CN114462336B - Method for calculating average temperature of coolant of main pipeline of nuclear reactor - Google Patents

Method for calculating average temperature of coolant of main pipeline of nuclear reactor Download PDF

Info

Publication number
CN114462336B
CN114462336B CN202210371397.XA CN202210371397A CN114462336B CN 114462336 B CN114462336 B CN 114462336B CN 202210371397 A CN202210371397 A CN 202210371397A CN 114462336 B CN114462336 B CN 114462336B
Authority
CN
China
Prior art keywords
main pipeline
temperature
model
coolant
average temperature
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202210371397.XA
Other languages
Chinese (zh)
Other versions
CN114462336A (en
Inventor
周新志
孙梓云
刘艺璇
朱加良
何正熙
徐涛
董晨龙
刘丹会
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sichuan University
Original Assignee
Sichuan University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Sichuan University filed Critical Sichuan University
Priority to CN202210371397.XA priority Critical patent/CN114462336B/en
Publication of CN114462336A publication Critical patent/CN114462336A/en
Application granted granted Critical
Publication of CN114462336B publication Critical patent/CN114462336B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/28—Design optimisation, verification or simulation using fluid dynamics, e.g. using Navier-Stokes equations or computational fluid dynamics [CFD]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2113/00—Details relating to the application field
    • G06F2113/08—Fluids
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/08—Thermal analysis or thermal optimisation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
    • G06F2119/14—Force analysis or force optimisation, e.g. static or dynamic forces

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • Computer Hardware Design (AREA)
  • General Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Geometry (AREA)
  • Molecular Biology (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Algebra (AREA)
  • Fluid Mechanics (AREA)
  • Mathematical Analysis (AREA)
  • Mathematical Optimization (AREA)
  • Pure & Applied Mathematics (AREA)
  • Monitoring And Testing Of Nuclear Reactors (AREA)

Abstract

The invention relates to a method for calculating the average temperature of coolant of a main pipeline of a nuclear reactor, which comprises the following steps: establishing a combined simulation model according to the acquired geometric parameters by using fluid dynamics simulation software; setting condition data for the combined simulation model, and calculating the temperature distribution of the coolant of the main pipeline; extracting temperature values at thermocouple group positions from main pipeline coolant temperature distribution as input of a machine learning model, extracting average temperature values of 200 points uniformly distributed on a cross section where the thermocouple groups are located as output of the machine learning model, and forming a mapping relation between the input and the output; selecting a plurality of thermocouple groups with different sections in the main pipeline, establishing a data set by using the formed mapping relation, dividing the data set into a training set, a testing set and a verification set, and inputting the training set into a convolutional neural network for training so as to obtain an average temperature calculation model. This scheme can promote measurement accuracy under the prerequisite that does not change original temperature measuring device.

