Intelligent determination method and system for reaction end point of polyester modified organic amine

Abstract

The present application relates to the technical field of chemical process determination, and discloses an intelligent determination method and system for the reaction endpoint of polyester-modified organic amine, which comprises the following steps: obtaining an oilfield application working condition to construct a basic database, building a two-dimensional performance plane coordinate system, dividing a rhombic performance target area, and determining a performance point with the fastest value improvement; constructing a process performance potential field model based on the performance point to obtain performance potential energy, defining a reaction trajectory and calculating a trajectory potential barrier, and determining an optimal process state vector; constructing an endpoint phase coordinate system and a convergence area, generating an intention state set, and generating a reaction termination instruction through differential judgment. The present application realizes accurate determination of the reaction endpoint of polyester-modified organic amine, solves the problems of low efficiency and poor data reliability of traditional methods, reduces the comprehensive cost of reaction, guarantees product performance and application safety, and improves production stability and economy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of chemical process determination, more particularly, the present application relates to an intelligent determination method and system for the reaction endpoint of polyester-modified organic amine. BACKGROUND

[0002] When the polyester-modified organic amine reaction is applied in the oil field, due to the differences in working conditions such as reservoir temperature, formation water salinity, and rock type, the traditional method has significant technical pain points. The gelation time and dissolution rate measurement efficiency are low, the data reliability is poor, and the performance optimization target is ambiguous, making it difficult to adapt to different oil field conditions. The gap between real-time performance and target performance during the reaction process is difficult to measure, the optimal path of comprehensive cost under multiple reaction trajectories is difficult to screen, and the future optimal process state is difficult to predict. At the same time, the reaction endpoint control faces the problems of difficult visualization of multi-dimensional state, difficult identification of trajectory deviation, and difficult accurate determination, which easily leads to substandard product performance, increased application risk, or waste of production resources. How to realize the precise quantification of the performance and risk of the polyester-modified organic amine reaction, the optimal control of comprehensive cost, and the accurate determination of the reaction endpoint has become a technical difficulty that the industry urgently needs to solve.

[0003] In the prior art, a kind of automatic potential titration method and system based on machine learning are disclosed in Chinese patent with authorization announcement No.CN114414648B. The method obtains the concentration of titrant by identifying the titration tube label, combines the initial potential of the solution to be measured, and uses historical titration data to train artificial neural network and logistic regression model to predict the titration endpoint. The titration process is terminated automatically and the ion concentration is calculated by real-time collection of potential and dosage data and determination of whether the termination condition is reached, thereby improving the accuracy and automation level of endpoint determination. This scheme belongs to the endpoint determination idea assisted by computer and machine learning, and is representative of the closed loop of "endpoint prediction-real-time verification-automatic shutdown". At the same time, a process endpoint detection enabled by artificial intelligence is disclosed in Chinese patent with authorization announcement No.CN111915551B. The imaging device obtains the sample surface image, and the pattern recognition and learning algorithm containing convolutional neural network is used to extract local features to determine whether the process endpoint is reached. If not, continue material removal or finishing, and issue a termination instruction when the endpoint is reached to realize automatic endpoint detection for charged particle microscope sample preparation and other scenarios. This scheme emphasizes image-based machine learning endpoint identification and process linkage control.

[0004] However, the above two prior arts provide effective ideas in end-point intelligent judgment, but still have difficulty in solving the problems of multi-index coordinated quantification and process navigation of polyester-modified organic amine reaction in oilfield application scenarios: the Chinese patent with the authorization announcement number CN114414648B focuses on single-point termination judgment of single-type physicochemical signal and single potential titration, does not establish a multi-source data fusion framework of residual organic amine monomer concentration, dissolution rate, gelation time and comprehensive application risk index, does not construct a two-dimensional performance plane coordinate system and a rhombus performance target area, and lacks modeling of performance points and performance potential in the direction of simultaneous decrease of time and risk; the Chinese patent with the authorization announcement number CN111915551B mainly uses image feature-driven end-point recognition, does not combine oilfield working condition library and historical production batches for similar working condition retrieval and comprehensive measurement of trajectory potential barrier, and does not introduce the optimal process state vector, end-point phase coordinate system and safety navigation mechanism of convergence area solved by particle swarm. Both of them do not form a closed-loop control chain from performance potential, trajectory potential barrier, optimal process state, intention state set to differentiated stopping, and it is difficult to realize precise, low-risk and navigable end-point intelligent judgment of the modified reaction under complex reservoir conditions. SUMMARY

[0005] The present application is applicable to various oilfield polyester-modified organic amine reaction scenarios, such as carbonate rock and sandstone reservoir working conditions, and can meet the reaction end-point judgment requirements under different reservoir temperatures and formation water salinities; through the basic database, two-dimensional performance plane coordinate system and rhombus performance target area, the present application realizes rapid matching of oilfield working conditions and precise quantification of reaction performance risk, and avoids the high-risk situation of time and risk simultaneously reaching the tolerance limit of traditional rectangular target area; the process performance potential field model combines with deep neural network to associate real-time process state vector with performance point coordinates, quantifies the gap between real-time performance and target, calculates the trajectory potential barrier by geometric mean method of energy consumption emission cost and process abnormal cost, and selects the reaction trajectory with optimal comprehensive cost; the end-point phase coordinate system is combined with differentiated judgment to generate precise reaction termination instructions, ensuring that the reaction is terminated when it is close to the optimal process state vector and dynamically tends to be stationary, improving product quality stability, and reducing production energy consumption and abnormal loss.

[0006] To achieve the above object, the present application provides the following technical scheme:

[0007] The intelligent judgment method for the end-point of polyester-modified organic amine reaction comprises the following steps:

[0008] An oilfield application working condition is obtained to construct a basic database simulating different working conditions; a two-dimensional performance plane coordinate system is constructed according to the basic database, and a rhombus performance target area dedicated to the working condition is divided on the two-dimensional performance plane coordinate system; a performance point with the fastest value increase is determined in the rhombus performance target area;

[0009] A performance potential field model is constructed based on the performance point to obtain performance potential energy for measuring the distance from the target; a reaction trajectory is defined, and a trajectory potential barrier representing the comprehensive cost is calculated for the reaction trajectory; the optimal process state vector is obtained by combining the performance potential energy and the trajectory potential barrier;

[0010] A two-dimensional end-point phase coordinate system is constructed based on the optimal process state vector, and a convergence area for safe navigation is marked for the reaction trajectory in the end-point phase coordinate system; intention judgment is performed on the convergence area to generate an intention state set; and a differentiated judgment is performed on the intention state set to obtain a reaction termination instruction for judging accurate parking.

[0011] Further, the construction method of the basic database comprises:

[0012] The oilfield application working conditions include reservoir basic parameters, rock types, and polymer types expected to be used in cooperation;

[0013] The rock types are extracted from the oilfield application working conditions, experiments are performed based on the chemical stability differences of rock components, the dissolution rate is detected, and the actual working conditions are simulated by the oilfield application working conditions to obtain the gelation time and the residual organic amine monomer concentration;

[0014] Experiments are performed on the simulated working conditions of different historical production batches to obtain the residual organic amine monomer concentration, the dissolution rate, and the gelation time corresponding to different historical production batches, and the residual organic amine monomer concentration, the dissolution rate, and the gelation time corresponding to different historical production batches are integrated to obtain the basic database.

[0015] Further, the method for obtaining the rhombic performance target area comprises:

[0016] The residual organic amine monomer concentration and the dissolution rate are combined to calculate the comprehensive application risk index by the comprehensive application risk index formula;

[0017] The oilfield application working conditions are matched and searched by the basic database to obtain N historical production batches most similar to the oilfield application working conditions, the gelation time and the comprehensive application risk index in the N historical production batches are extracted, and the arithmetic mean of the N gelation times and the arithmetic mean of the comprehensive application risk indexes are calculated; the arithmetic mean of the gelation time and the arithmetic mean of the comprehensive application risk index are taken as coordinate points, i.e., the target center, the arithmetic mean of the gelation time corresponds to the horizontal axis value, and the arithmetic mean of the comprehensive application risk index corresponds to the vertical axis value;

[0018] A general procedure library of oilfield operations is obtained, an acceptable time tolerance and a risk tolerance are determined, and four vertex coordinates, i.e., an upper vertex, a lower vertex, a right vertex, and a left vertex, are marked in the two-dimensional performance plane coordinate system according to the time tolerance and the risk tolerance;

[0019] In the two-dimensional performance plane coordinate system, the target center is taken as the geometric center, and the upper vertex, the lower vertex, the right vertex and the left vertex are sequentially connected to form a closed diamond region, i.e. a diamond performance target region.

