) and fractal dimension (c j ) as the core, the typical data shows that the average diameter after crushing remains above 30μm, the fractal dimension does not increase by more than 0.4, and the number increase (c k ) and roundness attenuation (c m ) to construct a structural stability assessment matrix. Based on the normalized index data, the freezing point is measured experimentally to determine its quantitative relationship with the normalized parameters. The specific formula is as follows:

[0043] S=a×(a i D avg +a j N+a k σ)+b×(b i V+b j A+b k ω)+c×(c i D+c j F+c k N ad +c m C)

[0044] ΔT g =AS p +B

[0045] Where: a+b+c=1a i +a j +a k =1b i +b j +b k =1c i +c j +c k +c l =1

[0046] In the formula: S——normalized index, where a n 、b n 、c n is the weight of each parameter, see Table 1 for details;

[0047] ΔT g ——Pour point depression, °C, where A, B, and P are obtained by fitting experimental data;

[0048] D avg ——Average size of wax crystals, μm;

[0049] N——number of wax crystals per unit volume;

[0050] σ——wax crystal size dispersion;

[0051] V——wax crystal linear velocity, mm / s;

[0052] A——wax crystal acceleration, mm / s 2 ;

[0053] ω——angular velocity of wax crystal, rad / s;

[0054] D——wax crystal broken diameter, μm;

[0055] F——wax crystal fractal dimension;

[0056] N ad ——Increase in the number of wax crystals;

[0057] C——wax wafer roundness attenuation;

[0058] Table 1 Coefficient weight range table

[0059]

[0060] Before the microfluidic experiment, a small number of representative freezing point measurement experiments were carried out according to the conventional freezing point measurement method. The freezing point depression ΔT was obtained by measuring the freezing point of crude oil at different freezing point depressant concentrations. g The freezing point drop ΔT was established. g A quantitative functional relationship is established between the wax crystal microscopic motion behavior parameters and the final freezing point change under the effects of different pour point depressants, achieving a cross-scale quantitative characterization from microscopic structural evolution to macroscopic freezing point changes.

[0061] This quantitative evaluation system uses a collaborative analysis of 10 characteristic parameters. Compared to traditional single-parameter evaluation systems, this multi-dimensional analysis method significantly reduces evaluation errors through parameter coupling, resulting in more accurate results. More importantly, the established "micro-motion-macro-performance mapping model" not only reveals the mechanism of action of pour point depressants but also provides theoretical guidance for their molecular design and industrial application.

[0062] The data analysis and control module uses LSTM and PPO for deep synergy, enabling efficient pour point depressant screening from time-series dynamic modeling to intelligent decision-making optimization. The microfluidic chip collects pressure, viscosity, and wax crystal dynamic parameters (size, fractal dimension, and motion trajectory) in real time. The LSTM network extracts high-dimensional features within a 60-second time window (such as the evolution of wax crystal structure under temperature gradients) to construct a state vector representation system. The PPO reinforcement learning algorithm uses this state as input and makes decisions in a three-dimensional continuous action space. It drives strategy optimization through a weighted reward function (weight ratio 10:6:3:1) based on pour point depression, viscosity change rate, fractal dimension change rate, and cost control, forming a closed-loop intelligent control mechanism of "time-series perception-decision optimization."

[0063] A microfluidic chip-based intelligent screening and optimization method for pour point depressants with integrated multi-parameter detection:

[0064] (1) Step 1: Model pre-fitting experiment. Before conducting the microfluidic chip experiment, a small sample conventional freezing point determination experiment was first performed to systematically obtain the freezing point depression (ΔT g ) to establish basic data on the correlation between pour point depression and the subsequently proposed normalized index S. This preliminary experimental data will provide key parameter support for the subsequent optimization of intelligent pour point depressant formulations.

[0065] (2) Step 2, initialization and calibration phase. First, crude oil is injected into the chip through the injection channel, and the microfluidic valve is used to precisely control the fluid to enter the wax precipitation chamber at a stable flow rate. During this process, the chip and its auxiliary equipment are calibrated for multiple parameters: the temperature control system adjusts the initial temperature of the chamber to the experimental set value (such as 40°C) through a preset program and verifies its temperature fluctuation range; the pressure sensor is calibrated with a standard pressure source to ensure that the measurement error is less than 0.5%; the microscopic imaging system is calibrated for focal length and resolution through a calibration plate, and the image processing software simultaneously initializes the parameter settings, including setting the image acquisition frame rate (such as 30fps), adjusting the contrast threshold, and establishing a wax crystal recognition template. The entire calibration process can be completed within 3 minutes, providing accurate benchmark conditions for subsequent experiments.

[0066] (3) Step three: The microfluidic system achieves efficient mixing of pour point depressants through multi-channel coordinated control. Pour point depressant samples with different concentration ratios are introduced through the injection channel of the corresponding inner diameter. After preliminary mixing with crude oil in the storage chamber, they enter the micro-mixing chamber with integrated fishbone and cylindrical flow structures according to the preset program. During this process, the flow rate of each injection channel (0.1-5μL / min) is precisely controlled by the microfluidic valve group to ensure that different samples participate in the reaction at a stable flow ratio. The microstructure of the mixing chamber induces laminar flow disturbance, so that the crude oil and the pour point depressant can be fully contacted in a short time, and the mixing efficiency is improved several times compared with the traditional method.

[0067] (4) Step 4, data acquisition step. The data acquisition process of the microfluidic system realizes comprehensive parameter measurement through multi-sensor fusion technology. The temperature control system simulates the crude oil cooling and gelation process at a rate of 0.1℃ / min, and the micro-piezoelectric sensors (accuracy ±0.1kPa) integrated on both sides monitor the channel pressure fluctuations in real time. The micro-flow rate sensor and shear force sensor manufactured based on MEMS technology synchronously obtain fluid dynamics parameters, and the fluid viscosity change is calculated in combination with the Hagen-Poiseuille equation. The microscopic imaging system continuously shoots the dynamics of the wax precipitation chamber at a rate of 60 frames per second, capturing the morphological evolution (size and number, average size, size), motion trajectory (linear velocity, angular velocity, acceleration) and crushing behavior (wax crystal diameter, fractal dimension, wax crystal, roundness) of the wax crystals during the precipitation process. Different weights are given to the above parameters to construct the normalized parameter S. Based on the freezing point measured in step 1, its quantitative relationship with the normalized parameter is determined. After three-dimensional reconstruction and voxelization, all image data were extracted by feature statistical algorithm to extract wax crystal volume fraction (0.1%-15%), number density (102-104 / mm 3 ) and spatial distribution parameters (skewness 0.5-1.2, kurtosis 2.8-4.5), providing a complete data set for quantitative analysis of the effects of pour point depressants on the three-dimensional growth characteristics of wax crystals.

