A system-level twin model initialization method combining mechanism and optimization algorithm

Abstract

The present invention provides a system-level twin model initialization method combining mechanism and optimization algorithm, including: decoupling a complex running digital twin system into a single system and calculating boundary parameters; screening unit data associated with the system to be debugged from the unit operation data and performing preprocessing; classifying the input and output parameters of the simulation model to determine whether there are performance parameters with a definite relationship; for performance parameters with a definite relationship, using the processed unit operation data for reverse calculation according to physical principles and establishing a debugging experience library for fine-tuning; for performance parameters without a definite relationship, using an intelligent optimization algorithm for optimal calculation; executing a calculation program to enable the simulation model to calculate and reach a steady state. The present invention has a smaller error, is faster and more accurate than initialization that only relies on an intelligent optimization algorithm.

Description

Technical Field

[0001] The present invention relates to the technical field of system-level twin model initialization, and in particular to a system-level twin model initialization method combining a mechanism and an optimization algorithm. Background Art

[0002] Digital twin technology can realize the mirroring of physical entities in virtual space, reflecting the entire life cycle of physical entities. How to realize the interaction between digital twin models and physical entities is one of the key issues that need to be solved in digital twin related research. The operation digital twin of nuclear power units is based on system simulation technology. By connecting unit data to the simulation model, using unit data to verify and optimize the model, and continuously improving the consistency between the model and the unit, the digital twin system has the same engineering data, the same initial state, and the same response results based on the same operation as the physical reference unit. It is a digital twin realized from the perspective of unit operation.

[0003] In the study of digital twins of nuclear power plant operation, it is necessary to realize automatic initialization of the system model, that is, the system model automatically returns to the state consistent with the selected moment of the unit, ensure the consistency of the equipment state, and the error of the main process parameters is within the allowable range. However, due to the complexity of the digital twin model of nuclear power operation, coupled with the influence of factors such as actual equipment aging, scaling, seawater temperature changes, life span, and the simplification and assumptions of the simulation model itself, it is difficult to initialize the simulation model to a state that is completely consistent with the unit. At present, when the simulation model has been determined, the direction to reduce the difference is mainly to adjust the adjustable parameters of the model and use the unit data to reversely calculate the simulation model parameters. However, there are many adjustable parameters in the digital twin model of nuclear power plant operation, and there are different degrees of coupling between the parameters. It often happens that the parameters debugged according to A, B, and C will cause large deviations in the simulation results of D. Summary of the invention

[0004] The purpose of the present invention is to overcome the defects described in the prior art, thereby providing a system-level twin model initialization method that combines mechanism and optimization algorithm. The method can automatically calculate the optimal simulation parameters based on the actual operation data of the unit, combined with the calculation principles of the system simulation model and the intelligent optimization algorithm, so as to realize the automatic switching of the digital twin model to a state consistent with the selected moment of the physical entity, thereby realizing the twin model initialization.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A system-level twin model initialization method combining mechanism and optimization algorithm includes the following steps:

[0007] S1: Decouple the complex operational digital twin system into a single system and calculate boundary parameters;

[0008] S2: Filtering the unit data related to the system to be debugged from the unit operation data and performing preprocessing;

[0009] S3: Classify the input and output parameters of the simulation model to determine whether there are performance parameters with a definite relationship;

[0010] S4: For performance parameters with a definite relationship, the unit operation data is combined with the node topology result calculation, and an initialization program is added to the simulation model calculation program. According to the physical principle, the processed unit operation data is used for reverse calculation and a debugging experience library is established for fine-tuning;

[0011] S5: For performance parameters with no definite relationship, the optimization target, weight and fitness function are determined based on experience or correlation analysis results, and an intelligent optimization algorithm is used for optimization calculation;

[0012] S6: Execute the S4 calculation program to make the simulation model calculate and reach a steady state.

[0013] S2 includes:

[0014] S2.1: Filter out the associated unit data according to the three-digit system code;

[0015] S2.2: Unit data preprocessing.

[0016] In S2.1, the associated unit data includes: unit data including the name of the initialization system; unit data including the name of the heat exchange system; unit data including the name of the mass exchange boundary system.