[0020] Further, the performance point acquisition method comprises:

[0021] A local coordinate system is defined with the target center as the origin in the diamond performance target region, and the local coordinate system is a plane rectangular coordinate system.

[0022] A golden quadrant is defined as a connection region of the local coordinate system horizontal axis and the local coordinate system vertical axis and a diamond performance target region in the third quadrant of the local coordinate system, and the golden quadrant is a right triangle region geometrically formed by the target center coordinates, the left vertex and the lower vertex.

[0023] A ray is emitted along a direction representing time and risk synchronization optimal descent from the target center, and the intersection point of the ray and the connection line between the left vertex and the lower vertex is the performance point.

[0024] Further, the trajectory barrier acquisition method comprises:

[0025] The combustion consumption rate of equipment providing a heat source for the reaction kettle is monitored in real time;

[0026] The nitrogen oxide emission rate and the soot emission rate are calculated based on the combustion consumption rate through a pollutant emission factor model, the nitrogen oxide emission rate and the soot emission rate are combined to obtain a total pollutant equivalent, and the total pollutant equivalent is time-integrated along a reaction trajectory to obtain an energy consumption emission cost.

[0027] A sensor is deployed at a leakage point of the reaction kettle to obtain instantaneous concentrations of ammonia, hydrogen chloride and non-methane total hydrocarbon, and the instantaneous concentrations of ammonia, hydrogen chloride and non-methane total hydrocarbon are weighted and summed to obtain a process abnormality cost.

[0028] The process abnormality cost and the energy consumption emission cost are combined and calculated by a geometric mean method to obtain the trajectory barrier.

[0029] Further, the optimal process state vector acquisition method comprises:

[0030] A search space is defined through a basic database, and a particle swarm containing H particles is randomly generated in the search space.

[0031] Each particle in the particle swarm is calculated for a particle performance potential energy through a process performance potential field model, and a particle trajectory barrier is calculated through the trajectory barrier acquisition method.

[0032] The fitness of each particle in the particle group is calculated by a particle performance potential and a particle trajectory barrier, the individual historical best position of each particle in the search space is set by the fitness, the individual historical best position refers to the position corresponding to the minimum fitness of each particle in the search space, the individual historical best positions of all particles are traversed, and the particle with the minimum fitness is set as the initial global best position of the whole particle group;

[0033] The particles in the search space are subjected to cyclic iteration, an update operation is performed on each particle after each iteration, the fitness of each particle after iteration is calculated, if the fitness of the particle after iteration is less than the fitness of the particle before iteration, the position of the particle in the search space after iteration is marked as the individual historical best position, and the fitness of all particles after iteration is also traversed, and the particle with the minimum fitness is set as the global best position of the whole particle group;

[0034] If the global best position remains unchanged for a plurality of iterations, the particle corresponding to the global best position is taken as the optimal process state vector.

[0035] Further, the method for obtaining the convergence region comprises:

[0036] Based on the optimal process state vector, a state distance and a trajectory deviation degree are calculated;

[0037] A terminal phase coordinate system is constructed for the optimal process state vector by the state distance and the trajectory deviation degree, the terminal phase coordinate system takes the state distance as the horizontal axis and the trajectory deviation degree as the vertical axis;

[0038] In the terminal phase coordinate system, a region surrounded by two hyperbolas, i.e. the convergence region, is defined.

[0039] Further, the method for obtaining the intention state set comprises:

[0040] Each real-time process state vector is obtained, a state distance and a trajectory deviation degree of each real-time process state vector are calculated, and the state distance and the trajectory deviation degree of the real-time process state vector are taken as the coordinates of the real-time process state vector, i.e. the unique coordinates in the terminal phase coordinate system, which are marked as state points, the state distance and the trajectory deviation degree of the real-time process state vector are the horizontal axis coordinates and the vertical axis coordinates of the state points, respectively;

[0041] Based on the hyperbolic inequality derived from the convergence region, the state points are judged to obtain strong intention states and weak intention states;

[0042] The strong intention states and the weak intention states are combined to obtain the intention state set.

[0043] Further, the reaction termination instruction acquisition method comprises the following steps:

[0044] In the terminal phase coordinate system, a square region with the origin as the center and the side length of 2F is defined as the parking area, wherein F represents the minimum value of the final control accuracy, and the value is determined according to the final error range allowed by each component in the optimal process state vector;

[0045] The differentiation judgment includes the area entry judgment and the low-speed judgment, the weak intention state in the intention state set is subjected to the area entry judgment, if the horizontal axis coordinate of the state point corresponding to the weak intention state is less than F, it is judged that the area entry judgment condition is met; meanwhile, the weak intention state meeting the area entry judgment condition is the prerequisite condition for executing the low-speed judgment, the length change rate of the real-time process state vector corresponding to the state point is acquired, if the absolute value of the length change rate of the real-time process state vector corresponding to the state point is less than the preset idling speed threshold, it is judged that the low-speed judgment condition is met;

[0046] If the area entry judgment condition and the low-speed judgment condition are met at the same time, the final reaction termination instruction is generated.

[0047] The intelligent judgment system for the reaction end point of the polyester modified organic amine is used to realize the intelligent judgment method for the reaction end point of the polyester modified organic amine, and the system comprises:

[0048] The performance starting point module is used to acquire the oil field application working condition, construct a basic database for simulating different working conditions, construct a visual two-dimensional performance plane coordinate system according to the basic database, and divide a working condition exclusive diamond performance target area on the two-dimensional performance plane coordinate system; a performance point with the fastest value improvement is determined in the diamond performance target area;

[0049] The optimal end point module is used to construct a process performance potential field model based on the performance point, obtain a performance potential energy for measuring the distance degree from the target, define a reaction trajectory, and calculate a trajectory potential barrier representing the comprehensive cost for the reaction trajectory; the optimal process state vector is obtained by combining the performance potential energy and the trajectory potential barrier;

[0050] The process path module is used to construct a two-dimensional terminal phase coordinate system based on the optimal process state vector, and demarcate a convergence area for safe navigation in the terminal phase coordinate system for the reaction trajectory; intention judgment is performed on the convergence area, and an intention state set is generated; differentiation judgment is performed on the intention state set, and a reaction termination instruction for judging accurate parking is obtained.

[0051] Compared with the prior art, the present application has the following advantages:

[0052] The application realizes the rapid matching of oilfield working conditions and the accurate quantification of reaction performance and risk through the basic database, the two-dimensional performance plane coordinate system and the rhombus performance target area, solves the pain points of poor working condition adaptability and fuzzy performance optimization target in the traditional scheme; the process performance potential field model combines with the trajectory potential barrier calculation, converts the fuzzy monitoring of the reaction process in the traditional scheme into the quantitative analysis of performance potential, energy consumption and emission cost, process abnormal cost, accurately selects the reaction trajectory with the optimal comprehensive cost, and reduces the production energy consumption and abnormal loss; the end point phase coordinate system is combined with the differential judgment, converts the experience-based judgment of the reaction end point in the traditional scheme into the accurate control of the state distance, the trajectory deviation degree and the length change rate of the process state vector, ensures that the reaction is terminated when the optimal process state vector is approached, and guarantees the product quality stability, and meets the intelligent and low-cost production demand of the polyether modified organic amine reaction under the complex working conditions of the oilfield. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor on the basis of these drawings.

[0054] Figure 1 The method flow chart of the intelligent judgment method of the polyether modified organic amine reaction end point provided by the embodiment of the present application;

[0055] Figure 2 The spatial relationship diagram of the rhombus performance target area in the two-dimensional performance plane coordinate system;

[0056] Figure 3 The position relationship diagram of the performance point in the rhombus performance target area and the local coordinate system;

[0057] Figure 4 The functional module diagram of the intelligent judgment system of the polyether modified organic amine reaction end point provided by the embodiment of the present application. DETAILED DESCRIPTION

[0058] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0059] Embodiment 1:

[0060] Please refer to Figure 1As shown, the embodiment provides an intelligent determination method for the reaction end point of polyester modified organic amine, which comprises the following steps:

[0061] In step S10, the oilfield application conditions are obtained to construct a basic database simulating different working conditions; a two-dimensional performance plane coordinate system is constructed according to the basic database, and a rhombus performance target area dedicated to the working conditions is divided on the two-dimensional performance plane coordinate system; and a performance point with the fastest value improvement is determined in the rhombus performance target area.