[0068] (5) Step five, data analysis and formulation optimization. Normalize the data of fluid mechanics parameters (pressure, viscosity, flow rate) and wax crystal temporal characteristics (morphological parameters, kinematic parameters, dynamic parameters and voxelization parameters) to eliminate dimensional differences; arrange the sensor data by time to form sequence data of time steps; use 60 time cloths as the input units of LSTM; then extract the temporal features of LSTM, and train the wax crystal evolution data containing different pour point depressant ratios by designing the input layer (pressure, viscosity, wax crystal size, fractal dimension, roundness) and the output layer (128-dimensional vector). Import the temporal feature vector output from LSTM into the PPO algorithm to construct the state space, select the pour point depression, viscosity reduction rate, wax crystal fractal dimension change rate and cost penalty to design the reward function, and obtain the pour point depressant concentration optimization scheme and reward value by calculation. The relevant parameters are then input into the long-short-term memory network to construct a pour point depressant effect evaluation model. Finally, a genetic algorithm is used to perform formula optimization search. The pour point depressant content is used as the genetic code. Through selection, crossover, mutation and other operations, new formula combinations are continuously generated, and the fitness of each combination is calculated based on the evaluation model. After multiple generations of evolutionary iterations, it gradually converges to the optimal or near-optimal pour point depressant formula.

[0069] Example:

[0070] This microfluidic chip-based intelligent screening and optimization device for pour point depressants, featuring integrated multi-parameter detection, includes an experimental module, a data acquisition module, and a data analysis and control module. The experimental module comprises a power zone, a delivery zone, a temperature control zone, an imaging zone, and a recovery zone. The power zone includes a constant-flow pump and mechanical devices, which provide kinetic energy to ensure uniform entry of the medium into the microfluidic chip. The delivery zone includes a micro-sampler and a delivery hose. The temperature control zone primarily features a Peltier temperature control station, ensuring that the sample in the microfluidic chip remains stable within the experimental set temperature range. The imaging zone primarily includes a polarizing microscope and a microfluidic chip. The recovery zone includes a waste liquid recovery hose and a waste liquid recovery tank. The power zone transports the sample from the micro-sampler via the delivery hose to the imaging zone and ultimately into the recovery zone. The chip includes an injection channel, microfluidic valves, a storage channel, a micromixing chamber, and a wax precipitation chamber. The chip is primarily constructed from polydimethylsiloxane (PDMS) or glass, which has high chemical stability and good biocompatibility. The injection channel can be set to different sizes according to experimental requirements, with a width of 100μm-1000μm and a depth of 5μm-500μm. The opening and closing of each injection channel are precisely controlled by a microfluidic valve, and multiple depressant samples and the fluid to be treated can be connected at the same time; the micro-mixing chamber is located in the central area of ​​the chip, with a volume of 1μl-100μl. The inner wall of the chamber is coated with a super-hydrophobic and anti-adsorption coating. The micro-mixing structure adopts a staggered fishbone design, which uses the principle of fluid dynamics to promote efficient mixing of the depressant and crude oil in a short time; the wax precipitation chamber is a key part of the microfluidic chip depressant screening and formula optimization device, and is an important platform for studying the characteristics of wax crystals under different conditions. The wax precipitation chamber is designed to simulate the actual flow environment of crude oil and provide a place for studying the behavior of wax crystals under different conditions. By observing the morphological, kinematic and dynamic behaviors of wax crystals, relevant data are obtained to evaluate the effect of the depressant and provide a basis for screening and optimizing the depressant formula. A feasible microfluidic chip structure example is given below. Figure 3 The specific microfluidic chip structure and monitored parameters depend on the oil product and pour point depressant. The internal structure of this microfluidic chip example and the parameters measured in different observation zones are as follows:

[0071] (1) Part I: Wax crystal morphology observation area

[0072] Straight channels and straight channels with baffles at varying intervals were set up. The straight channel was used to observe the natural formation process of wax crystals, while the baffled channel promoted the aggregation and collision of wax crystals. Using a high-resolution microscope imaging system and image analysis algorithms, parameters such as wax crystal size, average size, and number were obtained, allowing for microscopic analysis of the effects of pour point depressants on wax crystal morphology.

[0073] a. Size and quantity of wax crystals

[0074] The concentration of pour point depressants and the composition of crude oil significantly influence the size and number of wax crystals. Pour point depressants alter the crystallization process of wax crystals, causing them to flocculate, increase their size, and change their number. For example, some pour point depressant molecules can adsorb on the surface of wax crystals, altering their surface energy and causing them to coalesce. Studying the size and number of wax crystals provides a direct understanding of how pour point depressants alter their microstructure. Larger, fewer wax crystals reduce interactions between them, helping to improve crude oil flow. In the first section of the wax separation chamber, straight channels and straight channels with baffles at varying intervals are designed to observe the formation, aggregation, and collision of wax crystals. High-resolution microscopy imaging systems and image analysis software are used to process the captured images and determine the size and number of wax crystals. A specific image recognition algorithm accurately identifies the outline of each wax crystal, allowing the lengths of its major and minor axes to be measured as a measure of crystal size. The number of wax crystals is directly determined using the algorithm's object counting function.

[0075] b. Average size of wax crystals

[0076] The type and amount of pour point depressant added are key factors affecting the average size of wax crystals. Different types of pour point depressants have different effects on the growth of wax crystals, resulting in different changes in average size. Increasing the amount of pour point depressant added will usually further flocculate the average size of the wax crystals. The change in average size reflects the ability of the pour point depressant to regulate the growth of wax crystals. A larger average size means that the pour point depressant can more effectively aggregate wax crystals, thereby improving the fluidity of crude oil. The average size of wax crystals is a comprehensive consideration of the sizes of many wax crystals. After obtaining a large amount of wax crystal size data, it is calculated by arithmetic averaging. This calculation method can more comprehensively reflect the overall size characteristics of wax crystals in the observation area. The specific formula is:

[0077]

[0078] Where: is the average size of wax crystals; i is each wax crystal; n is the total number of wax crystals; x i is the long axis of the i-th wax crystal; y i is the minor axis of the i-th wax crystal.