Description

Method for calculating average temperature of coolant of main pipeline of nuclear reactor
Technical Field
The invention relates to the technical field of temperature measurement, in particular to a method for calculating the average temperature of a coolant of a main pipeline of a nuclear reactor.
Background
The coolant in the main pipe of the reactor absorbs the heat energy of the reactor core and transfers the heat energy to the two-loop steam generator, so that the temperature measurement of the main pipe is very important for the normal operation and detection of the nuclear power plant. However, the temperature stratification of the coolant in the main pipeline exists, and the measured value difference of the average temperature in the pipeline is large and the accuracy is low due to the arrangement of the number and the orientation of different temperature measuring devices.
The existing method for measuring the temperature of a main pipeline of a nuclear reactor is to take an average value of the temperatures measured by a plurality of thermocouples as the integral average temperature of a coolant in the main pipeline. However, the number and the orientation of the temperature measuring devices are limited by actual environmental conditions, the influence of temperature stratification cannot be avoided, the existing main pipeline temperature measuring devices are installed, and a large amount of cost is consumed by readjusting.
Disclosure of Invention
The invention aims to improve the calculation accuracy of average temperature on the premise of not changing the original temperature measuring device, and provides a method for calculating the average temperature of a coolant of a main pipeline of a nuclear reactor.
In order to achieve the above object, the embodiments of the present invention provide the following technical solutions:
a method for calculating the average temperature of coolant of a main pipeline of a nuclear reactor comprises the following steps:
s1, acquiring geometric parameters of the reactor core and the main pipeline, and establishing a combined simulation model according to the acquired geometric parameters by using fluid dynamics simulation software; setting condition data for the combined simulation model, and calculating the temperature distribution of the coolant of the main pipeline;
step S2, extracting temperature values at thermocouple group positions from the main pipeline coolant temperature distribution as the input of a machine learning model, and extracting the average temperature values of 200 points uniformly distributed on the cross section of the thermocouple group as the output of the machine learning model, thereby forming a mapping relation between the input and the output;
step S3, selecting a plurality of thermocouple groups with different sections in the main pipeline, establishing a data set by using the formed mapping relation, dividing the data set into a training set, a testing set and a verification set, inputting the training set into a convolutional neural network for training, thereby obtaining an average temperature calculation model.
The step S1 specifically includes the following steps:
s1-1, acquiring geometric parameters of a reactor core and a main pipeline to construct a geometric model; the geometric parameters of the reactor core comprise the boundary shape of the reactor core and the boundary shape of the reactor core assembly, and the geometric parameters of the main pipeline comprise the boundary shape of the main pipeline, the length of the main pipeline and the diameter of the main pipeline;
step S1-2, carrying out mesh division on the geometric model, and obtaining a mesh model after determining the optimal mesh division mode and the optimal mesh division quantity through sensitivity analysis;
step S1-3, after setting condition data for the grid model by using fluid dynamics simulation software, carrying out steady state calculation, and simulating to obtain a temperature distribution result, wherein the temperature distribution result is the temperature distribution of the coolant of the main pipeline; the set condition data includes initial condition data, boundary condition data, and turbulent flow condition data.
Further comprising step S1-4: after the temperature distribution of the coolant of the main pipeline is obtained through simulation, error analysis is carried out on a temperature simulation value and a temperature measured value corresponding to the thermocouple arrangement position in the real main pipeline, parameters of the combined simulation model are adjusted by combining fluid dynamics simulation software until the error meets a set requirement, so that a fluid dynamics simulation software network model is obtained, and the temperature distribution of the coolant of the main pipeline is obtained through calculation by using the fluid dynamics simulation software network model.
The specific steps of the step S1-3 are as follows: setting initial condition data by using fluid dynamics simulation software; substituting boundary condition data into the grid model, wherein the boundary condition data comprises reactor core inlet flow distribution, flow size, reactor core outlet pressure and reactor core outlet temperature; and selecting turbulent flow condition data, and simulating to obtain the temperature distribution of the coolant of the main pipeline after steady-state calculation.
The step S2 specifically includes the following steps:
step S2-1, assuming that M thermocouples are arranged in each main pipeline, extracting temperature values of the M thermocouples from the temperature distribution of the coolant of the main pipeline as the input of a machine learning model;
step S2-2, extracting the average temperature value of 200 uniformly distributed points on the section where the M thermocouples are located as the average temperature value of the section, and taking the average temperature value as the output of the machine learning model;
and step S2-3, forming a mapping relation between the M thermocouple temperature values and the average temperature value of 200 points uniformly distributed on the cross section where the M thermocouple temperature values are located.
The step S3 specifically includes the following steps:
s3-1, selecting a plurality of thermocouple groups arranged on different sections in the main pipeline to obtain a plurality of groups of mapping relations, using the plurality of groups of mapping relations as data sets, and dividing the data sets into a training set, a testing set and a verification set according to the ratio of 8:1: 1;