[0062] Further, step S10 comprises:

[0063] In step S11, the oilfield application conditions are obtained, targeted experiments are performed on the oilfield application conditions, the gelation time and the dissolution rate reflecting the risk index are obtained, and the basic database is constructed according to the gelation time and the dissolution rate.

[0064] The oilfield application conditions include reservoir basic parameters, rock types, and expected polymer types to be used, for example, polyacrylamide, wherein the reservoir basic parameters include formation temperature and formation water salinity; the formation temperature represents the temperature of the rock and the fluid in the pore reaching thermal equilibrium for a long time at the underground depth of the reservoir, and the formation water salinity represents the total mass concentration of all inorganic salt substances dissolved in the naturally occurring underground water in the reservoir pore, such as sodium chloride, calcium chloride, and magnesium sulfate; the rock type represents the rock type of the reservoir site, which is used to determine the risk and adaptation direction of the product and the reservoir, and the expected polymer type to be used represents the polymer specified based on the development goal of the reservoir itself and required to be used with the polyester modified organic amine gel, for example, polyacrylamide; if it is required to use the polyester modified organic amine gel with the polymer, the finished product is a pre-mixed finished product of the polyester modified organic amine gel and the polymer in a specified ratio under simulated reservoir temperature conditions; if it is not required, the finished product is a standard single polyester modified organic amine gel finished product.

[0065] The rock type is extracted from the oilfield application conditions, experiments are performed based on the chemical stability difference of the rock composition, if the rock type is carbonate rock, such as limestone and dolomite, experiment one is performed to detect the dissolution rate; if the rock type is sandstone, such as non-carbonate rock with quartz and feldspar as the main components, no experiment is needed; at the same time, experiment two is performed to detect the gelation time of the product containing the polyester modified organic amine gel in the reservoir, and only the carbonate rock type rock is experimented because the core component of the carbonate rock is carbonate mineral, which has poor chemical stability and is easy to react with acidic substances, and the reaction will directly cause the dissolution of the rock structure, resulting in dissolution pores or expansion of the original cracks; and the mineral chemical stability of sandstone is very strong and almost does not react with acid and alkali under normal temperature and pressure.

[0066] Experiment 1: Use the core sample consistent with the rock type, and strictly simulate the provided reservoir temperature, formation water salinity working condition parameters; pretreat the core sample, remove the surface impurities, then measure the initial core calcium carbonate total content A; place the core sample in the reaction container simulating the working condition, add the product to be detected, and stand for a preset time, then take out the core sample after the preset time, measure its remaining mass and remaining calcium carbonate content B, obtain the mass of the calcium carbonate eroded in the core, and calculate the calcium carbonate erosion capacity, i.e. the erosion rate, by the mass of the calcium carbonate eroded in the core and the initial core calcium carbonate total mass, . Wherein, the product to be detected refers to the gel functional product that has completed production or experiment and is finally planned to be applied to the target reservoir, and the preset time is determined according to the actual contact period of the product and the rock in the reservoir development.

[0067] Experiment 2: Determine the adaptive sample through the above compounded product or single polyester modified organic amine gel product, that is, the adaptive sample is one of the compounded product or single polyester modified organic amine gel product, set a constant temperature reaction container for reducing the real contact environment through the reservoir temperature and formation water salinity, put the adaptive sample into a transparent sealed bottle, and add an electrolyte solution to form a reaction system simulating the reservoir environment, and put the reaction system into the constant temperature reaction container, and start timing at the same time. Due to the uncertainty of the state change in the gelation process, manual timing is required. For example, when the adaptive sample is still liquid at the initial stage, it is turned over once about 15 minutes, when the sample starts to become sticky, it is turned over once every 5 minutes, and when the reaction system first appears in the state of no flow and no dripping when inverted by 180°, the timer is paused and the time is recorded. This time is the initial gelation time, and a small amount of adaptive sample is taken from the sealed bottle for detection by a rheometer to measure the storage modulus of the adaptive sample. If the storage modulus of the adaptive sample is greater than or equal to the gel threshold value, it is determined to be an effective gel with practical application value. If the storage modulus of the adaptive sample is less than the gel threshold value, the remaining sample is put back into the sealed bottle for re-turning and re-recording the initial gelation time of the adaptive sample, and the storage modulus of the adaptive sample is judged until the effective gel is determined to be effective. When it is determined to be an effective gel, the timer is stopped, and the value of the timer at this time is the gelation time. The storage modulus directly reflects the elastic mechanical strength of the gel, and the gel threshold value is set based on the sealing requirements of the oil field to ensure that the gel can resist the flushing of the formation water. At the same time, there is a certain small error in the experiment, for example, the time used for detection by the rheometer after the adaptive sample is taken out. However, the gelation time is usually several hours, and the proportion of the small error time is extremely small, so the error can be ignored.

[0068] The residual organic amine monomer concentration of the same experimental batch is obtained, the residual organic amine monomer concentration is obtained by detecting the residual organic amine monomer concentration through an online mass spectrometer arranged at a sampling port of a product storage tank, the residual organic amine monomer concentration represents the content of free organic amine raw materials that have not completely participated in the reaction, and the residual organic amine monomer concentration is obtained to facilitate guarantee of personnel health and safety, avoidance of harmful substance contact risk, ensuring product performance, and avoidance of residual monomer interference with gel function.

[0069] A basic database is constructed based on the residual organic amine monomer concentration, the gelation time, the dissolution rate, and the oilfield application condition, the basic database stores performance test data of products of historical production batches under different simulated conditions, is constructed from experimental data of a plurality of historical production batches, and stores a set of associated data of reservoir condition parameters, product performance indexes, and application risk indexes in a structured manner, wherein the reservoir condition parameters include all data of the oilfield application condition, the product performance index refers to the gelation time, the application risk index includes the dissolution rate and the residual organic amine monomer concentration, and the reservoir condition parameters, the product performance index, and the application risk index of each test batch are associated with each other, that is, the reservoir condition parameters, the product performance index, and the application risk index form a group of data. The basic database covers most land oilfield temperature ranges for formation temperature, for example, the formation temperature is selected to be 50-180°C; covers medium-high salt oil reservoirs for formation water salinity, for example, the formation water salinity is 5000-20000; and selects carbonate rocks and conventional sandstone reservoir types to ensure coverage of most underground rocks of oil reservoirs for rock type. For the time range, the data timeliness needs to be ensured to avoid deviation of early formula data from current products, for example, production or experimental data of the last three years is selected; and the basic database is constructed to quickly match the condition, to quickly determine the reference range of the gelation time and the application risk index by searching for similar conditions in the basic database after the oilfield application condition is obtained, to avoid time cost of starting from zero, to verify the rationality of the current test result through historical data to ensure the reliability of the test data, and to dynamically optimize the product formula.

[0070] In step S12, a two-dimensional performance plane coordinate system is constructed according to the basic database, and a rhombus performance target area dedicated to the condition is divided on the two-dimensional performance plane coordinate system.

[0071] The two-dimensional performance plane coordinate system is a plane rectangular coordinate system with two core performance indexes as axes, the X axis corresponds to the gelation time, and the Y axis corresponds to the comprehensive application risk index R. The comprehensive application risk index is obtained by converting the two potential risks of the dissolution rate C and the residual organic amine monomer concentration D into comparable values, and is used to quantify the comprehensive risk degree of the product in the oil reservoir application; R is calculated through the comprehensive application risk index formula, and the comprehensive application risk index formula is as follows: wherein max(·) represents taking the maximum value, represents the upper limit of the dissolution rate standard, which is set according to the enterprise internal control standard, and an example is that the upper limit of the dissolution rate standard is set to 1.0%, represents the upper limit of the residual organic amine monomer concentration standard, which is set according to the occupational health and safety standard, and an example is that the upper limit of the residual organic amine monomer concentration standard is set to 5.0%. The maximum function is used in the comprehensive application risk index formula to select the larger value between the ratio of the dissolution rate to the upper limit of the dissolution rate standard and the ratio of the residual organic amine monomer concentration to the upper limit of the residual organic amine monomer concentration standard, which avoids the subjectivity and controversy of artificially setting weights, makes the risk assessment more objective, and at the same time accurately locks the most dangerous risk item, avoiding that secondary risks cover up the core problems; the square operation is used to amplify the influence of the larger ratio and strengthen the risk sensitivity.

