Rapid calculation method and system for reactor core power distribution of small power reactor

Abstract

The invention belongs to but is not limited to the technical field of reactor core power distribution, and discloses a small power reactor core power distribution rapid calculation method and system, and the method comprises the steps: selecting a proper physical calculation program, carrying out the modeling of a fuel assembly of a small power reactor core, designing proper input and output parameters, and constructing a power distribution data spectrum library; a full-connection neural network is selected to construct a data-driven power distribution calculation model; the training sample data set is substituted into the data-driven calculation model based on the full-connection neural network, multiple rounds of training are carried out, and a convergent data-driven calculation model is obtained; and performing test verification on the test sample set by using the convergence model to obtain a test result. The method provided by the invention has extremely high calculation speed, can achieve calculation precision similar to that of a physical calculation program of the reactor core, and can meet basic requirements of online supervision of the reactor core of the small reactor for the high-maneuverability ship.

Description

Technical Field

[0001] The present invention belongs to but is not limited to the technical field of core power distribution, and in particular relates to a method and system for quickly calculating core power distribution of a small power reactor. Background Art

[0002] The core size of a marine pressurized water reactor is small. In order to meet the mobility requirements of the ship, the operating conditions change more frequently, resulting in more serious distortion of the core power density distribution than in a nuclear power plant. The core operation process requires more focused supervision. However, due to factors such as security conditions, power distribution measurement points are usually not set inside the marine reactor core, or only a small number of measurement points are set, making it difficult to accurately control the core power distribution. Large-scale high-precision physical calculation programs can calculate the core power distribution more accurately based on the core structure and state parameters, but they usually consume a lot of computing resources and a long computing time, which is not conducive to online use.

[0003] Research status of data-driven core physics calculation methods:

[0004] The development of computer technology has led to a series of technological revolutions in various industries due to data-driven and artificial intelligence technologies, and has spawned the concept of new engineering. In the field of reactor physics calculations, data-driven and artificial intelligence solutions have also been gradually proposed by scholars and have received widespread attention. There are more and more applications of artificial intelligence in core power calculations. Many studies are no longer limited to simple machine learning, but have begun to be combined with some optimization algorithms, and the focus of application has also shown diversified development.

[0005] Li Wenhuai et al. proposed a method combining learning models with data assimilation, which provided new ideas and methods for nuclear reactor detection technology and improved the operating efficiency and safety of nuclear reactors. Wang Xingfang et al. introduced deep learning neural networks and a series of training optimization methods, learned a large number of power distributions under different control rod positions, and finally realized the prediction of the core power distribution of sodium-cooled fast reactors. Liu Dong et al. used a physical information-driven neural network algorithm to try to solve the multidimensional neutron diffusion equation, and combined a series of accelerated convergence measures and efficient parallel search technology for effective incremental coefficients (keff) to solve the above-mentioned equations and finally obtained some high-precision calculation results. In order to realize the digital twin of the reactor, Prantikos Konstantinos et al. used a physical information-driven neural network algorithm to solve the neutron diffusion equation under point reactor dynamics, and provided a more efficient solution for irregular problems and high-dimensional problems, verifying that the neural network algorithm has a great advantage in computational efficiency. At the same time, compared with traditional numerical solution methods, this method does not need to spend a lot of time on dividing more complex networks. Kharazmi Ehsand et al. further developed a physical information-driven neural network model with more precise effects, called the deep machine learning hp-VPINN model, which greatly improved the calculation accuracy. The results of this study have outstanding practicality.

[0006] The above research shows that the application of data-driven methods in the field of core physics calculations has been widely studied and has achieved considerable results. The research on data-driven core power distribution calculation agent models has a good theoretical foundation.

[0007] Through the above analysis, the problems and defects of the prior art are as follows:

[0008] (1) Due to the constraints of safeguard conditions and other factors, there are usually no power distribution measurement points set inside the core of a marine reactor, or only a small number of measurement points are set, making it difficult to accurately control the core power distribution. Large-scale high-precision physical calculation programs can accurately calculate the core power distribution based on the core structure and state parameters, but they usually consume a lot of computing resources and a long computing time, which is not conducive to online use.