[0017] In S2.2, the unit data stored in the PI system is analyzed and preprocessed for the following situations before being used for model initialization:

[0018] For regulating valves, the valve stem position is within 0-100%;

[0019] Identify the correct equipment status: Under normal circumstances, the equipment status cannot be switched frequently, and the change of equipment status will cause the related pressure instruments to change synchronously, and the temperature and flow instruments to change with delay;

[0020] Instruments with redundant measurements need to use data confirmation and coordination technology to eliminate errors so that the instrument measurement values ​​meet the principles of conservation of mass and energy.

[0021] In S3, the input parameters of the model include structural parameters, performance parameters, boundary parameters and initial parameters, and the output parameters include pressure, temperature, flow rate, etc.

[0022] Furthermore, the boundary parameters are determined according to the unit operation data, the structural parameters are determined according to the design parameters, and the performance parameters are divided into the following two categories according to whether there is a definite relationship between the performance parameters and the output results of the twin model:

[0023] Type 1: Performance parameters with a definite relationship, which are directly calculated through the unit operation data;

[0024] Type 2: Performance parameters with no definite relationship, which cannot be directly calculated.

[0025] S4 includes:

[0026] S4.1: Calculate theoretically based on the simulation model;

[0027] S4.2: Fine-tune according to the experience library. Since the numerical solution process is an iterative process and there are errors, it is impossible to obtain very accurate results for parameters such as density, specific enthalpy and temperature in the theoretical calculation process of S4.1. Therefore, there is room for fine-tuning based on the results of the simulation model theoretical calculation. By establishing a debugging experience library, matching is performed for each pipeline, and further adjusting the type one parameters to further reduce errors.

[0028] S5 includes:

[0029] S5.1: Optimize the target selection strategy and select the sensor strategy result as the optimization target;

[0030] S5.2: Optimize the weight setting strategy, set the optimization target value weight according to the sensor range and the accuracy required by the user.

[0031] In S5.1, the unit measurement values ​​are selected from multiple angles such as redundant sensors, water vapor properties, and conservation principles to obtain the optimization target.

[0032] In S6, S5 is started based on the result of S4, and the parameters optimized by S5 are brought into the mechanism model. S4 is executed again. After the simulation model reaches a steady state, the error is calculated and initialization is completed. The above steps are repeated until the error meets the accuracy requirements and is coupled with other systems.

[0033] Compared with the prior art, the system-level twin model initialization method based on the combination of mechanism and optimization algorithm provided by the present invention has the following beneficial effects:

[0034] The method provided by the present invention uses a method that combines mechanism and intelligent optimization algorithm to initialize the parameters of the digital twin model. The two methods promote each other and can obtain a set of parameter values ​​that meet the minimum error. Compared with the initialization that relies entirely on the mechanism, the error is smaller, and compared with the initialization that only relies on the intelligent optimization algorithm, the speed is faster and more accurate.

[0035] Figure 1 and Figure 2 Taking the operating data of a certain M310 nuclear power unit in April 2017 as an example, this method is used, with the unit's 100% operating condition as the benchmark operating condition. After two iterations, the error between the simulation results and the unit data is within 1%.

[0036] Analysis of the test results revealed that before the first iteration, the turbine efficiency directly adopted the design data, resulting in a large difference between the turbine power and the target value when the pressure and flow were mostly close. After the first optimization, the turbine efficiency changed significantly. The change in efficiency will affect the pressure and temperature. Therefore, after the second model initialization, when the pressure and flow were close, the turbine power was also close to the target value, verifying that the model initialization and optimization tools promote each other. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 This is a schematic diagram of the first optimization convergence curve of the benchmark working condition;

[0039] Figure 2 This is a schematic diagram of the second optimization convergence curve for the benchmark condition;

[0040] Figure 3 A schematic diagram of the decoupling of the digital twin model of nuclear power operation provided by the present invention;

[0041] Figure 4 Schematic diagram of the main interfaces between the steam system provided by the present invention and other systems;

[0042] Figure 5 A flow chart of a type 1 parameter initialization program provided by the present invention;

[0043] Figure 6 A flow chart of the pipeline admittance calculation method provided by the present invention;

[0044] Figure 7 This is a flow chart of the type 2 parameter initialization method (intelligent optimization algorithm calculation process) provided by the present invention. DETAILED DESCRIPTION

[0045] The following is further explained in detail through specific implementation methods.