[0072] After obtaining the oilfield application working condition, the nearest neighbor algorithm is used in the basic database to perform matching retrieval on the oilfield application working condition, and N historical production batches most similar to the oilfield application working condition are found out. The nearest neighbor algorithm for matching retrieval means that the core parameters of the oilfield application working condition are preprocessed, the preprocessed core parameters are subjected to similarity calculation, the comprehensive weighted distance with each historical batch is calculated, the smaller the distance, the higher the similarity, and the N historical production batches with the highest similarity are screened out. For example, the continuous parameters in the core parameters of the oilfield application working condition, such as formation temperature and formation water salinity, are normalized to eliminate dimensions, and the classification parameters, such as rock type and expected polymer model, are labeled and coded to obtain the preprocessed core parameters. The weighted distance method is used to calculate the comprehensive weighted distance between the target working condition and each historical batch. The weights of the formation temperature and the formation water salinity are each 0.3, and the weights of the rock type and the polymer model are each 0.2 for weighted calculation. The gelation time and the comprehensive application risk index in the N historical production batches are extracted. The arithmetic mean value M1 of the N gelation times and the arithmetic mean value M2 of the comprehensive application risk indexes are calculated by the arithmetic mean value formula. M1 and M2 form a coordinate point (M1, M2) in a two-dimensional performance plane coordinate system. The target represents the most ideal balance point of performance and risk under the oilfield application working condition according to historical experience. The selection of N should avoid too small to cause accidental fluctuations of a single or a few historical cases, and at the same time avoid too large to increase efficiency and cost. For example, N is 5;

[0073] The general procedure library of oilfield operation is called to determine an acceptable time tolerance and risk tolerance ; wherein the job general procedure library is a safety operation procedure or an operation manual set by each oilfield according to its own working condition; the time tolerance indicates that the gelation time is allowed to deviate from the target corresponding gelation time, i.e., the maximum reasonable range of M1, and the risk tolerance indicates that the comprehensive application risk index is allowed to deviate from the target corresponding comprehensive application risk index, i.e., the maximum reasonable range of M2; the purpose is to adapt to the actual scene and solve the contradiction between the theoretical optimum and the engineering landing. Taking the target as the geometric center, four vertex coordinates, i.e., an upper vertex P1, a lower vertex P2, a right vertex P3 and a left vertex P4, are calculated and marked in the two-dimensional performance plane coordinate system. P1 indicates that the time effectiveness is kept unchanged, and the risk is pushed to the upper limit of the acceptable range, ; P2 indicates that the time effectiveness is kept unchanged, and the ideal lower boundary of the risk is determined, ; P3 indicates that the risk level is kept unchanged, and the time is delayed to the upper limit of the acceptable range, ; P4 indicates that the risk level is kept unchanged, and the ideal boundary of the time advancement is determined, .

[0074] P1, P2, P3 and P4 are sequentially connected to form a closed diamond-shaped area, i.e., a diamond-shaped performance target area, as shown in Figure 2 ; the diamond-shaped performance target area clearly indicates that the tolerances of the time and the risk in the two dimensions are independent of each other, avoiding the extreme case that the time and the risk reach the tolerance limit at the same time in the four corners of the traditional rectangular target area, which is usually a high-risk area that needs to be avoided in actual production.

[0075] In step S13, a golden quadrant embodying the double optimization of performance and risk is screened out in the diamond-shaped performance target area, and a performance point with the fastest value improvement is determined in the golden quadrant.

[0076] A local coordinate system is set in the diamond-shaped performance target area, and the local coordinate system is also a plane rectangular coordinate system. The local coordinate system takes the target as the origin, the horizontal axis of the local coordinate system is defined as a relative time axis, the relative time axis is parallel to the X axis of the two-dimensional performance plane coordinate system, and the coordinate value on the relative time axis represents the deviation of the gelation time relative to the target corresponding gelation time M1. The vertical axis of the local coordinate system is defined as a relative risk index axis, the relative risk index axis is parallel to the Y axis of the two-dimensional performance plane coordinate system, and the coordinate value on the relative risk index axis represents the deviation amount of the comprehensive application risk index relative to the target corresponding comprehensive application risk index M2.

[0077] The region formed by the local coordinate system and the third quadrant of the rhombus performance target area, i.e. the region formed by the connection line of the horizontal axis and the vertical axis of the local coordinate system and the third quadrant of the rhombus performance target area in the local coordinate system, the coordinate points in the region correspond to negative gelation time and comprehensive application risk index, which means that the product has shorter gelation time than the target, and the comprehensive risk is also lower than the target, indicating that the product improves the safety while improving the performance. The region is defined as the golden quadrant. According to the definition of the rhombus performance target area, it can be determined that the golden quadrant is a right triangle region surrounded by the target center coordinates, the left vertex and the lower vertex in geometry. The golden quadrant solves the traditional fuzzy pursuit of the shortest time and the lowest risk, and lacks a clear and executable single target. In the golden quadrant, a ray along the direction of synchronous optimal decline of time and risk is emitted from the target center, and the intersection point between the ray and the connection line between the left vertex and the lower vertex is defined as the performance point, as shown in Figure 3 The performance point coordinates are (optimal gelation time, optimal comprehensive application risk index). The direction of the connection line between the target center and the intersection point is not subjective or fixed, but is objectively determined according to the principle of equal proportion value decline. Specifically: the coordinates of any point on the golden quadrant are represented as (u, v), u represents the gelation time of the coordinate point on the golden quadrant, and v represents the comprehensive application risk index of the coordinate point on the golden quadrant. According to the time tolerance Risk tolerance An equal proportion value equation is established according to the principle of equal proportion value decline, i.e. Then the equal proportion value equation is The ray is a line segment from the target center to the intersection of the equal proportion value equation and the connection line between the left vertex and the lower vertex. The performance point converts the fuzzy optimization concept into an accurate and repeatable geometric solving process, and finds the performance point that can be reached within the tolerance range, completely eliminating the subjectivity and uncertainty brought by artificial selection or random selection.

[0078] Step S10 solves the technical problems of significant influence of different rock types on the reaction performance of polyether modified organic amine, low efficiency of gelation time and dissolution rate measurement, poor data reliability, and fuzzy performance optimization target in traditional methods by constructing a basic database, building a two-dimensional performance plane coordinate system, dividing a diamond-shaped performance target area, and screening a golden quadrant. The precise quantification of the reaction performance and risk of polyether modified organic amine, the rapid matching of the adaptive working conditions, and the clear focus of the optimization target are realized. The basic database can quickly match the working conditions and verify the reasonableness of the current detection results, avoiding the time cost of starting from zero detection. The two-dimensional performance plane coordinate system unifies the separated performance and risk indicators into a comparable two-dimensional geometric space. The diamond-shaped performance target area avoids the high-risk situation of time and risk simultaneously reaching the tolerance limit in the traditional rectangular target area. The golden quadrant converts the fuzzy short time and low risk optimization requirements into a clear geometric area. The performance point determines the performance point within the tolerance range through the principle of equal proportion value decrease, excluding the subjectivity of manual selection.

[0079] Step S20, based on the performance point, constructs a process performance potential field model to obtain a performance potential that measures the distance from the target; defines a reaction trajectory and calculates a trajectory potential barrier representing the comprehensive cost for the reaction trajectory; and combines the performance potential and the trajectory potential barrier to obtain an optimal process state vector.

[0080] Further, step S20 includes:

[0081] Step S21, based on the performance point, constructs a process performance potential field model to obtain a performance potential that measures the distance from the target.

[0082] The sensor collects the instantaneous values of the conventional parameters in the reaction kettle in real time according to the preset sampling period. The conventional parameters include temperature, pH value, and viscosity. The preset sampling period is set by the characteristics of the conventional parameters to ensure the capture of effective data. For example, it is set to 10 seconds. The temperature is measured by a thermocouple inserted into the reaction kettle, which reflects the energy level in the reaction. The pH value is measured by an online pH meter, which is the key to controlling the reaction equilibrium. The viscosity is measured by an online rotary viscometer, and the change in viscosity directly reflects the degree of polymer growth and crosslinking. The first-order derivative of the conventional parameters with respect to time is calculated by the derivative formula to obtain a conventional parameter rate set, which includes temperature change rate, pH value change rate, and viscosity change rate. For example, the temperature change rate is calculated by the difference between the current instantaneous temperature value and the temperature instantaneous value at the last sampling time. The pH value change rate and the viscosity change rate are obtained in the same way.