[0009] (2) Small power reactors have small cores, frequent power changes, and more serious power distortion than nuclear power plants. However, there are few measurement points in the core, making it difficult to observe the core power distribution. High-precision physical calculation programs can calculate the core power distribution more accurately, but the poor real-time performance is not conducive to online use. Summary of the invention

[0010] In view of the problems existing in the prior art, the present invention provides a data-driven method and system for quickly calculating the core power distribution of a small power reactor.

[0011] The present invention is implemented as follows: a data-driven method for quickly calculating the core power distribution of a small power reactor comprises the following steps:

[0012] Step 1: Select a suitable physical calculation program, model the fuel assembly of the small power reactor core, design appropriate input and output parameters, and build a power distribution data spectrum library;

[0013] Step 2: Use a fully connected neural network to build a data-driven power distribution calculation model;

[0014] Step three, the training sample data set is brought into the data-driven computing model based on the fully connected neural network, and multiple rounds of training are performed to obtain a converged data-driven computing model; the converged model is then used to test and verify the test sample set to obtain the test results.

[0015] Further, the selection of physical computing programs and modeling methods:

[0016] The easily accessible and versatile component calculation program HELIOS is selected as the physical calculation program to construct the power distribution spectrum library.

[0017] Furthermore, the HELIOS program is divided into three modules, among which the AURORA module is a parameter input module, the ZENITH module is a parameter output module, and the HELIOS module is a calculation module; the HELIOS calculation module input card includes three parts: aurora-03.inp, aurora-31.inp and zenith.inp; the parameter input module AURORA edits the input material size, material type, kernel density, geometric distribution, boundary properties, burnup depth, etc. through input cards with different functions, and runs the program to output the required parameters.

[0018] Further, the design method of input and output parameters:

[0019] (1) Setting of the average core temperature range;

[0020] The core average temperature range is selected, and the temperature range is 300K~580K;

[0021] The temperature points of the reactor average temperature training sample set are: 300K, 350K, 400K, 450K, 480K, 490K, 500K, 510K, 520K, 530K, 540K, 550K, 560K, 570K, 580K, and the temperature points of the test set are: 485K, 505K, 525K, 545K, 565K;

[0022] (2) Reactor burnup depth interval setting;

[0023] The reactor burnup depth is set at the beginning, middle and end of its life.

[0024] Furthermore, taking the temperature and burnup settings into comprehensive consideration, different temperature points and different burnups are substituted into the fuel assembly power distribution calculation model constructed by the HELIOS program for calculation. Based on the calculated fuel assembly power distribution data set, a training data sample set and a test data sample set are constructed.

[0025] Furthermore, the method for constructing the data-driven power distribution calculation model is as follows:

[0026] The fully connected neural network (Fully Connected Neural Network) consists of an input layer, a hidden layer, and an output layer. The input layer consists of two node elements, corresponding to the input core temperature and burnup depth; the output layer has hundreds of node elements, corresponding to the power generated by hundreds of output fuel elements; the hidden layer consists of 64 node elements, which are used to express the main calculation logic.

[0027] Another object of the present invention is to provide a data-driven small power reactor core power distribution fast calculation system comprising:

[0028] The power distribution spectrum library construction module is used to select appropriate physical calculation programs, model the fuel assemblies of the small power reactor core, design appropriate input and output parameters, and build a power distribution data spectrum library;

[0029] Data-driven computing model design module, used to select fully connected neural network to build power distribution computing model;

[0030] The model testing and verification module is used to bring the training sample data set into the data-driven computing model based on the fully connected neural network, and conduct multiple rounds of training to obtain a converged data-driven computing model; then use the converged model to test and verify the test sample set to obtain the test results.

[0031] Another object of the present invention is to provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data-driven method for quickly calculating the power distribution of a small power reactor core.

[0032] Another object of the present invention is to provide a computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to execute the steps of the data-driven method for quickly calculating the power distribution of a small power reactor core.

[0033] Another object of the present invention is to provide an information data processing terminal, which includes the data-driven small power reactor core power distribution rapid calculation system.