[0046] The present invention provides a system-level twin model initialization method combining mechanism and optimization algorithm, comprising the following steps:

[0047] S1: Decouple the complex operating digital twin system into a single system and calculate the boundary parameters. The decoupling is mainly based on simulation experience, that is, whether the mechanism model is consistent and whether the system with consistent mechanism model can clarify the boundary conditions. The system with clear boundary conditions is decoupled. The purpose of decoupling is to reduce the complexity of the system while reducing the dimension of the intelligent optimization algorithm. The nuclear power operation twin model involves multiple disciplines such as core, thermal engineering, fluid, electrical and instrumentation and control systems, and it simulates the entire power plant process. The system is very complex. Therefore, it is necessary to decouple the multidisciplinary simulation model. The decoupling is based on whether the mechanism model is consistent and whether the system with consistent mechanism model can clarify the boundary conditions. The system with clear boundary conditions is further decoupled.

[0048] S2: Preprocessing of associated unit data. Filter the unit data associated with the system to be debugged from the unit operation data and preprocess it to improve the reliability of the unit data.

[0049] S2.1: Filter out the associated unit data according to the three-digit system code, including the following three categories:

[0050] (1) Unit data including the name of the initialization system;

[0051] (2) Unit data including the name of the heat exchange system;

[0052] (3) Unit data including the name of the mass exchange boundary system;

[0053] S2.2: Unit data preprocessing. Since the unit's operating data mainly comes from the PI system, through the analysis of the unit data stored in the PI system, it is found that the data quality is not high and there are the following problems. It needs to be preprocessed before it can be used for model initialization.

[0054] (1) For regulating valves, the valve stem position should be within 0-100%;

[0055] (2) Identify the correct equipment status: Under normal circumstances, the equipment status cannot switch frequently, and changes in equipment status will cause the related pressure instruments to change synchronously, and the temperature and flow instruments to change with delay.

[0056] (3) Instruments with redundant measurements need to use data confirmation and coordination technology to eliminate errors so that the instrument measurement values ​​meet the principles of conservation of mass and energy.

[0057] S3: Classification of simulation model input and output parameters. Taking the simulation model of the secondary circuit process system of a power plant as an example, the input parameters of the model include structural parameters, performance parameters, boundary parameters and initial parameters. The output parameters include pressure, temperature, flow, etc. The initial parameters in the input parameters have little effect on the results, and mainly affect the initialization completion time. The boundary parameters can be determined according to the unit operation data, and the structural parameters can be determined according to the design parameters. Therefore, the performance parameters are the main factors that have a greater impact on the output results of the twin model. According to whether there is a definite relationship between the performance parameters and the output results of the twin model, the performance parameters can be divided into the following two categories.

[0058] Type 1: Performance parameters with a definite relationship, that is, there is a definite relationship between the performance parameters and the output results of the twin model, and this type of performance parameters can be directly calculated through the unit operation data;

[0059] Type 2: Performance parameters with no definite relationship, that is, there is a relationship between the performance parameter and the output result of the twin model, but the performance parameter cannot be directly calculated.

[0060] S4: For type 1 performance parameters, they can be calculated by combining the unit operation data with the node topology results. An initialization program is added to the simulation model calculation program. According to physical principles, the processed unit operation data is used for reverse calculation and a debugging experience library is established for fine-tuning.

[0061] S4.1: Calculate theoretically based on the simulation model. Parameters such as pipeline admittance and heat transfer coefficient can be calculated in combination with the theoretical model, so that a result that basically meets the requirements can be quickly obtained.

[0062] S4.2: Fine-tune according to the experience library. Since the numerical solution process is an iterative process and there are errors, it is impossible to obtain very accurate results for parameters such as density, specific enthalpy and temperature in the S4.1 theoretical calculation process. Therefore, there is room for fine-tuning according to the results of the simulation model theoretical calculation. By establishing a debugging experience library, matching each pipeline, and further adjusting the type 1 parameters, the error can be further reduced.

