Method and program for predicting reactor core characteristics, and apparatus for predicting reactor core characteristics

Abstract

This technology provides a way to efficiently, quickly, and accurately predict and evaluate the output distribution and control rod value, even if they cannot be measured simultaneously, by using the measurement result of one measurement to predict the other measurement result. [Solution] A method for predicting core characteristics, comprising: a core characteristics calculation step of calculating core characteristics including power distribution and control rod value in multiple fuel loading patterns; a core characteristics evaluation step of evaluating core characteristics based on the calculation results; a training data creation step of creating training data for core characteristics based on the evaluation results; a dataset creation step of creating a dataset by extracting power distribution and control rod value from the training data; a correlation model creation step of creating a correlation model between power distribution and control rod value by performing machine learning using the dataset; and a prediction calculation program creation step of creating a prediction calculation program by incorporating the correlation model, wherein the prediction calculation program outputs the other by inputting measured values ​​of power distribution or control rod value.

Description

Technical Field

[0001] The present invention relates to a method for predicting and calculating core characteristics in a nuclear reactor, a prediction calculation program, and a device for predicting and calculating core characteristics.

Background Art

[0002] In order to safely operate a nuclear reactor at a predetermined output, it is essential to measure various core characteristics (for example, output distribution, neutron flux distribution, reactivity distribution, assembly burnup distribution, control rod worth, etc.) before startup and during operation to confirm that there are no problems with the core and fuel.

[0003] When measuring these core characteristics, it may take time or require special operations, but safety and rationalization are always required. Usually, it is preferable to have no special operations because it reduces disturbances and enhances safety. There are also core characteristics that are difficult to measure during normal operation, and for such core characteristics, they are managed not by measured values but by values evaluated in the core design before startup.

[0004] For example, during the operation of a nuclear reactor, the measurement of two particularly important core characteristics, namely, the output distribution and the control rod worth of the nuclear reactor, cannot currently be measured simultaneously and are measured separately.

[0005] As a specific example, in a pressurized water reactor (PWR), the measurement of the output distribution can be performed by inserting a movable detector called MD (Movable Detector) from outside the reactor into the fuel assemblies loaded inside the reactor. However, at that time, if the core is disturbed by inserting a control rod or the like, the output distribution will change, so it is difficult to measure other core characteristics during the measurement of the output distribution.

[0006] Regarding the measurement of the control rod worth, if disturbances are given by performing other operations, the pure control rod worth cannot be measured. Similarly, it is difficult to measure other core characteristics during the measurement of the control rod worth.

[0007] Thus, since the two core characteristics that cannot be measured simultaneously must be measured separately for each core characteristic, the measurement process takes a long time. As a result, currently, only the core characteristics that are essential to confirm for reactor operation are measured, and measurements of other core characteristics that are only for reference are sometimes not performed.

[0008] Under these circumstances, the present inventors have proposed a core calculation technique that, for the same core characteristics, obtains the results of a second calculation program based on a different physical model from the results of a first calculation program based on a certain physical model (Patent Document 1). [Prior art documents] [Patent Documents]

[0009] [Patent Document 1] Patent No. 7252100 [Overview of the project] [Problems that the invention aims to solve]

[0010] However, the technologies described above are only applicable to the same core characteristics and do not apply to different core characteristics such as the power distribution and control rod value mentioned above. There is a need for core characteristic prediction technologies that can efficiently, quickly, and accurately determine different core characteristics that cannot be measured simultaneously. In particular, there is a need for core characteristic prediction technologies that can efficiently, quickly, and accurately determine the reactor power distribution and control rod value, which are important for reactor operation and require time to measure.

[0011] Therefore, the object of the present invention is to provide a core characteristic prediction calculation technology that can predict and evaluate reactor power distribution and control rod value, which cannot be measured simultaneously with high accuracy, in a short time and with high accuracy. [Means for solving the problem]

[0012] The inventors of the present invention have diligently studied how to solve the above problems and have found that the above problems can be solved by the invention described below.