[0034] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0035] First, the present invention aims at the practical problem that the high-precision core physics calculation program consumes a lot of computing power, takes a long time to calculate, and is not easy to use online. It proposes to rely on the reactor power distribution data spectrum library pre-calculated by the physical calculation program, design and train a fully connected artificial neural network proxy calculation model to replace the core physics calculation model, quickly calculate the core power distribution, and realize online support for core supervision during reactor operation. It has been verified by numerical experiments that this method has an extremely fast calculation speed and can achieve a calculation accuracy similar to that of the core physics calculation program, which can meet the basic needs of online supervision of the core of a highly maneuverable small-scale marine reactor.

[0036] Second, the technical solution of the present invention solves the problems of slow speed, insufficient accuracy and reliance on traditional physical calculation models in the calculation of core power distribution in the existing technology through a data-driven method for rapid calculation of core power distribution in small power reactors. Most of the existing methods for calculating core power distribution rely on physical simulation programs, and the calculation process is complex and requires a large amount of computing resources, especially showing inefficiency and delay when processing dynamic input data. Therefore, this traditional method is difficult to meet the needs of fast and accurate power distribution calculation of small power reactors in industrial applications.

[0037] The present invention achieves rapidity and high accuracy in power distribution calculation by introducing a fully connected neural network based on deep learning for data-driven calculation. Compared with traditional physical model calculation, the present invention shows significant technical progress in the following aspects:

[0038] 1. Significant improvement in computational efficiency: Through a data-driven approach, this method no longer relies on lengthy physical solution processes in the calculation of core power distribution. After training convergence, the neural network model can calculate new input data at an extremely high speed, greatly shortening the time for power distribution analysis and achieving real-time performance.

[0039] 2. Improved accuracy and stability: The present invention uses a large amount of power distribution data to train the neural network for multiple rounds to obtain a highly accurate, convergent and stable model. Test results show that the accuracy of the power distribution calculation results of this model is significantly higher than that of traditional methods, especially under complex loads and dynamic changes.

[0040] 3. High adaptability and flexibility: The neural network model of the present invention can adjust parameters according to the characteristics of small power reactors in the initial training to obtain the best calculation results. Its flexibility makes the technology applicable to small power reactors of different structures and sizes, achieving strong versatility.

[0041] 4. Real-time monitoring and optimization capability: Since the present invention can quickly calculate the power distribution of the core, the results can be used in real time to monitor and adjust the operating parameters of the small power reactor, which helps to achieve online optimization and operation management of the core. This real-time feedback capability significantly improves the safety and operating efficiency of small power reactors.

[0042] In summary, the present invention has achieved remarkable technical progress in computing efficiency, accuracy, flexibility and real-time monitoring, which helps to realize more intelligent and efficient small power reactor management and operation optimization in industry, and provides effective support for technological innovation in related industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of a data-driven small power reactor core power distribution rapid calculation method provided by an embodiment of the present invention.

[0044] Figure 2 It is a structural block diagram of a data-driven small power reactor core power distribution rapid calculation system provided by an embodiment of the present invention.

[0045] Figure 3 The invention provides a core power distribution calculation agent model algorithm system architecture design.

[0046] Figure 4 It is a neural network structure diagram provided by an embodiment of the present invention.

[0047] Figure 5 It is a comparison chart of the initial calculation results of the 485K life provided by the embodiment of the present invention.

[0048] Figure 6 1 is a comparison diagram of power distribution calculation errors at different burnup depths at 485K provided by an embodiment of the present invention; (a) beginning of life; (b) middle of life; (c) end of life.

[0049] Figure 7 1 is a comparison diagram of power distribution calculation errors at different temperatures at the beginning of the life provided by an embodiment of the present invention; (a) 525K; (b) 565K. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0051] like Figure 1 As shown, a data-driven small power reactor core power distribution fast calculation method provided by an embodiment of the present invention includes the following steps:

[0052] S101, construction of power distribution spectrum library;

[0053] S102, Data-driven computational model design;

[0054] S103, model testing and verification.