step S3-2, inputting a training set into a convolutional neural network for training, selecting a mean square error for a loss function, monitoring an average absolute error in the training process, confirming a network hyperparameter by using a K-fold cross validation method, and obtaining an average temperature calculation model after convergence is finished;
step S3-3: the mean temperature calculation model was validated and tested using a validation set and a test set.
In the scheme, the temperature values at the positions of M thermocouples in the mapping relations and the average temperature value of 200 points uniformly distributed on the cross section where the thermocouple is located are used as a data set.
Compared with the prior art, the invention has the beneficial effects that:
according to the scheme, condition data of a reactor core and a main pipeline are set according to fluid dynamics knowledge, main pipeline coolant temperature distribution under various working conditions is obtained through simulation, temperature values of a plurality of fixed points (such as 4 thermocouples) are used as input on the basis of a Convolutional Neural Network (CNN), the average temperature value of 200 points uniformly distributed on the section where the fixed points are located is used as output, and an average temperature calculation model is established; therefore, in practical application, the average temperature calculated by the temperature of a plurality of points of the corresponding section is obtained through the average temperature calculation model by utilizing the temperature measurement data of the finite points, and the average temperature of the coolant in the main pipeline is calculated more accurately.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required in the embodiments will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and those skilled in the art can also obtain other related drawings based on the drawings without inventive efforts.
FIG. 1 is a flow chart of a method of calculating an average temperature according to the present invention;
FIG. 2 is a schematic cross-sectional view of 10 embodiments of the present invention;
fig. 3 is a schematic diagram of the training result of the average temperature calculation model in embodiment 2 of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. The components of embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the present invention, presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments of the present invention without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that: like reference numbers and letters refer to like items in the following figures, and thus, once an item is defined in one figure, it need not be further defined and explained in subsequent figures. Also, in the description of the present invention, the terms "first", "second", and the like are used for distinguishing between descriptions and not necessarily for describing a relative importance or implying any actual relationship or order between such entities or operations.
Example 1:
the invention is realized by the following technical scheme, as shown in figure 1, a method for calculating the average temperature of a coolant of a main pipeline of a nuclear reactor comprises the following steps:
s1, acquiring geometric parameters of the reactor core and the main pipeline, and establishing a combined simulation model according to the acquired geometric parameters by using fluid dynamics simulation software; and after condition data is set for the combined simulation model, the temperature distribution of the coolant in the main pipeline is calculated.
The final purpose of the scheme is to obtain an average temperature calculation model after training is completed, and in order to meet subsequent training requirements and improve the accuracy of the average temperature calculation model, the temperature distribution of the coolant section of each main pipeline in the nuclear reactor under a large number of different working conditions needs to be obtained.
The method comprises the steps of establishing a combined simulation model, wherein because the temperature distribution of the coolant section of the main pipeline of the real nuclear reactor is difficult to directly and accurately obtain, a plurality of groups of data are generated by utilizing the simulation of Fluid dynamic simulation software (CFD) to replace real data. When a combined simulation model is established, simulation is completed through construction of a multilayer model, and the combined simulation model comprises a geometric model and a grid model.
Specifically, it is well known that there are multiple main pipes in a nuclear reactor, but for convenience of explanation, the present scheme is first described by taking one main pipe as an example, and first, geometric parameters of the core and the main pipe are obtained to construct a geometric model, wherein the geometric parameters include, but are not limited to, a core boundary shape, a core assembly boundary shape, a main pipe length, a main pipe diameter, and the like. The reactor core assembly can comprise an upper reactor core plate, a hanging basket barrel, an outlet nozzle, an upper support assembly, a control rod guide barrel assembly, a heat pipe section, a temperature measuring device and the like.
And then, carrying out meshing on the geometric model, and determining the optimal meshing mode and the optimal meshing number through sensitivity analysis to obtain a mesh model.
And finally, setting initial condition data, boundary condition data and turbulence condition data for the grid model by using fluid dynamics simulation software. Because the nuclear reactor has different working conditions during actual operation, when the initial condition data and the boundary condition data of the grid model are set, the boundary condition data of the reactor core under different working conditions are brought into the grid model by using fluid dynamics simulation software to serve as corresponding input conditions, wherein the boundary condition data comprise reactor core inlet flow distribution, flow size, reactor core outlet pressure, reactor core outlet temperature and the like.
After turbulent condition data are selected, when steady state calculation is carried out, the following continuity equation, mass conservation differential equation, momentum conservation differential equation and energy conservation equation are established for describing the thermodynamic hydraulic process, so that the temperature distribution of the coolant of the main pipeline is obtained.
The continuity equation is established as follows:
Figure DEST_PATH_IMAGE002
（1）
Figure DEST_PATH_IMAGE004
in the formula (1), rho is the density of the fluid in the main pipeline,
Figure DEST_PATH_IMAGE006
is the fluid velocity vector, B is the fluid micelle volume, wherein
Figure DEST_PATH_IMAGE008
，
Figure DEST_PATH_IMAGE010