[0083] A sensor is arranged at a leakage point of the reactor, and the sensor detects and quantifies the instantaneous concentration of a specific substance escaping into the surrounding local air in real time at a sampling period corresponding to the acquisition of conventional parameters. The leakage point is selected as a position where microscale and unorganized escape is most likely to occur in the reactor, such as inside the mechanical seal cover of the stirring shaft or near the sealing gap of the flange of the top feeding port. The specific substance includes ammonia, hydrogen chloride, and non-methane total hydrocarbon, which are the most common by-products or unreacted monomers of the polyolefin modified organic amine reaction.

[0084] The conventional parameters, the conventional parameter rate set, and the instantaneous concentration of the specific substance are combined to obtain a process state vector, i.e., the process state vector includes nine vectors of the conventional parameters, the conventional parameter rate set, and the specific substance. The nine vectors in the process state vector also change for different sampling times.

[0085] Z historical batches are screened from the basic database, and Z is selected to ensure that the data has statistical significance. For example, Z is 500. The Z historical batches are batch data after invalid data or a large number of missing values caused by sensor failure, shutdown, and other abnormal situations are removed.

[0086] A process state vector sequence and a historical performance point coordinate (optimal gelation time, optimal comprehensive application risk index) of the corresponding historical batch are extracted from each historical batch. The process state vector sequence represents a sequence combination of process state vectors captured from the key turning period of the reaction from the middle to the end of the historical batch, which describes the state change process of the reaction at each time. The key turning period is determined based on a viscosity threshold. When the real-time monitored viscosity is greater than the viscosity threshold, the process state vectors at different times are collected and combined according to the above sampling period to obtain the process state vector sequence. The viscosity threshold is determined based on statistical analysis of historical data. For example, the viscosity threshold is set to 30% to 50% of the final required viscosity.

[0087] The different process state vectors in the process state vector sequence are normalized by feature scaling to obtain a normalized process state vector sequence. The historical optimal gelation time and the historical optimal comprehensive application risk index corresponding to the historical performance point coordinate of each historical batch are also normalized to obtain a normalized coordinate point (normalized gelation time, normalized comprehensive application risk index).

[0088] The process performance potential field model is constructed by training a deep neural network. The training process of the deep neural network is as follows: the normalized process state vector sequence and the normalized coordinate point are input into the input end of the deep neural network, the input end sends the normalized process state vector sequence to the time sequence feature extraction module, the time sequence feature extraction module is analyzed by using a long short-term memory network, the unimportant information in the normalized process state vector sequence is removed through the forgetting gate in the long short-term memory network, for example, the temperature fluctuation at the beginning of the reaction; and the data that needs to be focused on is filtered out through the input gate, for example, the sudden acceleration of viscosity growth, a hidden state sequence is obtained, and the hidden state sequence is analyzed for longer-term and more complex dependency relationships, for example, it can associate a small pH value drop event at the beginning of the sequence with a rising ammonia concentration event at the end of the sequence, thereby learning that an acidic environment may have induced a side reaction in the later stage, such a deep and cross-time sequence story line, and finally input a fixed-dimension context vector. The context vector is deeply combined and transformed by a fully connected network, the time sequence features and other potential non-time sequence features are preliminarily fused, the hidden interactions between the features are further mined, and finally the predicted coordinate point (predicted gelation time Y1, predicted comprehensive application risk index Y2) is output. The output end compares the predicted coordinate point with the normalized coordinate point through a loss function, and outputs the performance potential. The loss function calculates the performance potential of the predicted coordinate point (predicted gelation time Y1, predicted comprehensive application risk index Y2) and the normalized coordinate point (normalized gelation time Y3, normalized comprehensive application risk index Y4). The expression of the performance potential corresponding to the coordinate point is: In subsequent applications, the real-time process state vector sequence and the performance point coordinate described in step S13 are input into the trained model to obtain the real-time performance potential. The performance potential is an index for predicting the gap between the performance of the final product and the performance point coordinate. The lower the performance potential, the closer the product performance is to the performance point coordinate.

[0089] Step S22, defining a reaction trajectory and calculating a trajectory potential barrier representing the comprehensive cost for the reaction trajectory.

[0090] For future production batches, there are multiple routes that can reach the performance point coordinates from the target heart coordinates, define each route as a reaction trajectory, how to filter out a cost-optimal reaction trajectory from multiple reaction trajectories. The cost-optimal reaction trajectory represents the reaction trajectory with the lowest comprehensive cost from the target heart coordinates to the performance point coordinates. The cost-optimal reaction trajectory is obtained by comparing the trajectory barriers of each complete reaction trajectory. The trajectory barrier is a comprehensive cost indicator reflecting the reaction trajectory, which is obtained by two key cost items. The first cost item is the energy consumption and emission cost, which represents the explicit environmental and energy cost in the production process. Specifically: monitor the combustion consumption rate Q1 of the equipment providing heat source for the reaction kettle in real time, and calculate the emission rate of the pollutant factor according to the pollutant emission factor model. The pollutant emission factor model is a mathematical model that correlates the heat source equipment combustion consumption rate with the main atmospheric pollutant emission amount. Its role is to indirectly estimate the total pollutant emission amount which is difficult to measure directly and frequently, through a more easily measured fuel consumption rate. The pollutant factor includes the nitrogen oxide emission factor N1 and the soot emission factor N2. The nitrogen oxide emission factor represents the mass of nitrogen oxides produced per unit mass or volume of fuel consumed. The soot emission factor represents the mass of soot produced per unit mass or volume of fuel consumed. The calculation formula of the nitrogen oxide emission rate E1 is: The calculation formula of the soot emission rate E2 is: The total pollutant equivalent E is obtained by combining the nitrogen oxide emission rate and the soot emission rate. The calculation formula is: Wherein, and respectively represent the pollution equivalent value of the nitrogen oxide emission factor and the soot emission factor. The total pollutant equivalent is set by publicly released environmental protection regulations, standards or technical guidelines with legal effect. The energy consumption and emission cost is obtained by time integrating the total pollutant equivalent along the reaction trajectory.

[0091] The second cost item is the process abnormality cost, which represents the implicit loss and risk in the production process, such as intensified side reactions, unorganized material escape; it is obtained by the instantaneous concentration of the specific substance described in step S21. Specifically: the instantaneous concentrations of ammonia, hydrogen chloride and non-methane total hydrocarbons are weighted and summed, and the weights are determined by the toxicity or environmental impact factor of each substance. The weighted sum value is time integrated along the entire reaction trajectory to obtain the process abnormality cost.

[0092] The trajectory barrier is calculated by combining the energy consumption and emission cost and the process abnormality cost using the geometric mean method, The geometric mean method is used instead of the arithmetic mean because in the arithmetic mean, a very low key cost item can significantly lower the average, thus masking the risk brought by another very high key cost item. In the geometric mean method, any sharp increase in a cost will cause the trajectory barrier to rise disproportionately, and the reaction trajectory with the lowest trajectory barrier, i.e., the cost-optimal reaction trajectory, is obtained.

[0093] Step S23, define a search space, the search space combines the trajectory barrier and the performance potential to obtain the optimal process state vector.

[0094] A limited hypercube is defined from the historical batches of the basic database as a search space, the purpose is to predict the optimal process state vector in the future by summarizing the experience of Z historical batches, the boundary of the search space is determined by the reasonable range of the process state vector at the reaction endpoint, for example, the ph value [8.0, 10.5], the viscosity [3000, 8000], the concentration of ammonia, hydrogen chloride, and non-methane total hydrocarbon are all in [0, 5]. It is also a nine-dimensional process state vector.

[0095] In the above search space, a particle swarm containing H particles is randomly generated, and H is 100 in an exemplary embodiment, each particle is a candidate endpoint state vector, which also includes nine vectors in the process state vector, and each particle contains an initial random velocity vector for moving and exploring in the search space; the fitness of each particle in the particle swarm is calculated, the fitness represents the degree of approximation to the optimal process state vector, the smaller the fitness, the closer the corresponding particle to the optimal process state vector, each particle is input into the deep neural network trained by the process performance potential field model in step S21 to obtain the particle performance potential, and the method of step S22 is called to calculate the particle trajectory barrier, For each particle, as it is the first time to evaluate the fitness, the position of each particle is set as the individual historical best position of the current particle, and the corresponding fitness is recorded, the individual historical best positions of all particles are traversed, and the particle with the smallest fitness is set as the initial global best position of the entire particle swarm.