[0013] The invention described in claim 1 is, A method for predicting reactor core characteristics, which involves measuring one core characteristic and predicting the other core characteristic for two core characteristics that cannot be measured simultaneously in a nuclear reactor, The two core characteristics mentioned above are the power distribution and the control rod value. A core characteristics calculation step of calculating, for each of a plurality of fuel loading patterns, a plurality of core characteristics including at least the power distribution and the control rod value, based on a predetermined core calculation program, Based on the obtained calculation results, a core characteristic evaluation process is performed to evaluate each core characteristic, Based on the evaluation results obtained, a training data creation process is performed to create training data for each core characteristic, A dataset creation step involves extracting the output distribution and the control rod value from the obtained training data and creating a dataset of the output distribution and the control rod value, A correlation model creation step involves creating a correlation model between the output distribution and the control rod value by performing machine learning using the created dataset, The system includes a process for creating a predictive calculation program that incorporates the created correlation model, This is a method for predicting core characteristics, characterized in that by inputting a measured value of the power distribution or control rod value obtained from an operating reactor into the prediction calculation program, the program outputs the other control rod value or power distribution, which is used as the predicted value for evaluation.

[0014] The invention described in claim 2 is, One of the aforementioned multiple core characteristics is the power distribution, control rod value, neutron flux distribution, reaction rate distribution, or aggregate burnup distribution. In the step of creating the dataset, the teacher data of any one of the neutron flux distribution, reaction rate distribution, and assembly burnup degree distribution, or the teacher data created by combining two or more teacher data is created as a dataset together with the output distribution and the teacher data of the control rod worth, which is the method for predicting and calculating core characteristics according to claim 1.

[0015] The invention according to claim 3 is The method for predicting and calculating core characteristics according to claim 2, wherein the neutron flux distribution is two or more groups of neutron flux distributions classified based on the energy of neutrons.

[0016] The invention according to claim 4 is The method for predicting and calculating core characteristics according to claim 2, wherein the neutron flux distribution is the neutron flux distribution of the energy group absorbed by the element for neutron absorption used in the control rod.

[0017] The invention according to claim 5 is The method for predicting and calculating core characteristics according to any one of claims 1 to 4, wherein the machine learning is performed using deep learning by a multi-layer neural network.

[0018] The invention according to claim 6 is The method for predicting and calculating core characteristics according to any one of claims 1 to 4, wherein the nuclear reactor is a pressurized water reactor (PWR).

[0019] The invention according to claim 7 is A program for predicting and calculating core characteristics for predicting one core characteristic by measuring the other core characteristic for two core characteristics that cannot be measured simultaneously in a nuclear reactor, wherein the two core characteristics are an output distribution and a control rod worth, A core characteristic calculation step of calculating, for each of a plurality of fuel loading patterns, at least a plurality of core characteristics including the output distribution and the control rod worth, respectively, based on a predetermined core calculation program; A core characteristic evaluation step of evaluating each core characteristic based on the obtained calculation results; A teacher data creation step of creating teacher data for each core characteristic based on the obtained evaluation results; A data set creation step of extracting the output distribution and the control rod worth from the obtained teacher data and creating a data set of the output distribution and the control rod worth; A correlation model creation step of creating a correlation model between the output distribution and the control rod worth by performing machine learning using the created data set, and It is configured to incorporate the created correlation model and input a measured value of the output distribution or the control rod worth obtained from an operating nuclear reactor to output the other control rod worth or output distribution as a predicted value for evaluation. A prediction calculation program for core characteristics is characterized by this.

[0020] The invention according to claim 8 is One of the plurality of core characteristics is an output distribution, a control rod worth, a neutron flux distribution, a reaction rate distribution, or an assembly burnup distribution, In the data set creation step, teacher data of any one of the neutron flux distribution, the reaction rate distribution, and the assembly burnup distribution, or teacher data created by combining two or more pieces of teacher data is combined with the teacher data of the output distribution and the control rod worth to create a data set. The prediction calculation program for core characteristics according to claim 7 is characterized by this.

[0021] The invention according to claim 9 is The neutron flux distribution is two or more groups of neutron flux distributions classified based on the energy of neutrons. The prediction calculation program for core characteristics according to claim 8 is characterized by this.

[0022] The invention according to claim 10 is The core characteristics prediction calculation program according to claim 8 is characterized in that the neutron flux distribution is the neutron flux distribution of the energy group absorbed by the neutron-absorbing elements used in the control rods.

[0023] The invention described in claim 11 is, The core characteristics prediction calculation program according to any one of claims 7 to 10 is characterized in that the machine learning is performed using deep learning with a multilayer neural network.