[0063] S5: For type 2 performance parameters, determine the optimization target, weight and fitness function based on experience or correlation analysis results, and use an intelligent optimization algorithm to perform optimization calculations.

[0064] S5.1: Optimization target selection strategy. Generally, the sensor strategy result is selected as the optimization target, but not all relevant sensor measurement point information is suitable as the optimization target. It is necessary to select the unit measurement value from multiple angles such as redundant sensors, water vapor properties, and conservation principles.

[0065] S5.2: Optimization weight setting strategy. The selection of optimization target value weight should be set according to the sensor range and the accuracy required by the user. That is, the larger the sensor range, the smaller the weight should be. The higher the accuracy required by the user, the larger the weight should be.

[0066] S6: Execute the S4 calculation program to make the simulation model calculate and reach a steady state. Based on the results of S4, start S5, and bring the parameters optimized by S5 into the mechanism model, execute S4 again, and after the simulation model reaches a steady state, calculate the error and complete an initialization. Repeat the above steps until the error meets the accuracy requirements and is coupled with other systems.

[0067] For example, the following description will focus on the steam system as an example.

[0068] S1: System decoupling and boundary parameter calculation: The nuclear power unit operation digital twin system can be divided into two types according to the type of mechanism model and whether the boundary conditions can be determined from the unit data. Figure 3 Several subsystems are shown.

[0069] See the schematic diagram of the boundary between the steam system and other systems for details. Figure 4 , the calculation methods of various boundary parameters are shown in Table 1:

[0070] Table 1 Calculation method of steam system boundary parameters

[0071]

[0072]

[0073] S2: Preprocessing of associated unit data.

[0074] S2.1: Related data screening. The steam system in this example involves multiple power plant systems, and has flow interfaces with the steam generator, and pressure interfaces with the condenser, condensate system, and feedwater system. After scope sorting, the unit data list described in the following table was determined.

[0075] Table 2 Unit data associated with the steam system

[0076]

[0077]

[0078] S2.2: Unit data preprocessing.

[0079] (1) The valve position of the main regulating valve is limited between 0-100%. For valve positions less than 0, it is set to 0;

[0080] (2) Identifying the correct equipment status: No unexpected equipment movement was found during the steam system data analysis.

[0081] (3) Data processing and coordination: Based on the steam flow, feed water flow, and sewage flow, the data processing and coordination algorithm is called to calculate the coordinated steam flow.

[0082] S3: Analysis of simulation model input and output parameters. According to whether there is a definite relationship between the performance parameters and the output results of the twin model, the steam system type 1 parameters mainly include the admittance of each pipeline (the inverse variable of the resistance coefficient). Type 2 parameters are mainly shown in the following table:

[0083] Table 3 Parameters of steam system model type 2

[0084]

[0085] S4: Type 1 parameter initialization method.

[0086] S4.1: Calculate based on the simulation model theory.

[0087] Type 1 parameter initialization requires the development of an initialization program in the simulation platform based on the calculation method of the mechanism model. The calculation formula of the admittance in the mechanism model is:

[0088] ksi=f / (s*vlv*sqrt(2*df*(p1-p2+df*g*h))

[0089] in,

[0090] Ksi-pipe admittance;

[0091] f-pipeline flow rate, kg / s;

[0092] s-pipe cross-sectional area, m 2 ;

[0093] vlv-normalized valve flow area;

[0094] df - upstream node density, kg / m 3 ;

[0095] p1-upstream node pressure, pa;

[0096] p2-downstream node pressure, pa;

[0097] g - acceleration due to gravity;

[0098] h-height difference between upstream and downstream nodes;

[0099] As can be seen from the above formula, the pipeline admittance requires the known pressure difference and flow rate of its upstream and downstream nodes, but this information cannot be directly obtained from the unit operation data and needs to be processed according to the topological structure of the pipeline network. For pipelines with similar diameters and connected in series, the principle of average distribution can be used for the pressure drop. When there are pressure-reducing elements in the pipeline, such as regulating valves and orifice plates, the average distribution method can no longer be used due to the large resistance loss of the pressure-reducing elements. At this time, the upstream and downstream nodes of the pressure-reducing elements need to be set as key nodes. By adding key nodes and their pressure settings, the connection relationship between non-key nodes and key nodes is determined in combination with the network topology, and the series pipeline group between the two key nodes is found. The pressure of non-key nodes in the series pipeline group is calculated using the principle of average distribution. In addition, the user is required to set the pipeline flow rate as the target to achieve the preliminary calculation of the pipeline admittance using the admittance calculation formula.