u, v, w are respectively fluid edges
Figure DEST_PATH_IMAGE012
The velocity components of the three coordinate axes of x, y and z,
Figure DEST_PATH_IMAGE014
、
Figure DEST_PATH_IMAGE016
、
Figure DEST_PATH_IMAGE018
three coordinate axes of x, y and z respectivelyThe unit vector of (2).
From the perspective of fluid micelles, the differential form of the established conservation-of-mass equation is as follows:
Figure DEST_PATH_IMAGE020
（2）
the differential form of the momentum conservation equation is established as follows:
Figure DEST_PATH_IMAGE022
（3）
Figure DEST_PATH_IMAGE024
（4）
Figure DEST_PATH_IMAGE026
（5）
in the formula (3), the formula (4) and the formula (5),
Figure DEST_PATH_IMAGE028
fxto act on the volumetric force per unit mass in a plane perpendicular to the x-axis,
Figure DEST_PATH_IMAGE028A
fyto act on a volumetric force per unit mass in a plane perpendicular to the y-axis,
Figure DEST_PATH_IMAGE028AA
fzis the volumetric force per unit mass acting in a plane perpendicular to the z-axis; τ is a shear stress component, τxxRepresenting a shear stress component of force acting in a plane perpendicular to the x-axis and directed in the direction of the x-axis, τyxRepresenting a shear stress component of force acting in a plane perpendicular to the y-axis and directed in the direction of the x-axis, τzxRepresenting the component of shear stress directed in the direction of the x-axis of the force acting on a plane perpendicular to the z-axis, τyyRepresenting a shear stress component of force acting on a plane perpendicular to the y-axis and directed in the direction of the y-axis, τxyRepresenting a shear stress component of force acting in a plane perpendicular to the x-axis and directed in the direction of the y-axis, τzyRepresenting a shear stress component of force acting in a plane perpendicular to the z-axis and directed in the direction of the y-axis, τzzRepresenting a shear stress component of force acting in a plane perpendicular to the z-axis and directed in the direction of the z-axis, τxzRepresenting a shear stress component of force acting in a plane perpendicular to the x-axis and directed in the direction of the z-axis, τyzRepresenting a shear stress component of force acting in a plane perpendicular to the y-axis and directed in the z-axis direction; p is the pressure at the center of the fluid micelle.
The established energy conservation equation is as follows:
Figure DEST_PATH_IMAGE030
（6）
in the formula (6), the reaction mixture is,
Figure DEST_PATH_IMAGE032
is the fluid internal energy per unit mass;
Figure DEST_PATH_IMAGE034
heat radiation heat exchange received by unit mass fluid in unit time;
Figure DEST_PATH_IMAGE036
is the coefficient of thermal conductivity; t is the fluid micelle temperature; wherein
Figure DEST_PATH_IMAGE038
。
In detail, since turbulence affects both mass and energy transfer, appropriate turbulence condition data needs to be selected to quantify the turbulence viscosity. The analysis can be assisted based on a standard k-epsilon turbulence model, and turbulence condition data solves an equation of turbulence momentum k and dissipation rate ɛ thereof, and the expression is as follows:
Figure DEST_PATH_IMAGE040
（7）
Figure DEST_PATH_IMAGE042
（8）
in the formulas (7) and (8), k represents turbulent momentum, and ɛ represents the dissipation rate of the turbulent momentum;
Figure DEST_PATH_IMAGE044
Gkrepresenting the turbulent kinetic energy generated by the laminar velocity gradient,
Figure DEST_PATH_IMAGE046
Gbrepresenting the kinetic energy of the turbulence generated by buoyancy,
Figure DEST_PATH_IMAGE048
Ymrepresenting fluctuations due to diffusion of transitions in compressible turbulence,
Figure DEST_PATH_IMAGE050
Skand
Figure DEST_PATH_IMAGE052
Sɛrepresenting user-defined data, C1ɛ、C2ɛ、C3ɛIs a constant;
Figure DEST_PATH_IMAGE054
Figure DEST_PATH_IMAGE056
and
Figure DEST_PATH_IMAGE058
Figure DEST_PATH_IMAGE060
is the momentum of turbulent flow
Figure DEST_PATH_IMAGE062
The turbulence prandtl number of the k equation and the dissipation rate ɛ equation;
Figure DEST_PATH_IMAGE064
represents the kinetic viscosity coefficient;
Figure DEST_PATH_IMAGE066
denotes the turbulent viscosity coefficient, defined as
Figure DEST_PATH_IMAGE068
Figure DEST_PATH_IMAGE070
，
Figure DEST_PATH_IMAGE072
Is a constant;
Figure DEST_PATH_IMAGE074
、
Figure DEST_PATH_IMAGE076
tensor representations in the coordinate directions respectively;
Figure DEST_PATH_IMAGE078
is a tensor representation of fluid velocity.
The calculation aiming at the equation is converted into discrete equation calculation, namely, the modeling area is subjected to grid division by using fluid dynamics simulation software to realize discretization, the temperature distribution of a plurality of sections of the main pipeline is obtained, and therefore the complete main pipeline coolant temperature distribution is obtained.
After the temperature distribution of the coolant of the main pipeline is obtained through simulation, temperature values (temperature simulation values) corresponding to M thermocouple positions of a real pipeline and temperature values (temperature true values) measured by the thermocouples of the real pipeline under the working condition are extracted from the temperature distribution of the coolant of the main pipeline to carry out error analysis, and grid division, condition data and the like are adjusted by combining with fluid dynamics simulation software until the errors meet the engineering requirements, so that a final fluid dynamics simulation software network model is obtained. The fluid dynamics simulation software network model can be used for further simulating more accurate main pipeline coolant temperature distribution, and the simulation process is the same as that described above, so that the description is omitted. The temperature distribution of the coolant in the main pipeline simulated by the network model of the fluid dynamics simulation software can be regarded as being in accordance with the reality, and can be used for acquiring training data by the following steps.
And step S2, extracting temperature values at thermocouple group positions from the main pipeline coolant temperature distribution as the input of a machine learning model, and extracting the average temperature value of 200 points uniformly distributed on the cross section of the thermocouple group as the output of the machine learning model, thereby forming a mapping relation between the input and the output.
In the simulation process of step S1, a plurality of main pipe coolant temperature distributions are generated by using different condition data, for example, there are 15 different working conditions in practical application, that is, 15 different condition data can be set, so that in step S1, 15 main pipe coolant temperature distributions can be obtained.