[0096] The particles in the search space are iterated, that is, constantly evolved through multiple iterations until the final condition is met and terminated. After each iteration, an update operation is performed on each particle. Specifically, the direction of each particle is updated using the standard update formula of the PSO algorithm. The principle includes three points. The first point is to let the particle maintain the original direction and fly to the second point. The second point is to learn from the individual's experience and fly to the historical best position corresponding to the single particle. The third point is to fly to the best position in the entire particle group, that is, the global best position. The fitness of the new position of the particle is calculated according to the new flight direction. If the fitness of the new position is less than the corresponding individual historical best position, the position of the new position is updated as the individual historical best position, which means that the particle has found a new position better than any place it has ever been. If the fitness of the new position is greater than or equal to the corresponding individual historical best position, the individual historical best position remains unchanged. After updating the individual historical best position, the global best position is updated. The updated individual historical best position of the particle group is traversed again to find the particle with the minimum fitness. The fitness of the global best position corresponding to the particle with the minimum fitness is compared with the first evaluation of the global best position. If it is less than the first evaluation of the global best position, the position of the particle with the minimum fitness is updated as the global best position. Otherwise, it remains unchanged. If the continuous multiple iterations, for example, set to 10 consecutive iterations, avoid too loose termination conditions, leading to premature convergence to a local optimum; the global best position remains unchanged, and the particle with the global best position remains unchanged as the optimal process state vector.

[0097] Step S20 solves the problems of difficult measurement of the gap between real-time performance and target performance, difficult screening of the optimal path of comprehensive cost under multiple reaction trajectories, and difficult prediction of the future optimal process state in the polyurethane modified organic amine reaction process by constructing a process performance potential field model, defining a trajectory potential barrier, and searching for an optimal process state vector. It realizes quantitative evaluation of the gap between reaction performance and target, accurate screening of the optimal reaction trajectory of comprehensive cost, and effective prediction of the future optimal process state. Among them, the process performance potential field model quantifies the gap between real-time performance and target, and can mine the long-term dependence relationship hidden in the time series data; the trajectory potential barrier avoids the single cost item from hiding the risk and accurately reflects the comprehensive cost of the reaction trajectory, helping to screen the cost-optimal trajectory; the search space determines the optimal process state vector, providing a clear state basis for reaction endpoint determination.

[0098] Step 30, based on the optimal process state vector, a two-dimensional endpoint phase coordinate system is constructed, and a convergence area for safe navigation is drawn for the reaction trajectory in the endpoint phase coordinate system; an intention judgment is performed on the convergence area to generate an intention state set; a differential judgment is performed on the intention state set to obtain a reaction termination instruction for judging accurate parking.

[0099] Further, step S30 includes:

[0100] Step S31, based on the optimal process state vector, construct a two-dimensional visual end-point phase coordinate system, and define a convergence area for the reaction trajectory in the end-point phase coordinate system for safe navigation.

[0101] With the unique target point defined by the optimal process state vector in the corresponding nine-dimensional space as the calculation reference, a new end-point phase coordinate system for describing the relative dynamic relationship is established, which is a two-dimensional plane rectangular coordinate system with state distance as the horizontal axis and trajectory deviation degree as the vertical axis. The physical meaning of the origin indicates that the state distance is zero, which means that the real-time measured process state vector has completely coincided with the optimal process state vector. Among them, the state distance L1 represents the Euclidean distance in the nine-dimensional space between the real-time process state vector L2 and the optimal process state vector L3 corresponding to the coordinate value on the horizontal axis, and the specific calculation method is: The trajectory deviation degree L7 represents the included angle between the current motion direction vector and the ideal approaching direction vector, wherein the current motion direction vector L4 represents the actual running direction and speed of L2 in the nine-dimensional space, which is obtained by time difference calculation on L2, that is, L5 represents the real-time process state vector of L2 at the previous sampling time. The ideal approaching direction vector L6 represents the direction of L2 pointing to L3, that is, The calculation result of the trajectory deviation degree is a scalar value represented in radians, which directly reflects whether the current reaction evolution direction is aligned with the end point. The angle of zero indicates perfect alignment, and the larger the angle, the more serious the deviation.

[0102] In the end-point phase coordinate system, an area surrounded by two hyperbolas , is defined as the convergence area of the above-mentioned real-time reaction trajectory, and the purpose is to serve as a safe navigation area for the reaction process. Among them, K is calculated according to the specific reaction kinetic characteristics, the time tolerance and the risk tolerance determined through negotiation in step S12, that is, ​, to ensure that the geometry of the convergence region matches the acceptable diamond-shaped performance target tolerance range. The convergence region is geometrically a horn-shaped region symmetric about the X-axis, embodying the dynamic precision requirement of the polyurethane modified organic amine reaction endpoint control. When the state deviation is large, it indicates that the distance from the optimal process state vector is far, and the upper limit of the allowed trajectory deviation is large, providing space for large adjustment of the reaction process. When the state deviation is small, it is close to the optimal process state vector, and the upper limit of the allowed trajectory deviation is sharply reduced, requiring the change direction of the reaction process to be highly accurate. This gradually tightening constraint strategy conforms to the objective law that the control precision needs to be exponentially improved at the end of the chemical reaction, ensuring the stability of the final product quality. The construction of the endpoint phase coordinate system successfully reduces the complex control problem of nine-dimensional vectors to a two-dimensional plane, providing an intuitive visual monitoring interface for operators and automatic control systems, and realizing the transformation from abstract multi-dimensional data to concrete geometric figures.

[0103] In step S32, intention judgment is performed on the convergence region to generate an intention state set.

[0104] For each real-time process state vector, the state distance and trajectory deviation defined in step 31 are calculated, and the state distance and trajectory deviation are taken as the coordinates of the real-time process state vector, i.e. (state distance, trajectory deviation), to obtain the unique coordinates of each real-time process state vector in the endpoint phase coordinate system, which is marked as a state point, and the state point coordinates are L2(X(B), Y(B)), where L2 is consistent with the above definition, representing the real-time process state vector; X(B) and Y(B) represent the state distance value and the trajectory deviation value of the real-time process state vector on the horizontal axis and the vertical axis of the endpoint phase coordinate system, respectively.

[0105] Based on the hyperbola of the above convergence region, the state point is judged, and the specific judgment process is as follows: if the coordinates X(B) and Y(B) in the state point coordinates L2(X(B), Y(B)) satisfy the inequality , then it indicates that the coordinates are located in the internal region of the convergence region and are marked as a weak intention state. The inequality is derived from the hyperbola of the convergence region, i.e. the mathematical condition for judging whether a coordinate is in the region is that the absolute value of the Y coordinate is less than the boundary value corresponding to the X coordinate , which represents, in a physical sense, that the current reaction is converging towards the optimal process state vector along an acceptable and stable reaction trajectory, and can guide relevant personnel to maintain this weak intention state. For example, only minor adjustments can be made to the actuator such as the valve opening of the heating system, the cooling water flow, the speed of the material dropping pump, etc. The purpose is to maintain the current good trend and suppress potential minor disturbances; if the coordinates X(B) and Y(B) in L2(X(B), Y(B)) satisfy the inequality When the coordinate has touched or crossed the convergence region, on the basis of satisfying the inequality , if the component on the longitudinal axis of the terminal phase coordinate system is the same as the sign of Y(B), for example > 0 and Y(B) > 0 represent the same sign, it is indicated that the coordinate has a trend of out of the convergence region, at this time a strong intention state is generated. Wherein, the component is a first-order difference derivative on the discrete Y coordinate sequence, that is , Y(B-1) represents the trajectory deviation degree of the previous sampling time of Y(B), the calculation result represents the instantaneous motion speed of the state point in the Y axis direction, that is, the longitudinal axis direction, at this time the relevant personnel should be guided to perform a large amount of correction operation. For example, quickly close the heating valve and increase the cooling water flow, the purpose is to forcibly pull the reaction trajectory back to the inside of the convergence region, the composite judgment of the strong intention state is different from the simple boundary alarm, which can effectively distinguish the harmless fluctuation caused by process noise and the real out-of-control precursor with deviation trend, so that the control behavior is more intelligent and stable. The strong intention state and the weak intention state are combined to obtain an intention state set.