[0055] like Figure 2 As shown, an embodiment of the present invention provides a data-driven small power reactor core power distribution fast calculation system, including:

[0056] The power distribution spectrum library construction module is used to select appropriate physical calculation programs, model a group of fuel assemblies in a small reactor core, design appropriate input and output parameters, and calculate and construct a power distribution data spectrum library;

[0057] Data-driven computing model design module, used to select fully connected neural network to build power distribution computing model;

[0058] The model testing and verification module is used to construct a training data sample set and a test data sample set based on the calculated fuel assembly power distribution data set, bring the training data sample set into the designed data-driven computing model based on the fully connected neural network, conduct multiple rounds of training, obtain a converged model, and then use the converged model to test and verify the test sample set.

[0059] like Figure 3 As shown in the figure, the core power distribution calculation agent model design system consists of three parts: construction of power distribution spectrum library, data-driven calculation model design, and testing and verification of the calculation model.

[0060] 1. Construction of power distribution spectrum library

[0061] Select a suitable physical calculation program, model a group of fuel assemblies in a small reactor core, design appropriate input and output parameters, and calculate and construct a power distribution data spectrum library.

[0062] 1.1 Physical computing program selection and modeling

[0063] The easily accessible and versatile component calculation program HELIOS is selected as the physical calculation program to construct the power distribution spectrum library.

[0064] The HELIOS program has great adaptability in dealing with geometric problems. Its internal principle uses the subgroup method, which has high calculation accuracy for resonance processing and can handle most grid-based problems. Therefore, it is widely used in two-dimensional grid physical calculations. The HELIOS program is divided into three modules, of which the AURORA module is the module for input parameters, the ZENITH module is the module for output parameters, and the HELIOS module is the module for calculation. The HELIOS calculation module input card includes three parts: aurora-03.inp, aurora-31.inp and zenith.inp. The input module AURORA edits the input material size, material type, kernel density, geometric distribution, boundary properties, burnup depth, etc. through input cards with different functions, and runs the program to output the required parameters.

[0065] The HELIOS program was used to model a group of fuel assemblies for a small pressurized water reactor. The cross-section of the fuel assembly is a square structure, using rod-shaped UO2 fuel elements. The cladding material is zirconium alloy. Control rods are not considered to be placed in the elements. Each fuel assembly contains hundreds of fuel elements arranged in a matrix structure.

[0066] 1.2 Input and output parameter design

[0067] The average core temperature and reactor burnup are two physical calculation parameters that are easy to obtain in real time during reactor operation. Therefore, the algorithm mainly attempts to use the core temperature and reactor burnup depth as input parameters to calculate the power distribution in the fuel assembly.

[0068] (1) Setting of average core temperature range

[0069] The selection of the core average temperature range is mainly based on the consideration of covering the operating temperature range of common pressurized water reactors from cold to hot, that is, the temperature range of 30℃~300℃(303K~573K), and finally decided to select the temperature range of 300K~580K. When designing the step size, considering the actual situation that the monitoring demand for core power distribution is small in the cold state and the supervision demand is large in the hot state, a larger step size is designed in the cold state, and a smaller step size is designed in the hot state to meet the needs of model training.

[0070] The temperature points of the final reactor average temperature training sample set are: 300K, 350K, 400K, 450K, 480K, 490K, 500K, 510K, 520K, 530K, 540K, 550K, 560K, 570K, 580K, and the temperature points of the test set are: 485K, 505K, 525K, 545K, 565K.

[0071] (2) Reactor burnup depth range setting

[0072] The setting of reactor burnup depth mainly considers three typical situations: the beginning of life, the middle of life and the end of life.

[0073] Taking the temperature and burnup settings into consideration, different temperature points and different burnups were substituted into the fuel assembly power distribution calculation model constructed by the HELIOS program for calculation. A total of 45 sets of training data and 15 sets of test data were obtained, as shown in Tables 1 and 2.

[0074] Table 1. Training dataset design

[0075]

[0076]

[0077] Table 2 Test dataset design

[0078]

[0079] Each set of data consists of a pair of 2-dimensional input vectors and hundreds of dimensions of output vectors, as shown in the following formula.