And extracting the temperature value at the thermocouple group position from the main pipeline coolant temperature distribution output by the simulation as the input of the machine learning model. For example, if 4 thermocouples (respectively, the 1 st thermocouple, the 2 nd thermocouple, the 3 rd thermocouple, and the 4 th thermocouple shown in fig. 2) are provided in a main pipe, the temperature values at the positions where the 4 thermocouples are provided are extracted from the coolant temperature distribution of the main pipe output by simulation and are used as the input of the machine learning model. Referring to the direction of the arrows in fig. 2, 4 thermocouples may be disposed at a distance of 0.05m from the outlet of the pipe or at a distance of 0.4m from the outlet of the pipe in the axial direction of the main pipe. Assuming that 4 thermocouples may be disposed at any one of 10 positions in the axial direction of the main pipe, 10 sets of thermocouple temperature data may be obtained, each set including temperature values of 4 thermocouples.
Then, the average temperature values of 200 points which are uniformly distributed on the cross section where the 4 thermocouples are located are extracted as the average temperature values of the cross section, please refer to fig. 2 which is an axial schematic diagram of a certain main pipe, if the 4 thermocouples may be disposed at any one of 10 positions of the main pipe, 10 sets of thermocouple temperature data may be obtained. That is, 10 positions of the main pipe correspond to 10 sections, and 4 points on each section represent the positions where 4 thermocouples are provided.
In this example, the 1 st cross-section is 0.05m from the outlet of the pipe, the 2 nd cross-section is 0.1m from the outlet of the pipe, and so on, and the 10 th cross-section is 0.5m from the outlet of the pipe. And (3) extracting the average value of the temperature of 200 points uniformly distributed on each section as the output of the machine learning model, thereby forming a mapping relation between the temperature at the position of the thermocouple group of the main pipeline and the average value of the temperature of 200 points uniformly distributed on the section where the thermocouple group is located.
Step S3, selecting a plurality of thermocouple groups with different sections in the main pipeline, establishing a data set by using the formed mapping relation, dividing the data set into a training set, a testing set and a verification set, inputting the training set into a convolutional neural network for training, thereby obtaining an average temperature calculation model.
The input of the machine learning model in the mapping of step S2 is the temperature value of the thermocouple group, representing the characteristic value of the input data, and the output of the machine learning model is the cross-sectional average temperature value, representing the target value of the output.
Next, as illustrated above, assuming that there are 15 different working conditions in the practical application, that is, 15 different condition data are set, step S1 can obtain 15 main pipe coolant temperature distributions, and step S2 extracts temperature values at the positions where 4 thermocouples are set (there are 10 possible setting positions) from the 15 main pipe coolant temperature distributions as the input of the machine learning model, that is, 150 sets of temperature data (each set of temperature data includes temperature values of 4 thermocouples) are used as the input of the machine learning model.
Assume that the temperature values at the 4 thermocouple locations are (T) respectively1,T2,T3,T4) The average temperature value of 200 points uniformly distributed on the cross section of the 4 thermocouples is (T)average) Then there is a mapping relation (T)1,T2,T3,T4)→(Taverage) Then 150 sets of temperature data may generate 150 sets of mapping relationships.
Such a pair of mapping relationships represents a set of data, and in order to improve the accuracy of the average temperature calculation model for subsequent training, the scheme generates a large number of such mapping relationships. The traditional calculation method directly takes the average value of 4 thermocouple temperature values as the average temperature of the section, but is influenced by the temperature stratification of the main pipeline, and the temperature measurement precision of the method is low. Therefore, the mapping relation established by the scheme is the average temperature value of 200 points uniformly distributed on each section, and various working conditions, a plurality of main pipelines and a plurality of sections are considered, so that the calculation accuracy of the average temperature calculation model is superior to that of 4 thermocouple temperature values and the average value is directly obtained.
And taking the multiple groups of mapping relations as a data set, and dividing the data set into a training set, a testing set and a verification set according to the proportion of 8:1: 1. The data in the training set is used to train the model, the data in the test set is used to monitor errors in the training process to prevent over-training, and the data in the validation set is used to evaluate the accuracy of the model obtained by training.
According to the scheme, a Convolutional Neural Network (CNN) is selected as a training model, and the Convolutional Neural network can construct a larger, richer and multidimensional assumed space compared with other machine learning regression models, so that the empirical risk (local) of the model is closer to the expected risk (global). Moreover, tests show that compared with other machine learning regression models, the convolutional neural network has higher training speed and stability in convergence, the trained model has better generalization performance, and can be quickly adapted to other regression tasks of the same type under the condition of modifying a few network hyper-parameters.
Training a training set by using a convolutional neural network, adjusting parameters of the convolutional neural network, determining an optimal optimizer of the model by using a K-fold cross-validation method, training loss function parameters, and training a monitoring index function, wherein the optimal optimizer selects an Adam optimizer, the loss function selects a Mean Square Error (MSE), monitoring an average absolute error (MAE) in the training process, properly adjusting and optimizing the parameters of the model, and obtaining an average temperature calculation model after convergence is completed. And finally, verifying and testing the trained average temperature calculation model by using a verification set and a test set.