[0079] c. Size of wax crystals

[0080] Since wax crystals are irregular in shape, the concept of equivalent diameter is introduced to more accurately describe their size. Assuming that the wax crystal is round, the equivalent diameter refers to the diameter of a circle with the same area as the wax crystal. The calculation formula is:

[0081]

[0082] Where A is the wax crystal area obtained by image analysis; d is the equivalent diameter of the wax crystal.

[0083] By calculating the equivalent diameter, wax crystals of varying shapes can be compared on a unified, quantitative scale. The equivalent diameter is influenced by the action of the pour point depressant and the crude oil environment. Pour point depressants alter the crystallization behavior of wax crystals, causing them to change their growth morphology, which in turn affects the equivalent diameter. Environmental factors such as crude oil temperature and pressure also influence the growth and morphology of wax crystals, thereby altering the equivalent diameter. Graphical analysis of the equivalent diameter allows for a more intuitive comparison of wax crystal size differences under different conditions, allowing assessment of the impact of pour point depressants on their macroscopic size.

[0084] (2) Part II: Wax crystal movement behavior observation area

[0085] This section includes an annular channel and channels with inclined surfaces at varying angles. The annular channel induces circular motion in the wax crystals, allowing for the measurement of linear and angular velocities. The inclined channel modulates the direction and velocity of the wax crystals, acquiring acceleration data. Using high-speed cameras and digital image correlation technology, the wax crystals' motion is tracked, their motion parameters calculated, and the effects of pour point depressants on their motion are investigated.

[0086] a. Linear velocity of wax crystal movement

[0087] Linear velocity refers to the distance that a wax crystal moves per unit time on its trajectory, and is used to describe the speed of the translational motion of wax crystals in microfluidic channels. The concentration of the pour point depressant and the characteristics of the crude oil significantly affect the linear velocity of the wax crystals. The pour point depressant changes the microstructure and surface properties of the wax crystals, reduces the interaction force between them and the crude oil molecules, makes the wax crystals easier to move in the crude oil, and increases the linear velocity. Different crude oils have different hindering effects on the movement of wax crystals due to differences in composition. The wax precipitation chamber in this part is equipped with a straight channel and a channel with inclined surfaces at different angles. A high-speed camera combined with image tracking technology is used to measure the linear velocity of the wax crystals. By analyzing the continuous images taken, the position changes of the wax crystals at adjacent moments are determined, and the linear velocity is calculated according to the linear velocity calculation formula. The formula is as follows:

[0088]

[0089] Where Δs is the displacement of a wax crystal within a time interval Δt. Studying the linear velocity provides an intuitive understanding of the effect of pour point depressants on the migration of wax crystals in crude oil. An increase in linear velocity indicates more active movement of wax crystals within the crude oil, potentially leading to improved crude oil fluidity.

[0090] b. Angular velocity of wax crystal motion

[0091] Angular velocity is used to measure the speed at which a wax crystal rotates around a fixed point. It is important when studying the rotational behavior of wax crystals in a circular channel. This observation area is set up in a circular channel. By analyzing the motion images of wax crystals in the circular channel captured by a high-speed camera, specific marking points on the wax crystals are selected to track their angular changes at different times. Angular velocity is calculated using the angular velocity calculation formula. The formula is as follows:

[0092]

[0093] The angular velocity is calculated by the angle that the marked point Δθ rotates within the time interval Δt. The structural parameters of the annular channel of the microfluidic chip and the properties of the pour point depressant affect the angular velocity of the wax crystal. Structural factors such as the annular flow channels of different diameters in the channel determine the flow pattern of the fluid in the channel, which in turn affects the rotation of the wax crystal. Calculating the angular velocity of the wax crystal helps to gain a deeper understanding of the effect of the pour point depressant on the rotation behavior of the wax crystal in a complex flow field, and reflects the effect of the pour point depressant on the microscopic flow characteristics inside the crude oil.

[0094] c. Acceleration of wax crystal motion

[0095] Acceleration describes how quickly a wax crystal changes velocity, including linear acceleration and angular acceleration. In channels with inclined surfaces and variable diameters, wax crystals experience non-uniform force fields, generating acceleration. By processing the linear or angular velocity data of wax crystals at different times, acceleration can be calculated using the acceleration calculation formula. The formula is as follows:

[0096]

[0097] Where: a is the linear acceleration; a w is the angular acceleration. Changes in channel geometry and the action of the pour point depressant affect wax crystal acceleration. The angle of the inclined channel and the contraction rate of the variable-diameter channel determine the degree of change in the fluid's force on the wax crystals. The pour point depressant modifies the force characteristics of the wax crystals, making them more susceptible to acceleration or deceleration. Studying acceleration can reveal the impact of pour point depressants on the motion response of wax crystals in dynamic environments and reflect the dynamic process by which pour point depressants enhance crude oil fluidity.

[0098] (3) Part III: Wax crystal dynamic behavior observation area

[0099] A variable diameter channel with internal strain gauges is used. The variable diameter channel varies the fluid velocity, exerting varying shear forces on the wax crystals. When crude oil containing wax crystals flows through a variable diameter pipeline, the fluid velocity changes due to the change in pipeline diameter. According to the fluid continuity equation, the flow velocity increases in the constriction section of the pipeline and decreases in the expansion section. This change in flow velocity causes the fluid to exert varying degrees of shear forces on the wax crystals. In the constriction section, the flow velocity increases, increasing the shear forces on the wax crystals; in the expansion section, the shear forces decrease. Different pour point depressant concentrations alter the internal structure and surface properties of the wax crystals, thereby affecting their ability to withstand shear forces. An appropriate pour point depressant concentration enhances the structural stability of the wax crystals, making them less susceptible to breakage under high shear forces. The pour point depressant concentration plays a key role in the breakage of wax crystals. When the pour point depressant concentration is too low, the wax crystals become fragile and easily break under low shear forces. While excessively high pour point depressant concentrations enhance the stability of the wax crystals, they may alter the overall properties of the crude oil and affect its fluidity. In addition to the pour point depressant, the crude oil's temperature, pressure, and the initial size and morphology of the wax crystals also influence the crushing process. Increasing temperature increases the flexibility of wax crystals, making them relatively less susceptible to breakage. Pressure changes alter the interaction between wax crystals and crude oil molecules, indirectly influencing the crushing process. Furthermore, strain gauges were installed within the channel to measure the resistance of the wax crystals to motion. Using a microscope imaging system and a high-speed camera, combined with image processing algorithms, parameters such as wax crystal diameter, fractal dimension, number, and roundness were measured to analyze the effect of pour point depressant concentration on the wax crystal crushing process.