[0080] 2. Data-driven computing model design

[0081] From the data structure, we can see that the problem that needs to be solved by using a neural network is a typical multi-input and multi-output highly complex machine learning problem. A fully connected neural network (Fully Connected Neural Network) is also called a multilayer perceptron (MLP). Each neuron in each layer is connected to all neurons in the previous layer, and each connection has a certain weight. This complex cross-structure can handle nonlinear separable problems well and has strong information expression capabilities. Therefore, a fully connected neural network is selected to construct a power distribution calculation model.

[0082] The fully connected neural network designed for this problem consists of three parts: input layer, hidden layer and output layer, such as Figure 4 As shown, the input layer consists of 2 node elements, corresponding to the input core temperature and burnup depth; the output layer has hundreds of node elements, corresponding to the power generated by hundreds of output fuel elements; the hidden layer consists of 64 node elements, which are used to express the main calculation logic.

[0083] The activation function of the node element uses the simple and efficient ReLU (Rectified Linear Unit) activation function. First, it is simple to calculate and has certain advantages in dealing with nonlinear problems. Second, there is no negative value in the input of this problem, that is, Dead ReLU will not occur.

[0084] The optimization algorithm uses the adaptive learning rate algorithm Adam (Adaptive Moment Estimation), which combines the advantages of momentum and RMSprop algorithms. It can converge quickly, has low computational cost and high computational efficiency. In addition, it can make corresponding adjustments according to different situations, which has great advantages in problems with more parameters.

[0085] Usually, the improvement targets of data-driven methods include accuracy, efficiency, robustness, and interpretability. Since the main task of the present invention is to quickly calculate the power distribution of fuel assemblies, the improvement targets mainly consider accuracy and efficiency.

[0086] Based on the calculated fuel assembly power distribution data set, training data sample sets and test data sample sets are constructed. First, the training data sample sets are brought into the designed data-driven computing model based on a fully connected neural network. Multiple rounds of training are performed to obtain a converged model. The converged model is then used to test and verify the test sample set.

[0087] 1. Data sample set construction

[0088] The data in the power distribution spectrum library is classified into training sample sets and test sets. This process strictly ensures the mutual exclusivity between different data sets and avoids the decrease in prediction accuracy caused by the crossover of information. On this basis, the data is processed with certain noise and some unreasonable or meaningless data is removed to ensure the quality of the data. In order to facilitate neural network recognition and improve training efficiency, the output (power value) is normalized to form a training and test data sample set for training and testing neural networks. Its structure is shown in Table 3.

[0089] Table 3 Training\test data sample structure

[0090]

[0091] 2. Test results analysis

[0092] First, the training sample data set was introduced into the designed fully connected neural network computing model, and 30,000 rounds of training were performed to obtain a converged data-driven computing model. After the convergence, the average relative error of the data-driven computing model was 2% and the mean square error was 0.0012.

[0093] Then, the test sample data set is brought into the fully connected neural network computing model for testing to obtain test results. The delay of all test calculation processes is within seconds, which can meet the online support requirements. Next, we will focus on comparative analysis of calculation accuracy.

[0094] Figure 5The core temperature is 485K, and the calculation results at the beginning of the life cycle are compared. The horizontal axis in the figure represents the fuel element number in the fuel assembly, the actual value curve in the vertical axis represents the nuclear power value of the fuel element calculated by the HELIOS program, and the predicted value curve represents the nuclear power value of the fuel element calculated by the data-driven model. It is not difficult to find that the two are very consistent through observation of the curves, with only a very small error at the inflection points such as the power peak and trough.

[0095] 2.1 Comparison of calculation results at different fuel consumption at the same temperature

[0096] When the core temperature is 485K, the calculation results of core power distribution at different burnup depths are compared and analyzed. Figure 6 shown. Figure 6 The horizontal axis of each curve represents the fuel element number in the fuel assembly, and the vertical axis represents the relative error of the power of each fuel element calculated by the data-driven calculation model relative to the nuclear power of the fuel element calculated by the HELIOS physical calculation program.