In this step, the Convolutional Neural Network (CNN) adopts a three-layer fully-connected layer structure, and the fully-connected layer connection formula is:
Figure DEST_PATH_IMAGE080
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE082
the ith neuron representing the l-th fully-connected layer,
Figure DEST_PATH_IMAGE084
a jth neuron representing a l-1 th fully-connected layer;
Figure DEST_PATH_IMAGE086
Figure DEST_PATH_IMAGE088
to represent
Figure DEST_PATH_IMAGE090
Figure DEST_PATH_IMAGE082A
And
Figure DEST_PATH_IMAGE092
Figure DEST_PATH_IMAGE084A
the connection weight of (1);
Figure DEST_PATH_IMAGE094
Figure DEST_PATH_IMAGE096
to represent
Figure DEST_PATH_IMAGE082AA
And
Figure DEST_PATH_IMAGE084AA
the connection bias term of (1).
The formula of the training optimization function of the Adam optimizer is as follows:
Figure DEST_PATH_IMAGE098
Figure DEST_PATH_IMAGE100
Figure DEST_PATH_IMAGE102
wherein the content of the first and second substances,
Figure DEST_PATH_IMAGE104
wkrepresents the connection weight of the full connection layer after the kth learning, wk-1Showing the connection weight of the full connection layer after the kth-1 th learning;
Figure DEST_PATH_IMAGE106
Figure DEST_PATH_IMAGE108
in order to learn the step size,
Figure DEST_PATH_IMAGE110
is the learning rate;
Figure DEST_PATH_IMAGE112
is the exponential moving average of the k-th learning gradient,
Figure DEST_PATH_IMAGE114
for the squared gradient of the kth learning,
Figure DEST_PATH_IMAGE116
is the exponential moving average of the k-1 learning gradient,
Figure DEST_PATH_IMAGE118
squared gradient for the k-1 th study;
Figure DEST_PATH_IMAGE120
Figure DEST_PATH_IMAGE122
and
Figure DEST_PATH_IMAGE124
are respectively as
Figure DEST_PATH_IMAGE126
Figure DEST_PATH_IMAGE112A
And
Figure DEST_PATH_IMAGE114A
the number of corrections of (d);
Figure DEST_PATH_IMAGE128
and
Figure DEST_PATH_IMAGE130
in order to control the exponential decay constant of the,
Figure DEST_PATH_IMAGE132
represent
Figure DEST_PATH_IMAGE128A
To the power of k of (a),
Figure DEST_PATH_IMAGE134
to represent
Figure DEST_PATH_IMAGE130A
To the k power of;
Figure DEST_PATH_IMAGE136
Figure DEST_PATH_IMAGE138
the first derivative (gradient).
The training loss function is formulated as:
Figure DEST_PATH_IMAGE140
the formula of the training monitoring index function is as follows:
Figure DEST_PATH_IMAGE142
wherein, a and a' respectively represent a target value and a predicted value output by the convolutional neural network; n is the total number of training samples, and i represents the ith training sample.
And step S4, taking the M thermocouple temperature values measured in the real pipeline as the input of an average temperature calculation model, and outputting the average temperature value of the coolant of the main pipeline by the average temperature calculation model.
Example 2:
this embodiment is the experimental verification process of the above embodiment 1, in order to simplify the experimental process, only one main pipeline simulation process is set up, and since most of the actual main pipeline coolant inlet temperature distributions under different working conditions have the characteristic of irregular single-peak or double-peak distribution, in order to construct more assumed spaces as much as possible to fit the actual temperature distributions, the inlet boundary condition data of the pipeline is simulated by using the fluid dynamics simulation software, and a large number of single-peak or double-peak function models t = f with different parameters are usedi(x, y) to fit the boundary condition data, wherein fiA unimodal or bimodal function model representing the different parameters, (x, y) the coordinates of the inlet cross section, and t the coolant temperature at that point of the coordinates.
In the experiment, 50 single-peak temperature fields and 40 double-peak temperature fields are fitted as condition data of a hydrodynamic simulation software simulation pipeline under different working conditions, that is, a total of 90 working conditions are used as input of a machine learning model, the temperature distribution of a main pipeline coolant simulated under each working condition is a 10-group temperature average value generated by 10 sections (one section is taken every 0.05 m) closest to an outlet (the temperature distribution of the section of the pipeline is more uniform as the outlet is closer, and the training of the model is facilitated), and please refer to a schematic diagram of fig. 2 for extracting 10 sections. Each set of data includes (T)1,T2,T3,T4,Taverage) Thus, a total of (50 + 40) × 10=900 sets of data is generated.
In order to reduce the difference of the input data characteristics of the convolutional neural network and make model training easier, each characteristic of the input data is subjected to standardization processing. The processed training data was divided into 810 training sets and 90 test sets (using the 90 test sets as validation sets) in a 9:1 ratio.
A convolutional neural network with a three-layer full-connection layer structure is constructed by using prior experience, and network parameters are shown in a table 1:
TABLE 1
Figure DEST_PATH_IMAGE144
The convolutional neural network uses an Adam optimizer, a Mean Square Error (MSE) is selected by a loss function, an average absolute error (MAE) is monitored in the training process, then a K-fold cross-validation method is used for confirming network hyper-parameters, if the number of training rounds (epochs) is set to be 200, the batch size (batch size) is set to be 64, and the training result is shown in figure 3, so that the convergence speed is high and stable.
In the verification and test process, the average temperature value of 200 uniformly distributed points on the cross section is used as the true value of the average temperature of the cross section, and the results of predicting 90 groups of verification sets by using a trained average temperature calculation model and the results of directly averaging the data characteristics of the verification sets and the results of error analysis of the true value are shown in table 2:
TABLE 2
Figure DEST_PATH_IMAGE146
According to experiments, compared with the traditional method for calculating the average temperature of the coolant of the main pipeline, the method can achieve higher measurement precision, can improve the measurement precision on the premise of not changing the original temperature measuring device, and has great engineering significance.
The above description is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of the changes or substitutions within the technical scope of the present invention, and all the changes or substitutions should be covered within the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (3)