[0100] For the steam system, the following types of nodes need to be set as key nodes.

[0101] (1) Nodes connected to pressure sensors: Node pressure is derived from unit data;

[0102] (2) Nodes before and after the regulating valve: The pressure is set based on experience. The pressure drop in the pipeline where the main regulating valve is located = 4-5 times the pressure drop in the pipeline where the main steam valve is located;

[0103] (3) Inlet and outlet nodes of high-pressure cylinder, medium-pressure cylinder and low-pressure cylinder stage group - pressure is obtained according to the heat balance diagram.

[0104] S4.2: Fine-tune based on the experience base.

[0105] Since the density in the above method is taken as the value at a certain moment in the process of calculating the admittance, considering the complexity of the pipe network structure, in order to make the result of the model initialization closer to the target value, it is necessary to further fine-tune it based on experience. The debugging experience is organized into an experience library, and the current flow network satisfies which experience and automatically fine-tunes it based on the experience. The debugging experience based on the calculation process is as follows:

[0106] (1) When the pipeline flow rate is less than the set flow rate, the upstream node pressure is greater than the set pressure, and the downstream node pressure is less than the set pressure, the pipeline conductance should be increased according to a certain step size;

[0107] (2) When the pipeline flow rate is greater than the set flow rate, the upstream node pressure is less than the set pressure, and the downstream node pressure is greater than the set pressure, the pipeline conductance should be reduced according to a certain step size;

[0108] (3) When the pipeline flow rate is greater than the set flow rate, and the upstream node pressure is greater than the set pressure, and the downstream node pressure is greater than the set pressure, the pipeline conductance should be reduced according to a certain step size;

[0109] (4) When the pipeline flow rate is less than the set flow rate, the upstream node pressure is less than the set pressure, and the downstream node pressure is less than the set pressure, the pipeline conductance should be increased according to a certain step size;

[0110] The above debugging experience is applicable to all pipelines, including steam pipes and drain pipes, but the flow rate in the drain pipe should be determined by the drain flow rate.

[0111] The pipeline admittance initialization process is as follows Figure 5 and Figure 6 shown.

[0112] S5: Type 2 parameter initialization method.

[0113] Type II parameters are generally parameters related to performance or efficiency. A multi-objective optimization algorithm based on a single objective is used for calculation. The learning factors c1 and c2 are both set to 0.5, the initial population size is set to 30, and the maximum number of iterations is set to 100.

[0114] S5: Type 2 parameter initialization method.

[0115] Type II parameters are generally parameters related to performance or efficiency. A multi-objective optimization algorithm based on a single objective is used for calculation. The learning factors c1 and c2 are both set to 0.5, the initial population size is set to 30, and the maximum number of iterations is set to 100.

[0116] (1) Optimize target value selection.

[0117] The optimization target value should select the variables that the user is concerned about and have an important impact on the system operation. The sensor measurement value is the best choice for the model optimization target value. The following table shows the step-by-step situation of the steam system sensor. Through analysis, it was found that there are 78 sensors or measurement point signals in the steam system, but there is a lot of redundancy in these measurement point data. Therefore, not all sensor measurements are suitable as target values.

[0118] Table 4 Steam system sensor list

[0119]

[0120]

[0121]

[0122] Whether the sensor is suitable as the target value is mainly analyzed from the following situations:

[0123] 1) Redundant sensors.

[0124] In order to ensure sufficient safety, the unit has multiple pressure, flow or temperature sensors at the same measuring point. However, due to the error of the sensor itself, the values ​​of the redundant sensors are slightly different. To solve this problem, the redundant sensors need to be processed. In this solution, if the redundant sensors are processed in the DCS and there is a variable output, the result after DCS processing is directly taken. If not, the average of the redundant sensors is taken or the maximum and minimum values ​​are removed and the middle value is taken.

[0125] 2) Physical properties of water vapor.