[0106] In step S33, differential judgment is performed on the intention state set to obtain a reaction termination instruction for precise parking.

[0107] In the terminal phase coordinate system, a square region with the origin as the center and the side length of 2F is defined as a parking area, the parking area is a minimum square region embedded in the narrowest part of the convergence region and centered on the origin, the purpose is to impose strict and independent final constraints on the state distance and the trajectory deviation degree, wherein F represents the minimum value of the final control precision, the value is determined according to the final error range allowed by each component in the optimal process state vector.

[0108] The differential judgment is performed on the weak intention state in the intention state set, and for the strong intention state, the judgment on the state point corresponding to the strong intention state is repeatedly performed according to the guidance method in step S32 until the weak intention state is obtained, and then the differential judgment is entered, wherein the differential judgment includes in-zone judgment and low-speed judgment.

[0109] The in-zone judgment condition is: if the horizontal coordinate of the state point coordinate corresponding to the weak intention state, i.e., the state distance, is less than F, it is judged that the in-zone judgment condition is met, indicating that the state point corresponding to the weak intention state enters the parking zone. The weak intention state meeting the in-zone judgment condition is a prerequisite for performing the low-speed judgment. The low-speed judgment condition is: if the absolute value of the length change rate of the real-time process state vector corresponding to the state point is less than the preset idling threshold, it is judged that the low-speed judgment condition is met. The length change rate of the process state vector is obtained by performing time difference calculation on the length of the real-time process state vector. The length of the real-time process state vector G1 and the length of the process state vector corresponding to the previous sampling time of the real-time process state vector G2 are calculated by the length formula, and the length change rate G of the process state vector is obtained by first-order differentiation. The length change rate of the process state vector is used to quantify the degree of change of the real-time process state vector in the nine-dimensional space. The closer the length change rate of the process state vector is to zero, the more stable and stationary the reaction is. The idling threshold is a very small positive value, which represents an upper limit of the macroscopic static state indicating basic stop, and is determined according to the statistical distribution of the length change rate of the process state vector of the historical successful batches in the last stage before the reaction ends in the basic database.

[0110] If the in-zone judgment condition and the low-speed judgment condition are met at the same time, the final reaction termination instruction is triggered, which means that the state point not only approaches the optimal process state vector in position, but also tends to be macroscopically stationary in dynamic evolution.

[0111] Step S30 obtains the reaction termination instruction by constructing the end-point phase coordinate system and the convergence area, generating the intention state set, and performing differential judgment, which solves the problems of difficult visualization of multi-dimensional state, difficult identification of reaction trajectory deviation, and difficult precision of end-point determination in the polyurethane modified organic amine reaction end-point control, realizes intuitive monitoring of reaction state, early warning of trajectory deviation, and precise parking of reaction end-point, and provides a complete solution for scientific determination of reaction termination. Among them, the end-point phase coordinate system reduces the nine-dimensional process state vector to two-dimensional plane coordinates, making the complex state relationship more intuitive; the convergence area defines the safe navigation area through hyperbola, gradually tightens the constraint as the state approaches the end-point, and conforms to the law of improving accuracy at the end of the reaction; the intention state set distinguishes between weak intention state and strong intention state, which can maintain stable reaction trend and identify precursors of loss of control in advance; differential judgment through double conditions ensures that the reaction is terminated when the position approaches the optimal state and the dynamic tends to be stationary, eliminating the error of single condition determination.

[0112] Example 2:

[0113] This embodiment provides an intelligent determination system for the end-point of the polyurethane modified organic amine reaction on the basis of Example 1, as shown inFigure 4 illustrated, comprising:

[0114] Performance starting point module: used for obtaining oilfield application conditions to build a basic database simulating different working conditions; a two-dimensional performance plane coordinate system is built according to the basic database, and a rhombus performance target area dedicated to the working condition is divided on the two-dimensional performance plane coordinate system; a performance point with the fastest value increase is determined in the rhombus performance target area;

[0115] Optimal endpoint module: used for building a process performance potential field model based on the performance point to obtain a performance potential measuring the distance from the target; a reaction trajectory is defined, and a trajectory potential barrier representing the comprehensive cost is calculated for the reaction trajectory; the optimal process state vector is obtained by combining the performance potential and the trajectory potential barrier;

[0116] Process path module: used for building a two-dimensional endpoint phase coordinate system based on the optimal process state vector, and defining a convergence area for safe navigation of the reaction trajectory in the endpoint phase coordinate system; an intention state set is generated by performing intention judgment on the convergence area; a reaction termination instruction for judging accurate parking is obtained by performing differentiated judgment on the intention state set.

[0117] In the performance starting point module, the basic database simulating different working conditions is obtained for oilfield application conditions; a two-dimensional performance plane coordinate system is built according to the basic database, and a rhombus performance target area dedicated to the working condition is divided on the two-dimensional performance plane coordinate system; a performance point with the fastest value increase is determined in the rhombus performance target area, comprising:

[0118] Step S11, obtaining oilfield application conditions, performing targeted experiments on the oilfield application conditions to obtain gelation time and dissolution rate reflecting risk indicators, and building a basic database according to the gelation time and the dissolution rate;

[0119] Step S12, building a two-dimensional performance plane coordinate system according to the basic database, and dividing a rhombus performance target area dedicated to the working condition on the two-dimensional performance plane coordinate system;

[0120] Step S13, screening out a golden quadrant representing double optimization of performance and risk in the rhombus performance target area, and determining a performance point with the fastest value increase in the golden quadrant.

[0121] In the optimal endpoint module, the process performance potential field model is built based on the performance point to obtain a performance potential measuring the distance from the target; a reaction trajectory is defined, and a trajectory potential barrier representing the comprehensive cost is calculated for the reaction trajectory; the optimal process state vector is obtained by combining the performance potential and the trajectory potential barrier, comprising:

[0122] Step S21, building a process performance potential field model based on the performance point to obtain a performance potential measuring the distance from the target;

[0123] Step S22, defining a reaction trajectory, and calculating a trajectory barrier representing the comprehensive cost for the reaction trajectory;

[0124] Step S23, defining a search space, combining the trajectory barrier and the performance potential to obtain an optimal process state vector.

[0125] In the process path module, a two-dimensional end-point phase coordinate system is constructed based on the optimal process state vector, and a convergence region for safe navigation is delineated for the reaction trajectory in the end-point phase coordinate system; intention judgment is performed on the convergence region to generate an intention state set; and differential judgment is performed on the intention state set to obtain a reaction termination instruction for judging accurate parking, including:

[0126] Step S31, constructing a two-dimensional visual end-point phase coordinate system based on the optimal process state vector, and delineating a convergence region for safe navigation for the reaction trajectory in the end-point phase coordinate system;

[0127] Step S32, performing intention judgment on the convergence region to generate an intention state set;

[0128] Step S33, performing differential judgment on the intention state set to obtain a reaction termination instruction for accurate parking.

[0129] The method and device, apparatus of the present application can be implemented in many ways. For example, the method and device, apparatus of the present application can be implemented by software, hardware, firmware, or any combination of software, hardware, firmware. The above-described order of steps for the method is only for illustration, and the steps of the method of the present application are not limited to the above specifically described order, unless otherwise specifically stated.

[0130] In addition, the above technical solutions provided in the embodiments of the present application have not been described in detail, which are consistent with the implementation principles of the corresponding technical solutions in the prior art, so as not to be too verbose.

[0131] The above description is only for the embodiments of the present application and the explanation of the technical principles used. Those skilled in the art should understand that the scope of protection involved in the present application is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the technical concept. For example, the above features are replaced with the technical features disclosed in the present application (but not limited to) having similar functions to form technical solutions.