[0024] The invention described in claim 12 is, A reactor core characteristic prediction calculation device that predicts and calculates the other reactor core characteristic by measuring one of the two reactor core characteristics that cannot be measured simultaneously in a nuclear reactor, The system is equipped with a core characteristics prediction calculation program as described in any one of claims 7 to 10, An input means for inputting a previously measured output distribution or control rod value measurement, The core characteristic prediction calculation device is characterized by comprising an output means that outputs the result calculated by the core characteristic prediction calculation program as a predicted value for the other core characteristic. [Effects of the Invention]

[0025] According to the present invention, it is possible to provide a core characteristic prediction calculation technique that can predict and evaluate reactor power distribution and control rod value with high accuracy in a short time, even if these cannot be measured simultaneously with high accuracy. [Brief explanation of the drawing]

[0026] [Figure 1] This flowchart shows the procedure for a method for predicting core characteristics according to one embodiment of the present invention. [Figure 2] This flowchart shows the procedure for a method of predicting core characteristics according to another embodiment of the present invention. [Modes for carrying out the invention]

[0027] [1] Background leading to the completion of the present invention 1. Problems with conventional technology As mentioned above, the technology presented in Patent Document 1, proposed by the present inventors, is a technology that targets the same core characteristics and does not target different core characteristics such as power distribution and control rod value. Furthermore, this technology targets calculated values ​​and does not target measured values.

[0028] Furthermore, since the measurement methods for core characteristics differ for each core characteristic, it was not thought that it would be possible to obtain measured values ​​for other core characteristics using different measurement methods by combining calculated values ​​for multiple core characteristics with measured values ​​for a particular core characteristic.

[0029] 2. Basic Concept of the Invention Under these circumstances, the inventors focused on the fact that, among several core characteristics, there is a strong physical correlation between the power distribution and the control rod value.

[0030] In other words, the power distribution, a core characteristic that shows the spatial distribution of thermal output within the reactor core, is a measurement of the fission rate distribution (spatial distribution of fission rate) for each fuel assembly and has a strong physical correlation with the neutron flux distribution. On the other hand, the control rod value is a measurement of the reactivity of the control rods (effectiveness of the control rods) and similarly has a strong physical correlation with the neutron flux distribution. Thus, since both the power distribution and the control rod value have a strong physical correlation with the neutron flux distribution, it can be considered that the two have a strong physical correlation through the neutron flux distribution.

[0031] The physical correlation between this power distribution and the value of control rods—that is, the tendency for the value of control rods to increase at locations with high power (high neutron flux)—has been known for some time, but it had only been used to confirm the general trend.

[0032] The inventors, through their research, found that when data information related to other factors such as neutron flux distribution, reaction rate distribution, and aggregate burnup distribution is compiled into a database along with data on power distribution and control rod value, it is possible to accurately predict one of the two data points, control rod value or power distribution, from the other. However, in this case, processing a vast amount of data is required, making it difficult to find a highly accurate correlation between power distribution and control rod value, two core characteristics.

[0033] Therefore, the inventors conducted further intensive studies and came up with the idea of ​​employing machine learning in processing this data. When they actually performed prediction calculations of core characteristics, they confirmed that if they could obtain the measurement result of one of the output distributions or control rod values, they could predict the other measurement result with a reasonably high degree of accuracy, thus completing the present invention.

[0034] In other words, the core characteristics prediction calculation method according to the present invention is a method for predicting core characteristics in which, for two core characteristics that cannot be measured simultaneously in a nuclear reactor, namely power distribution and control rod value, one core characteristic is measured and the other core characteristic is predicted and calculated. A core characteristics calculation step in which, for each of a plurality of fuel loading patterns, a plurality of core characteristics including at least the power distribution and the control rod value are calculated based on a predetermined core calculation program, Based on the obtained calculation results, a core characteristic evaluation process is performed to evaluate each core characteristic, Based on the evaluation results obtained, a training data creation process is performed to create training data for each core characteristic, A dataset creation step involves extracting the output distribution and the control rod value from the obtained training data and creating a dataset of the output distribution and the control rod value, A correlation model creation step involves creating a correlation model between the output distribution and the control rod value by performing machine learning using the created dataset, The process includes a step for creating a predictive calculation program that incorporates the created correlation model. Then, by inputting the power distribution or control rod value measurements obtained from an operating reactor into this prediction calculation program, the program outputs the other control rod value or power distribution, which is used as the predicted value for evaluation.