[0126] For the reference unit, the SG outlet is saturated steam, the entire high-pressure cylinder and part of the low-pressure cylinder are saturated steam. For saturated steam, pressure and temperature can be calculated based on physical properties. Considering that the measurement accuracy of temperature sensors is generally higher than that of pressure sensors, the data in a saturated state and with both temperature and pressure measurement points are corrected for pressure according to the temperature sensor table lookup. In the optimization, the corrected pressure sensor value is taken as the optimization target. In addition, the entire medium-pressure cylinder is in a superheated steam state, and the pressure and temperature measurement results should ensure that the steam is in a superheated state.

[0127] According to the above principles, the data of the measuring points in the saturated state in the steam system are analyzed. For VVP and GSS systems, when the measurement results of the pressure sensor are processed as absolute pressure, they are closer to the difference results of the corresponding temperature sensor. For AHP systems, when the measurement results of the pressure sensor are processed as relative pressure, they are closer to the difference results of the corresponding temperature sensor.

[0128] 3) Conservation principle.

[0129] According to the Bernoulli equation, the pressure on the same streamline should gradually decrease. It is necessary to constrain the unit data according to the conservation principle.

[0130] 4) Whether the sensor measurement value is sufficient.

[0131] Whether the type and arrangement of sensors can uniquely determine the system status. The unit data lacks the flow rate, specific enthalpy or dryness of each level of extraction pipeline, making it difficult to determine the system status. This phenomenon is common in various units. The extraction flow rate under the design condition is used as a constraint, and the SG outlet steam dryness is assumed to be 0.997. At the same time, when optimizing the baseline condition, the extraction flow rate is added as a constraint. When the flow rate is within 5% deviation, the fitness increase caused by the extraction flow rate deviation is not calculated. Otherwise, the fitness caused by the extraction flow rate deviation is calculated.

[0132] (2) Optimize the target value weight setting.

[0133] For multi-objective optimization algorithms based on a single objective, the weight setting has a direct impact on the optimization results. If the weight setting is not good, it is very likely that the optimization results will not meet the requirements.

[0134] The selection of the weight of the target value for simulation model optimization is set according to the sensor range and the accuracy required by the user. Generally speaking, the larger the range, the smaller the weight should be, and the higher the accuracy requirement, the larger the weight should be. For example, the range of electric power is 0-1300MW, which is relatively large compared to other parameters. According to the principle that the larger the range, the smaller the weight, the weight of this variable should be set smaller. However, power plants generally require that the error of electric power is within 2MW and the accuracy is within 0.2%, so the weight of electric power should be increased.

[0135] The final fitness function is:

[0136]

[0137] Among them: cost-fitness;

[0138] Pi - target value;

[0139] Psim - simulation value;

[0140] b-weight.

[0141] Type 2 parameter initialization calculation process is as follows Figure 7 As shown in the figure, the optimization algorithm controls the operation and freezing of the simulation model as needed through the interface program. The calculation process of this method is relatively simple, but it requires the development of additional communication programs to control the operation, freezing, and variable reading and writing of the simulation platform where the simulation model is located. In the application of the optimization algorithm, it is necessary to pay attention to the judgment method of the stability of the simulation model calculation.

[0142] 1) Normally, the method to judge the stability of model calculation should be that the difference between the two calculation results of all variables is very small. However, if this is used as the judgment condition, it will cause errors when the fluctuation caused by the change of optimization parameters has not yet propagated after the last steady state. The solution is to set a suitable timer (set to 10s during steam system optimization) according to the node scale and calculation cycle of the simulation model to ensure that the disturbance has propagated to more than half of the nodes, and then calculate the difference between the two calculations of the variables. When the difference is within a certain range, it is considered that the simulation model has reached a steady state.

[0143] 2) During the iteration process, it is very likely that the optimization parameters of a certain calculation cannot make the model stable or even diverge. Such optimization results need to be eliminated. In this case, the timer in 1) and the difference between the two calculations of all pipeline flows are used to assist in identification. When the timer exceeds the preset value and the difference between the two calculations of all pipeline flows is greater than the preset value, it is considered that this optimization result needs to be eliminated.