[0097] By comparing the calculation results at different burnup depths, it can be found that the maximum calculation error is about 1.5% at the beginning of the life, about 1.8% in the middle of the life, and about 1.6% at the end of the life. The calculation error first increases and then decreases with the increase of burnup depth, and the position of the maximum error is also not the same. However, the maximum calculation error of the three stages is less than 2%, which can meet the core supervision accuracy requirements during reactor operation.

[0098] 2.2 Comparison of calculation results at different temperatures for the same fuel consumption

[0099] The calculation errors of the two temperature points at 525K and 565K at the beginning of the life cycle are compared, as shown in Figure 7 As shown. Figure 7 It can be seen that under the same burnup, the maximum error of fuel element power calculation shows a slight increasing trend with the increase or decrease of core temperature, and the position of the maximum error is basically the same, but the maximum value of the relevant calculation error is less than 2%, which can meet the core supervision accuracy requirements during reactor operation.

[0100] In order to solve the practical problems that high-precision core physics calculation programs consume a lot of computing power and take a long time to calculate core power distribution, and are not easy to use online, this paper proposes a reactor power distribution data spectrum library pre-calculated by the physical calculation program, designs and trains a fully connected artificial neural network proxy calculation model to replace the core physics calculation model, quickly calculates the core power distribution, and realizes online support for core supervision during reactor operation. The method has been verified by numerical experiments to have an extremely fast calculation speed and can achieve a calculation accuracy similar to that of the core physics calculation program, which can meet the basic needs of online supervision of the core of highly maneuverable small-scale marine reactors.

[0101] The present invention provides a data-driven method for quickly calculating the power distribution of a small power reactor core. By combining physical modeling with deep learning, the power distribution of a small power reactor core can be quickly and accurately calculated. The detailed working principle of the method includes the following steps:

[0102] 1. Modeling and data library construction: First, select a physical calculation program suitable for the core of a small power reactor to physically model the fuel assembly inside the core. This modeling process determines the key structural and behavioral characteristics of the fuel assembly, and designs the necessary input parameters (such as fuel assembly material properties, geometric structure, etc.) and output parameters (such as power distribution). Based on this physical model, a large amount of power distribution sample data is generated, and finally a comprehensive power distribution data library is constructed. This data library provides the basic data required for subsequent neural network training and model construction.

[0103] 2. Construction of data-driven model: Based on the modeling data, a fully connected neural network is used to construct a power distribution calculation model. The fully connected neural network is a deep learning model with strong generalization ability, which is suitable for fitting nonlinear and high-dimensional data. In order to adapt to the complex physical characteristics of small power reactors, the neural network is designed with a multi-layer structure and a multi-node connection method, so that the model can effectively capture the complex characteristics of the core power distribution during the learning process.

[0104] 3. Model training and convergence: Combine the constructed data-driven model with the sample data in the data library obtained in step 1 to train the model. By bringing a large amount of training sample data into the neural network, gradually adjusting the weights and biases of the model, using the loss function to evaluate the error, and continuously iterating and optimizing the model. Through multiple rounds of training and parameter adjustment, the model will eventually converge, that is, the error reaches the preset range, and a stable power distribution calculation model with high accuracy is obtained.

[0105] 4. Model verification and optimization: Use the converged calculation model to verify the test sample set to confirm the generalization ability and accuracy of the model. During the test, the accuracy and applicability of the model are evaluated by comparing the calculation results with the actual power distribution data. If the error does not meet the standard, it is necessary to return to the previous step to readjust and train the model to ensure that the final model achieves the best performance.

[0106] 5. Fast calculation and application: After completing the test verification, the converged model can be used for fast calculation of the actual core power distribution. Given new input parameters, the data-driven model can output accurate power distribution results in a short time. This fast calculation method greatly reduces the time for step-by-step solution based on physical calculation programs, and improves the calculation efficiency of the core power distribution of small power reactors.

[0107] 6. Result output and adjustment suggestions: The calculated power distribution results can be used to monitor and optimize core performance. Combined with real-time data acquisition and power distribution adjustment requirements, the model provides efficient computing support, provides a reliable basis for real-time optimization of core power distribution, and helps improve the safety and operating efficiency of the core.