[0035] Furthermore, the core characteristics prediction calculation program according to the present invention is a core characteristics prediction calculation program for predicting the other core characteristics by measuring one core characteristic, for two core characteristics, power distribution and control rod value, which cannot be measured simultaneously in a nuclear reactor. A core characteristics calculation step in which, for each of a plurality of fuel loading patterns, a plurality of core characteristics, including at least the power distribution and the control rod value, are calculated based on a predetermined core calculation program, Based on the obtained calculation results, a core characteristic evaluation step is performed to evaluate each core characteristic, Based on the evaluation results obtained, a training data creation step is performed to create training data for each core characteristic, A dataset creation step involves extracting the output distribution and the control rod value from the obtained training data and creating a dataset of the output distribution and the control rod value, The system includes a correlation model creation step, which involves performing machine learning using the created dataset to create a correlation model between the output distribution and the control rod value. The system is then configured to incorporate this created correlation model and, by inputting either the power distribution or control rod value measurements obtained from an operating reactor, output the other control rod value or power distribution, which is then used as a predicted value for evaluation.

[0036] Furthermore, the core characteristic prediction calculation device according to the present invention is a core characteristic prediction calculation device that predicts and calculates the other core characteristic by measuring one of two core characteristics, power distribution and control rod value, which cannot be measured simultaneously in a nuclear reactor, and is equipped with the above-mentioned core characteristic prediction calculation program. It also includes an input means for inputting pre-measured power distribution or control rod value measurements, and an output means for outputting the result calculated by the core characteristic prediction calculation program as a predicted value for the other core characteristic.

[0037] [2] Embodiment The present invention will be described in detail below based on embodiments.

[0038] 1. Method for predicting core characteristics Figure 1 is a flowchart showing the procedure for a method of predicting reactor core characteristics according to one embodiment of the present invention. Figure 2 is a flowchart showing the procedure for a method of predicting reactor core characteristics according to another embodiment of the present invention.

[0039] Note that in Figures 1 and 2, the only difference is whether power distribution and control rod value are used as core characteristics when creating the dataset from the training data (Figure 1), or whether other core characteristics are also used (Figure 2). The flowchart flow is the same. Note that in Figure 2, neutron flux distribution, reaction rate distribution, and aggregate burnup distribution are listed as other core characteristics, but these are not limiting, and other core characteristics may be used. The following describes each step separately.

[0040] (1) Training data creation process Prior to creating training data, first, for each of the multiple fuel loading patterns, at least multiple core characteristics, including power distribution and control rod value, are calculated based on a predetermined (known) core calculation program (core characteristic calculation step: not shown), and each core characteristic is evaluated (core characteristic evaluation step: not shown).

[0041] Subsequently, training data for each core characteristic is created based on the obtained evaluation results. This training data can be obtained by organizing and accumulating the evaluation results of each core characteristic under various fuel loading patterns using various core calculation programs.

[0042] Furthermore, publicly known core calculation programs can be used. Here, various core calculation programs can be based on the same physical model, or programs based on different physical models can be used depending on the core characteristics.

[0043] Furthermore, the "multiple fuel loading patterns" mentioned above include those with different fuel arrangements within the reactor, as well as those with the same fuel arrangement but different core conditions. Here, core conditions can be exemplified by core thermal output, control rod position, cycle burnup, etc.

[0044] Furthermore, as shown in Figure 1, only two core characteristics, power distribution and control rod value, may be selected. However, as shown in Figure 2, neutron flux distribution, reaction rate distribution, aggregate burnup distribution, etc., may also be selected in addition to power distribution and control rod value. By calculating and evaluating many core characteristics in addition to power distribution and control rod value, and using the accumulated data as training data along with power distribution and control rod value, it is possible to predict the desired core characteristics with higher accuracy. In particular, since neutron flux distribution has a strong correlation with power distribution and control rod value as described above, it is preferable to use it as it enables more accurate prediction calculations.

[0045] The training data for each core characteristic that has been created can be stored in a memory device and used in conjunction with new training data for each core characteristic created based on newly obtained evaluation results for each core characteristic to create a new prediction program.

[0046] (2) Dataset creation process Next, using an extraction program, the output distribution and control rod values ​​are extracted from the obtained training data to create a dataset of output distribution and control rod values. A publicly known program can be used as the extraction program.

[0047] In this case, if there is training data for neutron flux distribution, reaction rate distribution, and aggregate burnup distribution, it is preferable to use one of these training data sets, or training data created by combining two or more training data sets, together with the training data for power distribution and control rod value to form a dataset.