[0144] S6: Execute the S4 calculation program to make the simulation model calculate and reach a steady state. Based on the results of S4, start S5, and bring the parameters optimized by S5 into the mechanism model, execute S4 again, and after the simulation model reaches a steady state, calculate the error and complete an initialization. Repeat the above steps until the error meets the accuracy requirements and is coupled with other systems.

[0145] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A system-level twin model initialization method combining mechanism and optimization algorithm, characterized in that: The following steps are involved: S1: Decouple the complex operational digital twin system into a single system and calculate boundary parameters; S2: Filter out the unit data associated with the system to be debugged from the unit operation data and perform preprocessing; S3: Classification of simulation model input and output parameters, classifying the performance parameters according to whether there is a definite relationship between the performance parameters and the output results of the twin model; S4: For performance parameters with a definite relationship, the unit operation data is combined with the node topology result calculation, and an initialization program is added to the simulation model calculation program. According to the physical principle, the processed unit operation data is used for reverse calculation and a debugging experience library is established for fine-tuning; S5: For performance parameters with no definite relationship, the optimization target, weight and fitness function are determined based on experience or correlation analysis results, and an intelligent optimization algorithm is used for optimization calculation; S6: Execute the S4 calculation program to make the simulation model calculate and reach a steady state.

2. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 1 is characterized in that S2 include: S2.1: Filter out the associated unit data according to the three-digit system code; S2.2: Unit data preprocessing.

3. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 2 is characterized in that: In S2.1, the associated unit data includes: unit data including the name of the initialization system; unit data including the name of the heat exchange system; unit data including the name of the mass exchange boundary system.

4. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 3 is characterized in that: In S2.2, the unit data stored in the PI system is analyzed and preprocessed for the following situations before being used for model initialization: For regulating valves, the valve stem position is within 0-100%; Identify the correct equipment status: Under normal circumstances, the equipment status cannot be switched frequently, and the change of equipment status will cause the related pressure instruments to change synchronously, and the temperature and flow instruments to change with delay; Instruments with redundant measurements need to use data confirmation and coordination technology to eliminate errors so that the instrument measurement values ​​meet the principles of conservation of mass and energy.

5. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 1 is characterized in that: In S3, the input parameters of the model include structural parameters, performance parameters, boundary parameters and initial parameters, and the output parameters include pressure, temperature and flow rate.

6. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 5 is characterized in that: The boundary parameters are determined according to the unit operation data, the structural parameters are determined according to the design parameters, and the performance parameters are divided into the following two categories according to whether there is a definite relationship between the performance parameters and the output results of the twin model: Type 1: Performance parameters with a definite relationship, which are directly calculated through the unit operation data; Type 2: Performance parameters with no definite relationship, which cannot be directly calculated.

7. The system-level twin model initialization method combining the mechanism and the optimization algorithm according to claim 1 is characterized in that S4 include: S4.1: Calculate theoretically based on the simulation model; S4.2: Fine-tune according to the experience library. Since the numerical solution process is an iterative process and there are errors, it is impossible to obtain very accurate results for density, specific enthalpy and temperature parameters in the theoretical calculation process of S4.

1. Therefore, there is room for fine-tuning according to the results of the simulation model theoretical calculation. By establishing a debugging experience library, matching is performed for each pipeline, and further adjusting the type one parameters to further reduce the error.

8. The system-level twin model initialization method combining the mechanism and the optimization algorithm according to claim 1 is characterized in that S5 include: S5.1: Optimize the target selection strategy and select the sensor strategy result as the optimization target; S5.2: Optimize the weight setting strategy, set the optimization target value weight according to the sensor range and the accuracy required by the user.

9. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 8 is characterized in that: In S5.1, the unit measurement values ​​are selected from multiple angles such as redundant sensors, water vapor properties, and conservation principles to obtain the optimization target.

10. The system-level twin model initialization method combining the mechanism and optimization algorithm according to claim 1, characterized in that: In S6, S5 is started based on the result of S4, and the parameters optimized by S5 are brought into the mechanism model. S4 is executed again. After the simulation model reaches a steady state, the error is calculated and initialization is completed. The above steps are repeated until the error meets the accuracy requirements and is coupled with other systems.

Images