[0100] a. Wax crystal diameter is a direct indicator of the degree of wax crystal fragmentation. During the fragmentation process, large-diameter wax crystals gradually break into smaller ones. By measuring and analyzing wax crystal diameters at different times, we can understand the degree and rate of wax crystal fragmentation. Using a microscope imaging system combined with image processing software, we can accurately measure wax crystal diameter.

[0101] b. Fractal dimension is used to describe the complexity of wax crystals. During the fragmentation process, the fractal dimension of wax crystals changes. Before fragmentation, the fractal dimension of wax crystals is relatively low and the structure is relatively regular. As fragmentation progresses, the shape of the wax crystals becomes more complex, and the fractal dimension increases. This change in fractal dimension can reflect the complexity and irregularity of the wax crystal fragmentation. Algorithms such as the box counting dimension can be used to calculate the fractal dimension of wax crystals.

[0102] c. The number of wax crystals increases during the breakup process. A larger wax crystal breaks into smaller ones. By counting the number of wax crystals, we can intuitively understand the extent of breakage. Using image recognition technology and counting algorithms, we can accurately count the number of wax crystals within a specific area.

[0103] d. Roundness measures the degree to which a wax crystal's shape approaches a circle. During the crushing process, the roundness of a wax crystal changes. Uncrushed wax crystals have a relatively high roundness, while crushed wax crystals, due to their irregular shapes, have a reduced roundness. This change in roundness reflects the shape changes during the crushing process. By calculating the circumference and area of ​​the wax crystal, a formula can be used to determine the roundness value.

[0104]

[0105] A is the area of ​​the wax crystal, P is the circumference of the wax crystal, and the closer the roundness is to 1, the closer the wax crystal is to a circle.

[0106] The system's temperature control system includes a microscopic temperature control device, which cools the crude oil in the observation channel to the desired experimental conditions, preparing for microscopic imaging of wax crystals in the presence of a pour point depressant. The imaging system primarily includes a polarizing microscope, which is used to capture and record the wax crystal formation process. The data acquisition module collects pressure data across the chip, uses microelectromechanical systems (MEMS) technology to obtain flow rate and shear force parameters from high-precision sensors within the microfluidic chip, and uses a microscope imaging system to capture images of wax crystal precipitation within the wax precipitation chamber. The data analysis and control module primarily uses engineering fluid mechanics theory to calculate the viscosity of the fluid within the chip and voxelizes images of the wax crystal precipitation process. Voxel-based statistical parameters are used to calculate the volume fraction, number density, and other statistical distribution parameters of the wax crystals.

[0107] This embodiment provides a method for rapid screening and intelligent optimization of pour point depressants on a microfluidic chip with integrated multi-parameter detection:

[0108] (1) Step 1: Model pre-fitting experiment. Before conducting the microfluidic chip experiment, a small sample conventional freezing point determination experiment was first performed to systematically obtain the freezing point depression (ΔT g ) to establish basic data on the correlation between pour point depression and the subsequently proposed normalized index S. This preliminary experimental data will provide key parameter support for the subsequent optimization of intelligent pour point depressant formulations.

[0109] (2) Step 2: Initialization and calibration. Inject the sample into the microfluidic chip through the injection channel, and use the microfluidic valve to control the sample to enter the wax precipitation chamber at a stable flow rate. At the same time, initialize and calibrate the microfluidic chip and related equipment, such as calibrating the temperature control system to make the initial temperature in the cavity consistent with the preset experimental temperature, checking the pressure gauge to ensure more accurate pressure measurement, and debugging the microscopic imaging system and image processing software to make them in the initial state.

[0110] (3) Step three, inject and mix the pour point depressant. Place the pre-prepared pour point depressants of different concentrations in the micro-sampler. At this time, the outlet valve of the storage channel is in a closed state. The pour point depressant under the action of the constant flow pump flows through the storage channels of different inner diameters and is stored. The crude oil is evenly transported from the injection channel to the storage channel. The valve is opened, and the crude oil and pour point depressant enter the fishbone-shaped micro-mixing chamber at the same time to be fully and evenly mixed. Then, they enter the wax precipitation channel, and the mixed crude oil in the wax precipitation channel is cooled according to the preset experimental requirements. The injection process requires precise control of the flow rate of the pour point depressant in the channel to ensure that the flow rate of different samples entering the reaction system is stable and can be precisely controlled.

[0111] (4) Step 4, data acquisition step. The temperature control system in the microfluidic chip is used to simulate the cooling process. The high-precision pressure gauges set on both sides of the chip can detect the pressure changes in the channel. In the microelectromechanical system, microsensors based on different principles can be designed and manufactured to achieve data measurement. When crude oil and pour point depressant flow in the microfluidic chip, the flow rate, shear force and other parameters of the fluid are obtained by combining microelectromechanical system technology. The change in fluid viscosity is indirectly calculated through the principle of fluid mechanics. At the same time, a microscopic imaging system and image acquisition are used to capture the image of the wax crystal precipitation process in the wax precipitation chamber, and the wax precipitation chamber dynamics are continuously photographed at a rate of 60 frames per second to capture the morphological evolution of the wax crystal precipitation process, namely morphology (size and number, average size, size), motion trajectory, namely kinematic behavior (linear velocity, angular velocity, acceleration) and crushing behavior, namely dynamic behavior (wax crystal diameter, fractal dimension, wax crystal, roundness). The monitored parameters are determined according to different oil products and pour point depressants. Different weights are given to the above parameters to construct a normalization parameter S. Based on the freezing point measured in step one, its quantitative relationship with the normalization parameter is determined. In addition, the microscopic images of wax crystals taken in multiple directions are voxelized. After all image data are three-dimensionally reconstructed and voxelized, the appropriate voxel size is selected according to the resolution of the image and the required accuracy to divide the uniform voxel grid, which is represented by a three-dimensional array, and each element of the array corresponds to a voxel. For each voxel, a corresponding attribute is assigned according to its position in the microfluidic chip. The presence state of wax crystals in the voxel area is represented by a binary attribute (1 for wax crystals and 0 for no wax crystals), or a continuous attribute is assigned, such as the local concentration, density or grayscale value of the wax crystals (for grayscale images). By assigning values, the distribution and properties of wax crystals can be discretized to prepare for subsequent quantitative analysis. The voxel-based characteristic statistical method was used to quantitatively calculate the wax crystal volume fraction, wax crystal number density, wax crystal standard deviation skewness and peak value, and analyze the distribution characteristics of wax crystals in three-dimensional space, providing an important basis for studying the formation and growth of wax crystals during the cooling process of crude oil under the action of different concentrations of pour point depressants.