[0048] In the above, the neutron flux distribution and reaction rate distribution are 10 with respect to neutron energy. -5 Since the values ​​differ for each energy level across a wide energy range from eV to 10 MeV, it is preferable to have two or more neutron flux distributions classified based on the neutron energy.

[0049] Specifically, two groups of neutron flux distributions can be cited as examples: fast neutron flux distributions and thermal neutron flux distributions. Fast neutron flux distributions are neutron flux distributions with relatively high energy, while thermal neutron flux distributions are neutron flux distributions with relatively low energy. Furthermore, three groups of neutron flux distributions can be cited as examples: fast neutron flux distributions, thermal neutron flux distributions, and epithermal neutron flux distributions. Exothermic neutron flux distributions are neutron flux distributions with slightly higher energy than thermal neutron flux distributions. By classifying neutron flux distributions into more groups, more precise evaluation results can be obtained.

[0050] Furthermore, the neutron flux distribution may be the neutron flux distribution of the energy group absorbed by the neutron-absorbing elements used in the control rods. The neutron-absorbing elements are not particularly limited and include, for example, boron (B), silver (Ag), indium (In), cadmium (Cd), hafnium (Hf), etc.

[0051] (3) Correlation model creation process Next, using the created dataset, machine learning is performed with a pre-built-in learning program to obtain the correlation between the output distribution and the control rod value, thereby creating a correlation model. Here, conventionally used learning devices and learning programs can be used. It is preferable to use deep learning with a multilayer neural network for machine learning. This allows anyone, regardless of their skill level, to find the patterns in the correlation between the output distribution and the control rod value with high accuracy and create a correlation model, even when it is difficult to grasp the underlying principles.

[0052] (4) Prediction calculation process Then, by inputting measured values ​​of the power distribution or control rod value obtained from an operating reactor into a prediction calculation program that incorporates a correlation model, the other control rod value or power distribution can be output as a predicted value for evaluation. In this case, as mentioned above, since the power distribution and control rod value are correlated with high accuracy, even if the power distribution and control rod value of a reactor cannot be measured simultaneously, one can be measured efficiently, and the other can be predicted and evaluated with high accuracy in a short time.

[0053] Specifically, while control rod values ​​are generally measured during reactor startup, this result can be used to accurately predict and evaluate the measured power distribution without having to measure the power distribution itself.

[0054] Furthermore, while the power distribution is measured periodically during reactor operation, this result can be used to accurately predict and evaluate the control rod value without having to measure it in the reactor during operation.

[0055] Furthermore, the core characteristic prediction calculation method described above can be applied to any type of reactor, including fast reactors, boiling water reactors (BWRs), and pressurized water reactors (PWRs).

[0056] The evaluation results of the control rod value and power distribution output as predicted values ​​in the present invention are predicted based on the correlation with the measured values ​​of the other control rod value or power distribution, and are therefore considered to be closer to the values ​​that would be obtained if measured, compared to the control rod value and power distribution output as calculation results of conventional core calculation programs. Accordingly, it is considered that the present invention can obtain equivalent evaluation results in a shorter time compared to when both the control rod value and power distribution are measured.

[0057] 2. Core characteristic prediction calculation program and prediction calculation device By creating a program where each step includes the process of creating a correlation model for each of the above processes, and by incorporating the created correlation model and inputting the measured values ​​of the power distribution or control rod value obtained from an operating reactor, the program can output the other control rod value or power distribution and use it as a predicted value for evaluation, thereby creating a program for predicting core characteristics.

[0058] Furthermore, by incorporating this core characteristic prediction calculation program, and providing an input means for inputting pre-measured power distribution or control rod value measurements, and an output means for outputting the results calculated by the core characteristic prediction calculation program as a predicted value for the other core characteristic, it can be made into a core characteristic prediction calculation device.

[0059] By using such a core characteristics prediction calculation device to perform prediction calculations of core characteristics, it is possible to efficiently and quickly grasp and evaluate the reactor's power distribution and control rod value with high accuracy, regardless of the engineer's skill level, thereby greatly contributing to the safe operation of the reactor.

[0060] Although the present invention has been described above based on embodiments, the present invention is not limited to the above embodiments. Various modifications can be made to the above embodiments within the same and equivalent scope as the present invention.