[0108] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. It can be understood by a person of ordinary skill in the art that the above-mentioned devices and methods can be implemented using computer executable instructions and / or contained in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium. The device and its modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0109] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principle of the present invention should be covered by the protection scope of the present invention.

Claims

1. A data-driven method for rapid calculation of core power distribution of a small power reactor, characterized in that: The following steps are involved: Step 1: Select a suitable physical calculation program, model the fuel assembly of the small power reactor core, design appropriate input and output parameters, and build a power distribution data spectrum library; Step 2: Use a fully connected neural network to build a data-driven power distribution calculation model; Step 3: The training sample data set is introduced into the data-driven computing model based on the fully connected neural network, and multiple rounds of training are performed to obtain a converged data-driven computing model; The convergence model is then used to test and verify the test sample set to obtain the test results.

2. The data-driven small power reactor core power distribution rapid calculation method according to claim 1, characterized in that: Selection of physical computing programs and modeling methods: The easily accessible and versatile component calculation program HELIOS is selected as the physical calculation program to construct the power distribution spectrum library.

3. The data-driven small power reactor core power distribution rapid calculation method according to claim 2, characterized in that: The HELIOS program is divided into three modules, among which the AURORA module is a parameter input module, the ZENITH module is a parameter output module, and the HELIOS module is a calculation module; the HELIOS calculation module input card includes three parts: aurora-03.inp, aurora-31.inp and zenith.inp; the parameter input module AURORA edits the input material size, material type, kernel density, geometric distribution, boundary properties, and burnup depth through input cards with different functions, and runs the program to output the required parameters.

4. The data-driven small power reactor core power distribution rapid calculation method according to claim 1, characterized in that: Design method of input and output parameters: (1) Setting of the average core temperature range; The core average temperature range is selected, and the temperature range is 300K~580K; The temperature points of the reactor average temperature training sample set are: 300K, 350K, 400K, 450K, 480K, 490K, 500K, 510K, 520K, 530K, 540K, 550K, 560K, 570K, 580K, and the temperature points of the test set are: 485K, 505K, 525K, 545K, 565K; (2) Reactor burnup depth interval setting; The reactor burnup depth is set at the beginning, middle and end of its life.

5. The data-driven small power reactor core power distribution rapid calculation method according to claim 4, characterized in that: Taking the temperature and burnup settings into comprehensive consideration, different temperature points and different burnups are substituted into the fuel assembly power distribution calculation model constructed by the HELIOS program for calculation. Based on the calculated fuel assembly power distribution data set, training data sample sets and test data sample sets are constructed.

6. The data-driven small power reactor core power distribution rapid calculation method according to claim 1, characterized in that: The method for constructing the data-driven power distribution calculation model: The fully connected neural network (Fully Connected Neural Network) consists of an input layer, a hidden layer, and an output layer. The input layer consists of two node elements, corresponding to the input core temperature and burnup depth; the output layer has hundreds of node elements, corresponding to the power generated by hundreds of output fuel elements; the hidden layer consists of 64 node elements, which are used to express the main calculation logic.

7. A data-driven small power reactor core power distribution fast calculation system according to any one of claims 1 to 6, characterized in that: include: The power distribution spectrum library construction module is used to select appropriate physical calculation programs, model the fuel assemblies of the small power reactor core, design appropriate input and output parameters, and build a power distribution data spectrum library; Data-driven computing model design module, used to select fully connected neural network to build power distribution computing model; The model testing and verification module is used to bring the training sample data set into the data-driven computing model based on the fully connected neural network, and conduct multiple rounds of training to obtain a converged data-driven computing model; The convergence model is then used to test and verify the test sample set to obtain the test results.

8. A computer device, characterized in that: The computer device includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the data-driven small power reactor core power distribution rapid calculation method as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, which, when executed by a processor, enables the processor to perform the steps of the data-driven method for quickly calculating the core power distribution of a small power reactor as described in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that: The information data processing terminal includes the data-driven small power reactor core power distribution rapid calculation system as described in claim 7.

Images