[0112] Confidentiality test:

[0113] 1) 10 key parameters such as size and quantity, average size, size, linear velocity, angular velocity, acceleration, wax crystal diameter, fractal dimension, wax crystal and roundness are selected to analyze the morphological evolution, motion trajectory and fragmentation behavior of wax crystals during precipitation. The calculation formula selected in the example is as follows:

[0114] ① Wax Crystal Size and Quantity: In the first section of the wax precipitation chamber, straight channels and straight channels with baffles of varying spacing are installed to observe the formation, aggregation, and collision of wax crystals. Using a high-resolution microscope imaging system and image analysis software, the captured images are processed to determine the size and quantity of the wax crystals. A specific image recognition algorithm accurately identifies the outline of each wax crystal, measuring its major and minor axis lengths as a measure of crystal size. The number of wax crystals is directly determined using the algorithm's target counting function.

[0115] ② Average size of wax crystals: The average size of wax crystals is a comprehensive indicator of the sizes of many wax crystals. After obtaining a large number of wax crystal size data, it is calculated by arithmetic average method. The selected calculation formula is:

[0116]

[0117] Where: is the average size of wax crystals; i is each wax crystal; n is the total number of wax crystals; x i is the long axis of the i-th wax crystal; y i is the minor axis of the i-th wax crystal.

[0118] ③ Wax crystal size: Due to the irregular shape of wax crystals, the concept of equivalent diameter is introduced to more accurately describe their size. Assuming that the wax crystal is round, the equivalent diameter refers to the diameter of a circle with the same area as the wax crystal. The calculation formula is:

[0119]

[0120] Where A is the wax crystal area obtained by image analysis; d is the equivalent diameter of the wax crystal.

[0121] ④ Linear velocity of wax crystal movement: The wax separation chamber in this part is equipped with straight channels and channels with inclined surfaces at different angles. The linear velocity of wax crystals is measured using a high-speed camera combined with image tracking technology. By analyzing the continuous images taken, the position changes of wax crystals at adjacent moments are determined, and the linear velocity is calculated according to the linear velocity calculation formula. The selected calculation formula is:

[0122]

[0123] Where Δs is the displacement of the wax crystal in the time interval Δt.

[0124] 5. Angular velocity of wax crystal movement: A circular channel is set up in the observation area. By analyzing the movement images of wax crystals in the circular channel captured by a high-speed camera, specific marking points on the wax crystals are selected to track their angular changes at different times. The angular velocity is calculated according to the angular velocity calculation formula. The selected calculation formula is:

[0125]

[0126] where Δθ is the angle that the marker point rotates during the time interval Δt to calculate the angular velocity.

[0127] ⑥ Acceleration of wax crystal movement: By processing the linear velocity or angular velocity data of wax crystals at different times, the acceleration is calculated according to the acceleration calculation formula. The selected calculation formula is:

[0128]

[0129] Where: a is the linear acceleration; a w is the angular acceleration.

[0130] ⑦ Wax crystal diameter is a direct indicator for assessing the degree of wax crystal fragmentation. During the fragmentation process, large-diameter wax crystals gradually break into smaller ones. By measuring and statistically analyzing wax crystal diameters at different times, we can understand the degree and rate of wax crystal fragmentation. Using a microscope imaging system combined with image processing software, we can accurately measure wax crystal diameter.

[0131] ⑧ Fractal dimension is used to describe the complexity of wax crystals. During the fragmentation process, the fractal dimension of wax crystals changes. Before fragmentation, the fractal dimension of wax crystals is relatively low and the structure is relatively regular. As the fragmentation progresses, the shape of the wax crystals becomes more complex, and the fractal dimension increases. This change in fractal dimension can reflect the complexity and irregularity of the wax crystal fragmentation. Algorithms such as the box counting dimension can be used to calculate the fractal dimension of wax crystals.

[0132] 9. The number of wax crystals increases during the breakup process. A larger wax crystal breaks into smaller ones. By counting the number of wax crystals, we can intuitively understand the extent of breakage. Using image recognition technology and counting algorithms, we can accurately count the number of wax crystals within a given area.

[0133] ⑩ Roundness measures the degree to which a wax crystal's shape approaches a circle. During the crushing process, the roundness of a wax crystal changes. Uncrushed wax crystals have a relatively high roundness, while crushed wax crystals, due to their irregular shapes, have a lower roundness. This change in roundness reflects the shape changes during the crushing process. By calculating the circumference and area of ​​the wax crystal, a formula can be used to determine the roundness value.

[0134]

[0135] Where: A is the area of ​​the wax crystal, P is the circumference of the wax crystal, and the closer the roundness is to 1, the closer the wax crystal is to a circle. The indicators obtained from the three observation parts are combined to evaluate the effect of the pour point depressant. During the evaluation process, different indicators are assigned corresponding weights. In the first part, considering that the average size has a greater impact on the fluidity of crude oil, it can be given a weight of 0.4, and the weights of the number and size of wax crystals are set to 0.3 respectively. In the second part, linear velocity and acceleration have a more critical impact on the movement of wax crystals and the overall fluidity of crude oil. They can be assigned weights of 0.4 and 0.3 respectively, and the weight of angular velocity is set to 0.3. In the third part, the diameter and fractal dimension of wax crystals can better reflect the breakage and structural changes of wax crystals under shear. The weight can be set to 0.35, and the weights of quantity and roundness are set to 0.3. By calculating the weighted comprehensive score, the scores of different pour point depressants under the same experimental conditions are compared. The higher the score, the better the effect of the pour point depressant in improving the characteristics of wax crystals and improving the fluidity of crude oil. The specific calculation method is:

[0136] S=0.4×(0.4D avg +0.3N+0.3σ)+0.35×(0.4V+0.3A+0.3ω)+0.25×(0.35D+0.35F+0.15N ad +0.15C)