Claims

1. A method for predicting reactor core characteristics, which involves measuring one core characteristic and predicting the other core characteristic for two core characteristics that cannot be measured simultaneously in a nuclear reactor, The two core characteristics mentioned above are the power distribution and the control rod value. A core characteristics calculation step of calculating, for each of a plurality of fuel loading patterns, a plurality of core characteristics including at least the power distribution and the control rod value, based on a predetermined core calculation program, Based on the obtained calculation results, a core characteristic evaluation process is performed to evaluate each core characteristic, Based on the evaluation results obtained, a training data creation process is performed to create training data for each core characteristic, A dataset creation step involves extracting the output distribution and the control rod value from the obtained training data and creating a dataset of the output distribution and the control rod value, A correlation model creation step involves creating a correlation model between the output distribution and the control rod value by performing machine learning using the created dataset, The system includes a process for creating a predictive calculation program that incorporates the created correlation model, A method for predicting core characteristics, characterized in that, by inputting a measured value of the power distribution or control rod value obtained from an operating reactor into the prediction calculation program, the program outputs the other control rod value or power distribution, which is used as the predicted value for evaluation.

2. One of the aforementioned multiple core characteristics is the power distribution, control rod value, neutron flux distribution, reaction rate distribution, or aggregate burnup distribution. The method for predicting core characteristics according to claim 1, characterized in that, in the dataset creation step, training data created by combining one of the neutron flux distribution, reaction rate distribution, and aggregate burnup distribution, or two or more training data, is created as a dataset together with the training data for the power distribution and control rod value.

3. The method for predicting core characteristics according to claim 2, characterized in that the neutron flux distribution is a neutron flux distribution of two or more groups classified based on the energy of neutrons.

4. The method for predicting core characteristics according to claim 2, characterized in that the neutron flux distribution is the neutron flux distribution of the energy group absorbed by the neutron-absorbing elements used in the control rods.

5. The method for predicting reactor core characteristics according to any one of claims 1 to 4, characterized in that the machine learning is performed using deep learning with a multilayer neural network.

6. The method for predicting core characteristics according to any one of claims 1 to 4, characterized in that the reactor is a pressurized water reactor (PWR).

7. A core characteristic prediction calculation program for predicting the core characteristic of a nuclear reactor, which involves measuring one core characteristic and calculating the other core characteristic, for two core characteristics that cannot be measured simultaneously in a nuclear reactor. The two core characteristics mentioned above are the power distribution and the control rod value. A core characteristics calculation step in which, for each of a plurality of fuel loading patterns, a plurality of core characteristics, including at least the power distribution and the control rod value, are calculated based on a predetermined core calculation program, Based on the obtained calculation results, a core characteristic evaluation step is performed to evaluate each core characteristic, Based on the evaluation results obtained, a training data creation step is performed to create training data for each core characteristic, A dataset creation step involves extracting the output distribution and the control rod value from the obtained training data and creating a dataset of the output distribution and the control rod value, The system includes a correlation model creation step, which involves performing machine learning using the created dataset to create a correlation model between the output distribution and the control rod value. A core characteristics prediction calculation program characterized by incorporating a created correlation model and taking input measurements of power distribution or control rod value obtained from an operating reactor to output the other control rod value or power distribution, which is then used as a predicted value for evaluation.

8. One of the aforementioned multiple core characteristics is the power distribution, control rod value, neutron flux distribution, reaction rate distribution, or aggregate burnup distribution. The core characteristics prediction calculation program according to claim 7, characterized in that, in the dataset creation step, training data created by combining one of the neutron flux distribution, reaction rate distribution, and aggregate burnup distribution, or two or more training data, is created together with the training data for the power distribution and control rod value to form a dataset.

9. The core characteristics prediction calculation program according to claim 8, characterized in that the neutron flux distribution is two or more neutron flux distributions classified based on the energy of neutrons.

10. The core characteristics prediction calculation program according to claim 8, characterized in that the neutron flux distribution is the neutron flux distribution of the energy group absorbed by the neutron-absorbing elements used in the control rods.

11. The core characteristics prediction calculation program according to any one of claims 7 to 10, characterized in that the machine learning is performed using deep learning with a multilayer neural network.

12. A reactor core characteristic prediction calculation device that predicts and calculates the other reactor core characteristic by measuring one of the two reactor core characteristics that cannot be measured simultaneously in a nuclear reactor, The system is equipped with a core characteristics prediction calculation program according to any one of claims 7 to 10, An input means for inputting a previously measured output distribution or control rod value measurement, A core characteristic prediction calculation device characterized by comprising an output means for outputting the results calculated by the core characteristic prediction calculation program as a predicted value for the other core characteristic.

Images