1. A method for calculating the average temperature of a coolant of a main pipeline of a nuclear reactor is characterized by comprising the following steps: the method comprises the following steps:
s1, acquiring geometric parameters of the reactor core and the main pipeline, and establishing a combined simulation model according to the acquired geometric parameters by using fluid dynamics simulation software; setting condition data for the combined simulation model, and calculating the temperature distribution of the coolant of the main pipeline;
step S2, extracting temperature values at the thermocouple group position from the main pipe coolant temperature distribution as the input of a machine learning model, and extracting the average temperature value of 200 points uniformly distributed on the section where the thermocouple group is located as the output of the machine learning model, thereby forming a mapping relation between the input and the output;
step S3, selecting a plurality of thermocouple groups with different sections in a main pipeline, establishing a data set by using the formed mapping relation, dividing the data set into a training set, a testing set and a verification set, and inputting the training set into a convolutional neural network for training so as to obtain an average temperature calculation model;
the step S1 specifically includes the following steps:
s1-1, acquiring geometric parameters of a reactor core and a main pipeline to construct a geometric model; the geometric parameters of the reactor core comprise the boundary shape of the reactor core and the boundary shape of the reactor core assembly, and the geometric parameters of the main pipeline comprise the boundary shape of the main pipeline, the length of the main pipeline and the diameter of the main pipeline;
step S1-2, carrying out mesh division on the geometric model, and obtaining a mesh model after determining the optimal mesh division mode and the optimal mesh division quantity through sensitivity analysis;
step S1-3, after setting condition data for the grid model by using fluid dynamics simulation software, carrying out steady state calculation, and simulating to obtain a temperature distribution result, wherein the temperature distribution result is the temperature distribution of the coolant of the main pipeline; the set condition data comprises initial condition data, boundary condition data and turbulence condition data;
further comprising step S1-4: after the temperature distribution of the coolant of the main pipeline is obtained through simulation, carrying out error analysis on a temperature simulation value and a temperature measured value corresponding to the thermocouple arrangement position in the real main pipeline, adjusting parameters of a combined simulation model by combining fluid dynamics simulation software until the error meets a set requirement so as to obtain a fluid dynamics simulation software network model, and calculating the temperature distribution of the coolant of the main pipeline by using the fluid dynamics simulation software network model;
the specific steps of the step S1-3 are as follows: setting initial condition data by using fluid dynamics simulation software; substituting boundary condition data into the grid model, wherein the boundary condition data comprises reactor core inlet flow distribution, flow size, reactor core outlet pressure and reactor core outlet temperature; and selecting turbulent flow condition data, and simulating to obtain the temperature distribution of the coolant of the main pipeline after steady-state calculation.
2. The method for calculating the average temperature of the coolant of the main pipeline of the nuclear reactor according to claim 1, wherein: the step S2 specifically includes the following steps:
step S2-1, assuming that M thermocouples are arranged in each main pipeline, extracting temperature values of the M thermocouples from the temperature distribution of the coolant of the main pipeline as the input of a machine learning model;
step S2-2, extracting the average temperature value of 200 uniformly distributed points on the section where the M thermocouples are located as the average temperature value of the section, and taking the average temperature value as the output of the machine learning model;
and step S2-3, forming a mapping relation between the M thermocouple temperature values and the average temperature value of 200 points uniformly distributed on the cross section where the M thermocouple temperature values are located.
3. The method for calculating the average temperature of the coolant of the main pipeline of the nuclear reactor according to claim 1, wherein: the step S3 specifically includes the following steps:
s3-1, selecting a plurality of thermocouple groups arranged on different sections in the main pipeline to obtain a plurality of groups of mapping relations, using the plurality of groups of mapping relations as data sets, and dividing the data sets into a training set, a testing set and a verification set according to the ratio of 8:1: 1;
step S3-2, inputting a training set into a convolutional neural network for training, selecting a mean square error for a loss function, monitoring an average absolute error in the training process, confirming a network hyperparameter by using a K-fold cross validation method, and obtaining an average temperature calculation model after convergence is finished;
step S3-3: the mean temperature calculation model was validated and tested using a validation set and a test set.
CN202210371397.XA 2022-04-11 2022-04-11 Method for calculating average temperature of coolant of main pipeline of nuclear reactor Active CN114462336B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210371397.XA CN114462336B (en) 2022-04-11 2022-04-11 Method for calculating average temperature of coolant of main pipeline of nuclear reactor