[0137] ΔT g =AS p +B

[0138] Where: D avg ——Average size of wax crystals, μm;

[0139] N——number of wax crystals per unit volume;

[0140] σ——wax crystal size dispersion;

[0141] V——wax crystal linear velocity, mm / s;

[0142] A——wax crystal acceleration, mm / s 2 ;

[0143] ω——angular velocity of wax crystal, rad / s;

[0144] D——wax crystal broken diameter, μm;

[0145] F——wax crystal fractal dimension;

[0146] N ad ——Increase in the number of wax crystals;

[0147] C——wax wafer roundness attenuation;

[0148] 2) In this example, five key parameters of the voxel feature statistical method, such as wax crystal volume fraction statistics, wax crystal number density statistics, standard deviation, skewness, and kurtosis, are selected for statistical analysis. The calculation formula selected in the example is as follows:

[0149] ① Wax crystal volume fraction statistics: By counting the number of voxels containing wax crystals and comparing them to the total number of voxels, the volume fraction of wax crystals is calculated. Changes in the wax crystal volume fraction at different temperatures can reflect the growth or dissolution process of wax crystals, as well as the effect of the pour point depressant on wax crystal growth.

[0150] V voxel =a 3

[0151] V wax =n×s×h

[0152]

[0153] Where: a is the side length of the voxel; n is the number of pixels in the wax crystal area within the voxel; s is the actual area corresponding to each pixel; h is the thickness of the wax crystal in the direction perpendicular to the imaging plane; V voxel is the volume of a single voxel; V wax,i is the volume of wax crystals contained in each voxel i; F i is the volume fraction of wax crystals in the voxel.

[0154] By analyzing the average volume fraction of wax crystals and their distribution in different regions within the entire three-dimensional space, we can understand the relative content distribution of wax crystals in the three-dimensional space. The formula for calculating the average volume fraction of wax crystals is as follows:

[0155]

[0156] Where n is the total number of voxels; F i is the volume fraction of wax crystals in the voxel; is the average area fraction of wax crystals.

[0157] ② Wax crystal density statistics: First, the number of individual wax crystals is determined through methods such as connected component analysis. The wax crystal density is then calculated based on the total volume of the voxelized space. By calculating the wax crystal density across the entire three-dimensional space under varying pour point depressant concentrations, we can understand the density and distribution of wax crystals within the space.

[0158]

[0159] where N wax,i is the number of wax crystals in each voxel; ρ i is the number density of wax crystals, i.e. the number of wax crystals per unit volume; is the average number density of wax crystals in the entire three-dimensional space.

[0160] ③ Standard Deviation: This can be used to measure the degree of dispersion of wax crystals in three-dimensional space, reflecting the uniformity of wax crystal distribution. A higher standard deviation indicates an uneven distribution of wax crystals within the microfluidic chip, potentially affected by the temperature gradient and varying concentrations of the pour point depressant. A lower standard deviation may indicate a relatively uniform distribution of wax crystals.

[0161]

[0162] Where n is the total number of voxels; F i is the volume fraction of wax crystals in the voxel; is the average area fraction of wax crystals, and σ is the standard deviation of wax crystals.

[0163] ④ Skewness: This can be used to characterize the asymmetry of wax crystal distribution. Skewness reveals the spatial orientation of wax crystals. A positive skewness may indicate that wax crystals tend to be more concentrated on one side of the chip or in a certain local area. A skewness less than 0 indicates the opposite.

[0164]

[0165] Where: n is the total number of voxels; F i is the volume fraction of wax crystals in the voxel; is the average area fraction of wax crystals, σ is the standard deviation of wax crystals; S is the skewness.

[0166] ⑤ Kurtosis: This reflects the sharpness or flatness of the wax crystal distribution. A high kurtosis may indicate a high concentration of wax crystals in certain local areas, while a low kurtosis may indicate a more dispersed distribution. This helps to further understand the aggregation and dispersion patterns of wax crystals in space.

[0167]

[0168] Where: n is the total number of voxels; F i is the volume fraction of wax crystals in the voxel; is the average area fraction of wax crystals, σ is the standard deviation of wax crystals; S is the skewness of wax crystals within the voxel; K is the kurtosis of wax crystals within the voxel.

[0169] (5) Step five, data analysis and formula optimization step. Normalize the data such as fluid mechanics parameters (pressure, viscosity, flow rate) and wax crystal temporal characteristics (morphological parameters, kinematic parameters, dynamic parameters and voxelization parameters) to eliminate dimensional differences; arrange the sensor data by time to form sequence data of time steps; use 60 time cloths as the input units of LSTM; then extract the temporal features of LSTM, and train the wax crystal evolution data containing different pour point depressant ratios by designing the input layer (pressure, viscosity, wax crystal size, fractal dimension, roundness) and the output layer (128-dimensional vector). Import the temporal feature vector output from LSTM into the PPO algorithm to construct the state space, select the pour point depression, viscosity reduction rate, wax crystal fractal dimension change rate and cost penalty to design the reward function, and obtain the pour point depressant concentration optimization scheme and reward value by calculation. The reward function formula is as follows:

[0170]

[0171] Where: ΔT g ——Depression of freezing point (℃), the difference between the initial and current freezing points;

[0172] ——Viscosity reduction rate, where μ t is the minimum viscosity of the current formula crude oil, μ max is the initial viscosity of crude oil;

[0173] ——The change rate of wax crystal fractal dimension, where F0 is the initial fractal dimension, F t is the current value;

[0174] C total ——Total drug concentration

[0175] The final dynamic result example is as follows:

[0176] Initial state: solidification point 28℃, viscosity Fractal dimension 1.6.

[0177] LSTM output: Predict that the freezing point will drop to 22°C in the next ten minutes.

[0178] PPO decision: Increase the pour point depressant concentration by 247 ppm.

[0179] Result: The actual freezing point dropped to 16°C, and the reward value increased by 2.7.

[0180] Furthermore, the model can achieve the optimal formulation for a given pour point depressant concentration. Data from the microfluidic chip is fed into a pour point depressant screening and optimization model collaboratively constructed using LSTM and PPO, accumulating large amounts of data and building a database. In subsequent field applications, by assigning different pour point depressant concentrations and using this concentration as the genetic code, new formulation combinations are continuously generated through selection, crossover, and mutation. The fitness of each combination is calculated based on the evaluation model. After multiple generations of evolutionary iterations, the model gradually converges to an optimal or near-optimal pour point depressant formulation.

[0181] This part achieves efficient dynamic adjustment of the pour point depressant formulation through LSTM dynamic time series modeling and PPO policy gradient optimization. Its core advantages are: LSTM captures the dynamic evolution of wax crystal behavior, addressing the "time blind spot" of traditional static experiments. It can accurately predict the optimal time point for the crude oil to react with the pour point depressant, reducing time costs; the PPO algorithm utilizes these rules to adjust the formulation in real time, avoiding the policy lag caused by traditional methods that ignore the time dimension.

[0182] The present invention can accurately simulate the dynamic flow environment of crude oil from a pipeline within the chip through a miniaturized flow channel network (width 50-500μm) and an integrated control system. Its core advantages are: reconstructing the shear field of crude oil flow through micron-level flow channel design, using the temperature control system to simulate the temperature gradient changes, and combining optical detection to track the growth of wax crystals and the network formation process. Compared with traditional methods, microfluidic technology only requires a few milliliters of crude oil samples to complete the screening of pour point depressants. The experimental cycle is shortened from several weeks to several hours, and multi-dimensional data such as the evolution of wax crystal morphology and changes in rheological parameters can be obtained simultaneously. More importantly, this technology can dynamically simulate the complex mechanical and thermodynamic conditions in actual working conditions, providing an experimental platform that is closer to real scenarios for the performance evaluation of pour point depressants, and significantly improving the correlation between laboratory data and field application effects.

Claims

1. A microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection, characterized by: This microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection includes an experimental module, a data acquisition module, and a data analysis and control module. The experimental module includes a power area, a conveying area, a temperature control area, an imaging area, and a recovery area. The imaging area includes a polarizing microscope and a microfluidic chip. The microfluidic chip includes an injection channel, a storage chamber, a micro-mixing chamber, a wax precipitation chamber, and a temperature control system. Multiple pour point depressant samples and the fluid to be treated are simultaneously connected through each injection channel. The inner wall of the micro-mixing chamber is coated with a super-hydrophobic anti-adsorption coating. Combined with the staggered fishbone structure and cylindrical flow design, the fluid shear effect is used to achieve efficient mixing of the pour point depressant and crude oil; the wax precipitation chamber simulates the complex flow environment of a real oil pipeline, and is equipped with a wax crystal morphology observation area, a wax crystal movement behavior observation area, and a wax crystal dynamic behavior observation area. A combination of straight channels, annular channels, channels with inclined surfaces, and variable-diameter channels is used to observe the flow state of wax crystals under different environments and collect wax crystal dynamic parameters; The data acquisition module collects the pressure of the fluid in the microfluidic chip, obtains the flow rate and shear force parameters of the fluid through the microelectromechanical system, and uses the microscope imaging system and image acquisition to capture images of the wax crystal precipitation process in the fluid in the wax precipitation chamber to obtain the wax crystal volume fraction, wax crystal number density, wax crystal standard deviation skewness and peak value; the data analysis and control module intelligently screens the pour point depressant and dynamically optimizes the pour point depressant ratio strategy based on the long short-term memory network LSTM and PPO reinforcement learning algorithm.

2. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 1, characterized in that: The collection of dynamic parameters of wax crystals specifically includes: collecting morphological parameters of wax crystals, including the size, average size and number of wax crystals under cooling, collecting kinematic behavior parameters of wax crystals, including the linear velocity, angular velocity and acceleration of wax crystals, and collecting dynamic behavior parameters of wax crystals, including the size, number, fractal dimension and roundness of wax crystals under shear force.

3. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 2, characterized in that: The power zone includes a constant flow pump and a mechanical device. The constant flow pump provides kinetic energy so that the medium can evenly enter the interior of the microfluidic chip; the delivery zone includes a micro sampler and a delivery hose; the temperature control zone includes a Peltier temperature control table to stabilize the sample in the microfluidic chip in the experimental set temperature range; the recovery zone includes a waste liquid recovery hose and a waste liquid recovery barrel; under the action of the power zone, the sample in the micro sampler is transported to the imaging area through the delivery hose and finally flows into the recovery zone.

4. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 3, characterized in that: The microfluidic chip is made of polydimethylsiloxane (PDMS) or glass, and its injection channels are set to different sizes according to experimental requirements, with a width of 100μm-1000μm and a depth of 5μm-500μm. Each injection channel is precisely controlled to open and close by a microfluidic valve, and multiple pour point depressant samples and fluids to be treated are connected at the same time; the micro-mixing chamber is located in the central area of ​​the chip and has a volume of 1μl-100μl.

5. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 4, characterized in that: The wax crystal morphology observation area is provided with an ordinary straight channel and a straight channel with baffles. The straight channel with baffles is a straight channel in which a plurality of baffles are arranged, and the spacing between two adjacent baffles is different. The straight channel is used to observe the formation process of wax crystals in a natural state, and the straight channel with baffles promotes the aggregation and collision of wax crystals. Through a high-resolution microscope imaging system and an image analysis algorithm, the size, average size and number of wax crystals are obtained, and the influence of the pour point depressant on the wax crystal morphology is analyzed from a microscopic level.

6. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 5, characterized in that: The wax crystal motion behavior observation area includes an annular channel and an inclined surface channel. The inclined surface channel is a channel with inclined surfaces at different angles. The annular channel causes the wax crystal to produce circular motion, which is used to measure linear velocity and angular velocity. The inclined channel changes the direction and speed of wax crystal movement to obtain acceleration data; with the help of high-speed cameras and digital image correlation technology, the movement trajectory of wax crystals is tracked, the motion parameters are calculated, and the effect of pour point depressants on the movement state of wax crystals is studied.

7. The microfluidic chip-based intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 6, characterized in that: The wax crystal dynamic behavior observation area adopts a variable diameter channel and a strain gauge channel. The strain gauge channel is a variable diameter channel with strain gauges arranged in an internal array. The variable diameter channel changes the fluid velocity and applies different shear forces to the wax crystals. When crude oil containing wax crystals flows through the variable diameter channel, the flow velocity of the fluid will change. The change in flow velocity causes the fluid to apply different degrees of shear force to the wax crystals. In the contraction section, the flow velocity accelerates and the shear force on the wax crystals increases; in the expansion section, the shear force is relatively reduced; different pour point depressant concentrations change the internal structure and surface properties of the wax crystals, affecting their ability to resist shear force. The temperature, pressure of the crude oil and the initial size and morphology of the wax crystals also affect the wax crystal crushing process. As the temperature rises, the flexibility of the wax crystals increases and they are not easy to break. Pressure changes will change the interaction between the wax crystals and the crude oil molecules, indirectly affecting the crushing process; the strain gauge channel is used to measure the resistance to the movement of the wax crystals, and then a microscope imaging system and a high-speed camera are used in combination with an image processing algorithm to obtain the diameter, fractal dimension, number, and roundness of the wax crystals, and analyze the influence of the pour point depressant concentration on the wax crystal crushing process.

8. A method for screening and optimizing using the microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 6, characterized in that: The dynamic parameters of pressure, viscosity and wax crystals are collected in real time through the microfluidic chip. The high-dimensional features within the 60-second time window are extracted through the long short-term memory network LSTM, and the evolution law of the wax crystal structure under the high-dimensional feature temperature gradient is constructed. The state vector representation system is input into the PPO reinforcement learning algorithm, and decisions are made in the three-dimensional continuous action space. The strategy optimization is driven by the weighted reward function of freezing point depression, viscosity change rate, fractal dimension change rate and cost control. The weight ratio of freezing point depression, viscosity change rate, fractal dimension change rate and cost control is 10:6:3:1, forming a closed-loop intelligent control mechanism of temporal perception-decision optimization. Through the deep collaboration of the long short-term memory network LSTM and the PPO reinforcement learning algorithm, efficient pour point depressant screening and optimization from temporal dynamic modeling to intelligent decision optimization is achieved.

9. The method for screening and optimizing using the microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 8, characterized in that: Step 1: Model pre-fitting experiment: Before conducting the microfluidic chip experiment, a small sample conventional pour point measurement experiment is first performed to systematically obtain the pour point depression of the oil sample after adding pour point depressant, and establish basic data on the correlation between pour point depression and normalized index S; In the second step, efficient mixing of pour point depressants is achieved in the microfluidic chip through multi-channel coordinated control: Pour point depressant samples of different concentrations are introduced through injection channels of corresponding inner diameters. After preliminary mixing with crude oil in the storage chamber, they enter the micro-mixing chamber. The flow rate of each injection channel is precisely controlled by the microfluidic valve group at 0.1-5μL / min to ensure that different samples participate in the reaction at a stable flow rate ratio. The microstructure of the micro-mixing chamber induces laminar flow disturbance, allowing the crude oil and pour point depressant to achieve full contact in a short time. Step 3: Data Acquisition: The data acquisition process of the microfluidic chip uses multi-sensor fusion technology to achieve comprehensive parameter measurement. The temperature control system simulates the crude oil cooling and gelation process at a rate of 0.1°C / min. The micro-piezoelectric sensors integrated on both sides monitor the channel pressure fluctuations in real time. The micro-flow rate sensor and shear force sensor manufactured based on MEMS technology synchronously obtain fluid dynamic parameters, and the Hagen-Poiseuille equation is used to calculate the change in fluid viscosity. The microscopic imaging system continuously captures the dynamics of the wax precipitation chamber at a rate of 60 frames per second, capturing the morphological evolution, motion trajectory, and fragmentation behavior of the wax crystals during precipitation. The dynamic parameters of the wax crystals are collected and assigned different weights to construct the normalized parameter S. Based on the solidification point measured in step 1, the quantitative relationship between it and the normalized parameters was determined; after all the image data were 3D reconstructed and voxelized, the wax crystal volume fraction, number density and spatial distribution parameters were extracted by the characteristic statistical algorithm. The wax crystal volume fraction was 0.1%-15%, the number density was 10 2 -10 4 Pieces / mm 3 , spatial distribution parameters: skewness 0.5-1.2, kurtosis 2.8-4.5, providing a complete data set for quantitative analysis of the effect of pour point depressants on the three-dimensional growth characteristics of wax crystals; Step 4: Data analysis and formula optimization: Normalize the fluid mechanics parameters and wax crystal time series feature data. The fluid mechanics parameters include pressure, viscosity, and flow rate. The wax crystal time series features include morphological parameters, kinematic parameters, dynamic parameters, and voxelized parameters. Arrange the sensor data by time to form sequence data of time steps. Use 60 time cloths as the input units of LSTM. Next, extract the time series features of LSTM. Train the wax crystal evolution data with different pour point depressant ratios by designing 128-dimensional vectors of the input layer and output layer. The input layer includes pressure, viscosity, wax crystal size, fractal dimension, and roundness. The time series feature vectors output from TM are imported into the PPO algorithm to construct the state space. The pour point depression, viscosity reduction rate, wax crystal fractal dimension change rate and cost penalty are selected for reward function design. The pour point depressant concentration optimization scheme and reward value are obtained by calculation. The relevant parameters are then input into the long short-term memory network LSTM to construct a pour point depressant effect evaluation model. Finally, the PPO algorithm is used to perform formula optimization search. The pour point depressant content is used as the gene code. New formula combinations are continuously generated through selection, crossover and mutation. The fitness of each combination is calculated according to the evaluation model. After multiple generations of evolutionary iterations, it gradually converges to the optimal or near-optimal pour point depressant formula.

10. The method for screening and optimizing using the microfluidic chip intelligent screening and optimization device for pour point depressants with integrated multi-parameter detection according to claim 9, characterized in that: The index data after normalization in step 4 is measured experimentally to determine the quantitative relationship between the freezing point and the normalization parameter. The specific formula is as follows: S=a×(a i D avg +a j N+a k σ)+b×(b i V+b j A+b k ω)+c×(c i D+c j F+c k N ad +c m C) ΔT g =AS p +B Where: a+b+c=1a i +a j +a k =1b i +b j +b k =1c i +c j +c k +c l =1 Where: S is the normalized index, where a n 、b n 、c n is the weight of each parameter; ΔT g is the freezing point depression, ℃, where A, B, and P are obtained by fitting experimental data; D avg is the average size of wax crystals, μm; N is the number of wax crystals per unit volume; σ is the dispersion of wax crystal size; V is the linear velocity of wax crystals, mm / s; A is the acceleration of wax crystals, mm / s 2 ; ω is the angular velocity of the wax crystal, rad / s; D is the wax crystal crushing diameter, μm; F is the fractal dimension of the wax crystal; N ad is the increase in the number of wax crystals; C is the attenuation of the roundness of the wax crystals.

Images