Claims

1. A method for intelligent determination of the end point of a polyester-modified organic amine reaction, characterized in that The method comprises: Obtaining an oilfield application working condition to construct a basic database simulating different working conditions; constructing a two-dimensional performance plane coordinate system according to the basic database, and dividing a working condition exclusive rhombus performance target area on the two-dimensional performance plane coordinate system; determining a performance point with the fastest performance in the rhombus performance target area according to an equal proportion value equation; Based on the performance point, a process performance potential field model is constructed to obtain a performance potential energy for measuring the distance from the target; a reaction trajectory is defined, and a trajectory potential barrier representing the comprehensive cost is calculated for the reaction trajectory; the performance potential energy and the trajectory potential barrier are combined to obtain an optimal process state vector; Based on the optimal process state vector, a two-dimensional terminal phase coordinate system is constructed, and a convergence area for safe navigation is drawn for the reaction trajectory in the terminal phase coordinate system; an intention judgment is performed on the convergence area to generate an intention state set; a differential judgment is performed on the intention state set to obtain a reaction termination instruction for judging parking; The performance point comprises: defining a local coordinate system with the target center as the origin in the rhombus performance target area, and the local coordinate system is a plane rectangular coordinate system; The connecting line area of the horizontal axis and the vertical axis of the local coordinate system and the rhombus performance target area in the third quadrant of the local coordinate system is defined as the golden quadrant, which is a right triangle area surrounded by the target center coordinates, the left vertex and the lower vertex in geometry; Starting from the target center, a ray along the optimal descending direction representing the synchronization of time and risk is emitted, and the intersection point of the ray and the connecting line between the left vertex and the lower vertex is the performance point; The method for obtaining the reaction termination instruction comprises: in the terminal phase coordinate system, defining a square area with the origin as the center and the side length as 2F as a parking area, wherein F represents the minimum value of the final control accuracy; The differential judgment comprises an area entry judgment and a low speed judgment, and the area entry judgment is performed on the weak intention state in the intention state set, and if the horizontal axis coordinate of the state point corresponding to the weak intention state is less than F, it is judged that the area entry judgment condition is met; meanwhile, the weak intention state meeting the area entry judgment condition is a prerequisite for performing the low speed judgment, and the modulus change rate of the real-time process state vector corresponding to the state point is obtained, and if the absolute value of the modulus change rate of the real-time process state vector corresponding to the state point is less than a preset idle speed threshold, it is judged that the low speed judgment condition is met; If the area entry judgment condition and the low speed judgment condition are met at the same time, the final reaction termination instruction is generated.

2. The method of claim 1, wherein the polymeric polyol modified organic amine reaction endpoint smart determination method is characterized by, The method for constructing the basic database comprises: The oilfield application working condition comprises reservoir basic parameters, rock types and polymer types expected to be used together; The rock types are extracted from the oilfield application working condition, experiments are performed based on the chemical stability difference of rock components, the dissolution rate is detected, and the gelation time is obtained by simulating the actual working condition through the oilfield application working condition to obtain the residual organic amine monomer concentration; The working conditions of different historical production batches are simulated to obtain the residual organic amine monomer concentration, the dissolution rate and the gelation time corresponding to different historical production batches, and the residual organic amine monomer concentration, the dissolution rate and the gelation time corresponding to different historical production batches are integrated to obtain the basic database.

3. The method of claim 2, wherein the polymeric polyol modified organic amine reaction endpoint smart determination method is characterized by, The method for obtaining the rhombus performance target area comprises the following steps: The residual organic amine monomer concentration and the dissolution rate are combined, and the comprehensive application risk index is calculated by using a risk index formula; The oilfield application working condition is matched and searched through the basic database, N historical production batches most similar to the oilfield application working condition are obtained, the gelation time and the comprehensive application risk index in the N historical production batches are extracted, and the arithmetic average of the N gelation times and the arithmetic average of the comprehensive application risk indexes are calculated. The arithmetic average of the gelation time corresponds to the horizontal axis value, and the arithmetic average of the comprehensive application risk index corresponds to the vertical axis value. The operation general procedure library of the oilfield is obtained, an acceptable time tolerance and a risk tolerance are determined, and four vertex coordinates, i.e., an upper vertex, a lower vertex, a right vertex and a left vertex, are marked in the two-dimensional performance plane coordinate system according to the time tolerance and the risk tolerance.

4. The method of claim 3, wherein the polymeric polyol-modified organic amine reaction endpoint smart determination method is characterized by, The rhombus performance target area is formed by sequentially connecting the upper vertex, the lower vertex, the right vertex and the left vertex with the target heart as the geometric center. The method for obtaining the trajectory barrier comprises the following steps: The combustion consumption rate of the equipment providing heat source for the reaction kettle is monitored in real time; The nitrogen oxide emission rate and the soot emission rate are calculated by using a pollutant emission factor model based on the combustion consumption rate, the total pollutant equivalent is obtained by combining the nitrogen oxide emission rate and the soot emission rate, and the energy consumption emission cost is obtained by time integration along the reaction trajectory. The sensors are deployed at the leakage points of the reaction kettle to obtain the instantaneous concentrations of ammonia, hydrogen chloride and non-methane total hydrocarbon, and the process abnormal cost is obtained by weighted summation of the instantaneous concentrations of ammonia, hydrogen chloride and non-methane total hydrocarbon.

5. The method of claim 4, wherein the polymeric polyol modified organic amine reaction endpoint smart determination method is characterized by, The process abnormal cost and the energy consumption emission cost are combined and calculated by using the geometric mean method to obtain the trajectory barrier. The method for obtaining the optimal process state vector comprises the following steps: A search space is defined through the basic database, and H particles are randomly generated in the search space; The particle performance potential energy is calculated by using the process performance potential field model for each particle in the particle swarm, and the particle trajectory barrier is calculated by using the method for obtaining the trajectory barrier; The fitness of each particle in the particle swarm is calculated by using the particle performance potential energy and the particle trajectory barrier, the individual historical best position of each particle in the search space is set according to the fitness, the individual historical best position refers to the position corresponding to the minimum fitness of each particle in the search space, and the particle with the minimum fitness is set as the initial global best position of the whole particle swarm by traversing the individual historical best positions of all particles; The particles in the search space are iterated, the update operation is performed on each particle after each iteration, the fitness of each particle after iteration is calculated, and the position in the search space of the particle after iteration is marked as the individual historical best position if the fitness of the particle after iteration is less than the fitness of the particle before iteration. The fitness of all particles after iteration is traversed, and the particle with the minimum fitness is set as the global best position of the whole particle swarm. If the global optimal position remains unchanged after continuous iterations, the particle corresponding to the global optimal position is taken as the optimal process state vector.

6. The method of claim 5, wherein the polymeric polyol modified organic amine reaction endpoint smart determination method is characterized by, The method for obtaining the convergence region comprises: Based on the optimal process state vector, state distance and trajectory deviation degree are calculated; A terminal phase coordinate system is constructed for the optimal process state vector based on the state distance and the trajectory deviation degree, wherein the terminal phase coordinate system takes the state distance as the horizontal axis and the trajectory deviation degree as the vertical axis; In the terminal phase coordinate system, a region surrounded by two hyperbolas, i.e., the convergence region, is defined.

7. The method of claim 6, wherein the polymeric polyol modified organic amine reaction endpoint smart determination method is characterized by, The method for obtaining the set of intention states comprises: The state distance and the trajectory deviation degree of each real-time process state vector are calculated, and the state distance and the trajectory deviation degree of the real-time process state vector are taken as the coordinates of the real-time process state vector, i.e., the unique coordinates in the terminal phase coordinate system, which are marked as state points, wherein the state distance and the trajectory deviation degree of the real-time process state vector are the horizontal axis coordinate and the vertical axis coordinate of the state point, respectively; Based on the hyperbolic inequality derived from the convergence region, the state points are judged to obtain strong intention states and weak intention states; The strong intention states and the weak intention states are combined to obtain the set of intention states.

8. A smart determination system of the end point of a polyol-modified organic amine reaction for implementing the smart determination method of the end point of a polyol-modified organic amine reaction according to any one of claims 1 to 7, characterized in that, The system comprises: A performance starting point module is configured to obtain an oilfield application working condition to construct a basic database simulating different working conditions, to construct a two-dimensional performance plane coordinate system based on the basic database, to divide a working condition exclusive rhombus performance target area on the two-dimensional performance plane coordinate system, and to determine a performance point with the fastest value increase in the rhombus performance target area; An optimal terminal module is configured to construct a process performance potential field model based on the performance point to obtain a performance potential energy for measuring the distance from the target, to define a reaction trajectory and to calculate a trajectory potential barrier representing the comprehensive cost for the reaction trajectory, and to obtain an optimal process state vector by combining the performance potential energy and the trajectory potential barrier; A process path module is configured to construct a two-dimensional terminal phase coordinate system based on the optimal process state vector, to demarcate a convergence region for safe navigation of the reaction trajectory in the terminal phase coordinate system, to perform intention judgment on the convergence region to generate a set of intention states, and to perform differential judgment on the set of intention states to obtain a reaction termination instruction for judging stopping.

Images