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210371397.XA CN114462336B (en) 2022-04-11 2022-04-11 Method for calculating average temperature of coolant of main pipeline of nuclear reactor

Publications (2)

Publication Number Publication Date
CN114462336A CN114462336A (en) 2022-05-10
CN114462336B true CN114462336B (en) 2022-06-24

Family

ID=81417590

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210371397.XA Active CN114462336B (en) 2022-04-11 2022-04-11 Method for calculating average temperature of coolant of main pipeline of nuclear reactor

Country Status (1)

Country Link
CN (1) CN114462336B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117558472B (en) * 2024-01-11 2024-03-15 深圳大学 Nuclear reactor cooling system and cooling control method thereof

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108846190A (en) * 2018-06-05 2018-11-20 哈尔滨工程大学 A kind of nuclear heat coupling simulation method of PWR fuel assembly
CN110289111A (en) * 2019-07-09 2019-09-27 上海交通大学 The flow of natural convection loop determines method, apparatus and server

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104102782B (en) * 2014-07-18 2017-05-24 上海电力学院 Method for controlling temperatures of reactors of pressurized water reactor nuclear power stations by aid of RBF (radial basis function) neural networks
CN111882044B (en) * 2020-08-05 2021-09-14 四川大学 Eutectic prediction method and deep learning framework based on graph neural network
CN112632873B (en) * 2020-12-30 2022-10-28 西安交通大学 Method for determining position arrangement mode of measuring points of temperature thermocouple of nuclear reactor hot leg
CN113076684B (en) * 2021-02-23 2022-09-20 中国核动力研究设计院 Intelligent calculation method for transient parameters in rod adjusting process of nuclear reactor core
CN113486483B (en) * 2021-07-12 2022-12-09 西安交通大学 Reactor small-break multi-dimensional coupling analysis method

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108846190A (en) * 2018-06-05 2018-11-20 哈尔滨工程大学 A kind of nuclear heat coupling simulation method of PWR fuel assembly
CN110289111A (en) * 2019-07-09 2019-09-27 上海交通大学 The flow of natural convection loop determines method, apparatus and server

Also Published As

Publication number Publication date
CN114462336A (en) 2022-05-10

Similar Documents

Publication Publication Date Title
EP3944125A1 (en) Method and system for manufacturing a heat exchanger for supercritical pressure fluid
CN102419827B (en) Radial basis function (RBF) neural network-based boiling heat exchanging prediction method
CN106897520B (en) Heat transfer system reliability analysis method containing fuzzy parameters
CN114462336B (en) Method for calculating average temperature of coolant of main pipeline of nuclear reactor
CN106528990A (en) Hypersonic velocity pointed conical appearance heat flux density modeling approach based on functional optimization
Vidanović et al. Validation of the CFD code used for determination of aerodynamic characteristics of nonstandard AGARD-B calibration model
CN111144054A (en) Modeling method for natural circulation characteristic of villiaumite cooling high-temperature reactor passive waste heat discharge system
CN114444413A (en) Sub-channel-level three-dimensional thermal hydraulic analysis method for plate-shaped fuel reactor core
CN117189713A (en) Hydraulic system fault diagnosis method based on digital twin driving
Zhu et al. An integral type µ synthesis method for temperature and pressure control of flight environment simulation volume
Smith The difference between traditional experiments and CFD validation benchmark experiments
CN113990539B (en) Ultrasonic measurement reconstruction method for temperature distribution of coolant in hot section of main pipeline of nuclear reactor
CN114912364A (en) Natural gas well flow prediction method, device, equipment and computer readable medium
Thummar et al. Machine learning for vortex flowmeter design
WO2024051724A1 (en) Flow-induced vibration test apparatus and method, computer device, storage medium, and product
Aguirre et al. Numerical analysis for detecting head losses in trifurcations of high head in hydropower plants
CN117090831A (en) Hydraulic system fault diagnosis framework with twinning application layer
CN113536640B (en) Optimization design method for internal flow channel structure of air distributor based on orthogonal test
Cheng et al. Robust optimization and uncertainty quantification of a micro axial compressor for unmanned aerial vehicles
CN115470726A (en) Hypersonic inlet channel flow field rapid prediction method based on deep learning
CN115238416A (en) Flow prediction method and system for fuel atomizing nozzle of aircraft engine
CN111861011B (en) Supercritical pressure fluid convection heat exchange performance prediction method and system
CN114818377A (en) Heat exchange equipment life loss detection method and device and electronic equipment
CN110750756B (en) Real-time on-line instrument checksum diagnosis method through optimal support vector machine algorithm
Tinoco et al. Numerical simulation of industrial flows

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant