Thermal limit monitoring system, thermal limit monitoring device, and thermal limit monitoring method

Abstract

The objective is to provide a thermal limit monitoring system that is both high-speed and capable of improving the accuracy of thermal limit calculation. [Solution] A thermal limit monitoring system equipped with a thermal limit monitoring device, wherein the thermal limit monitoring device is a thermal limit calculation unit that calculates the thermal limit using machine learning with correction amounts used as training data in the calculation of the thermal limit for each fuel by a core performance calculation device.

Description

Technical Field

[0004] , , , , , ,

[0001] The present invention relates to a thermal limit value monitoring system, a thermal limit value monitoring device, and a thermal limit value monitoring method.

Background Art

[0002] As a technology related to a thermal limit value monitoring device provided in a boiling water reactor (BWR), there is the technology described in the following cited reference 1. In this cited reference 1, it is described that "as a conversion coefficient for converting the signal of the local power range monitor into an amount corresponding to the thermal limit value, it is a best fit value obtained from the actual thermal limit value and the signal of the local power range monitor during typical control rod operation or a value close thereto, and a conversion coefficient selected according to the in-core arrangement of the operating control rods from among a plurality of types of conversion coefficients is used, and the update of the conversion coefficient is performed by core performance calculation on a process computer with a shortened execution cycle, and when the amount corresponding to the thermal limit value approaches the limit value during operation, the core performance calculation by the process computer is started to update the conversion coefficient."

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] The thermal limit monitoring device described in Patent Document 1 above calculates the thermal limit from the signal of the local power range monitor using a conversion factor, thus enabling faster calculation of the thermal limit compared to the core performance calculation device, which takes several minutes to calculate the thermal limit. However, this thermal limit monitoring device cannot reflect the fuel / core design or the burnup of each fuel in the conversion factor, so the calculated thermal limit has a large error compared to the thermal limit calculated by the core performance calculation device during actual operation. Therefore, it is conceivable to adopt a calculation model using machine learning for calculating the thermal limit. However, even in this case, simply providing the fuel / core design as input data, calculating the conversion factor when the control rods and core flow rate are manipulated, and creating a calculation model using prior machine learning with these as training data will result in a difference between the calculated thermal limit and the thermal limit calculated by the core performance calculation device during actual operation.

[0005] Therefore, the present invention aims to provide a thermal limit monitoring system, a thermal limit monitoring device, and a thermal limit monitoring method that can improve the accuracy of thermal limit calculation while maintaining high speed. [Means for solving the problem]

[0006] To solve the above problems, for example, the configuration described in the claims may be adopted. The present invention includes several means for solving the above-mentioned problems, but one example is a thermal limit value monitoring system equipped with a thermal limit value monitoring device, wherein the thermal limit value monitoring device is a thermal limit value monitoring system equipped with a thermal limit value calculation unit that calculates the thermal limit value using machine learning with correction amounts used as training data in the calculation of the thermal limit value for each fuel by a core performance calculation device. [Effects of the Invention]

[0007] The present invention provides a thermal limit monitoring device, a thermal limit monitoring method, and a thermal limit monitoring system that can improve the accuracy of thermal limit calculation while maintaining high speed. [Brief explanation of the drawing]

[0008] [Figure 1] This is a block diagram showing the configuration of the thermal limit monitoring system according to the first embodiment. [Figure 2] This figure shows an example of a learning model generated by machine learning using the thermal limit monitoring system according to the first embodiment. [Figure 3] This diagram shows the configuration of the signal input / output control unit of the thermal monitoring system according to the first embodiment. [Figure 4] This is a diagram showing the input section of the thermal monitoring system according to the first embodiment. [Figure 5] This flowchart shows the pre-processing procedure in the thermal limit monitoring method according to the first embodiment. [Figure 6] This flowchart shows the procedure for the thermal limit monitoring method according to the first embodiment. [Figure 7] This is a block diagram showing the configuration of the thermal limit monitoring system according to the second embodiment. [Figure 8] This figure shows an example of a correction quantity learning model generated by machine learning using the thermal limit value monitoring system according to the second embodiment. [Figure 9] This figure shows an example of a learning model generated by machine learning using the thermal limit monitoring system according to the second embodiment. [Figure 10] This diagram shows the configuration of the signal input / output control unit of the thermal monitoring system according to the second embodiment. [Figure 11] This flowchart shows the pre-processing procedure in the thermal limit value monitoring method according to the second embodiment. [Figure 12] This is a block diagram showing the configuration of the thermal limit monitoring system according to the third embodiment. [Modes for carrying out the invention]

[0009] Hereinafter, embodiments to which the present invention is applied will be described in detail with reference to the drawings. In each embodiment, the same components are denoted by the same reference numerals, and redundant descriptions are omitted.

[0010] ≪First Embodiment≫ Figure 1 is a block diagram showing the configuration of a thermal limit monitoring system 1 according to a first embodiment. The thermal limit monitoring system 1 shown in this figure is a device for monitoring the thermal limits of the reactor core in a nuclear reactor. The thermal limits monitored by the thermal limit monitoring system 1 are two values: linear power density and limit power ratio. Such a thermal limit monitoring system 1 comprises a training data creation unit 10, a learning model generation unit 20, and a thermal limit monitoring device 30. Each of these units and devices is composed of one or separate computers. The computer may be a personal computer equipped with a CPU (Central Processing Unit), memory such as ROM (Read Only Memory) and RAM (Random Access Memory), and a network interface as needed. These components will be described below.

[0011] <Training Data Creation Unit 10> The training data creation unit 10 is an offline simulator that generates training data for the learning model generated by the learning model generation unit 20, which will be described below. The training data creation unit 10, which consists of an offline simulator, calculates thermal limits using the same calculation algorithm as the three-dimensional core performance calculation, which takes several minutes to calculate the thermal limits. Furthermore, this training data creation unit is designed to allow the same correction amounts as the core performance calculation to be set.

[0012] Here, the correction amount for core performance calculations (hereinafter simply referred to as the correction amount) is, as an example, the correction amount for cross-sectional data with respect to neutron flux. This correction amount is added to the core performance calculations so that the calculations reproduce the neutron flux values ​​measured by the Local Power Region Monitor (LPRM) and the Mobile Neutron Detector (TIP) during reactor operation.

[0013] The teacher data creation unit 10 has neutron flux distribution calculation data and thermal-hydraulic data based on the same fuel and core design as the core performance calculation, and performs simulations by giving a plurality of correction amounts (for example, -a, ±0, +a) for each fuel position. Thereby, the teacher data creation unit 10 sets the control rod operation amount and the core flow rate operation amount from the assumed initial control rod position and the initial core flow rate, and calculates the thermal limit value and the LPRM value after the operation. Through this process, the teacher data creation unit 10 creates teacher data with (1) the control rod operation amount, (2) the initial control rod position, (3) the core flow rate operation amount, (4) the initial core flow rate, (5) the LPRM ratio, and (6) the correction amount as input values and the change rate of the thermal limit value as the output value.

[0014] Note that in the teacher data creation unit 10, when creating the teacher data, the correction amounts given to each fuel position are set to the same values for fuels whose cross-sectional positions in the core are at symmetric positions (mirror symmetry or rotational symmetry). Also, the teacher data creation unit 10 may be configured to give only the correction amounts of the fuels arranged adjacent to each LPRM detector at substantially the same height, and give the correction amounts for other height positions by interpolation and extrapolation. As described above, the number of teacher data required for learning in the teacher data creation unit 10 can be reduced, and the learning time in the subsequent stage can be shortened.

[0015] <Learning model generation unit 20> The learning model generation unit 20 generates a learning model for a fraction of the fuels by machine learning using the teacher data created by the teacher data creation unit 10. FIG. 2 is a diagram showing an example of a learning model generated by machine learning by the thermal limit value monitoring system 1 according to the first embodiment. The learning model shown in FIG. 2 is a learning model of fuels generated by the learning model generation unit 20 by machine learning using a neural network.

[0016] In this learning model, the input layer 21 receives the input values ​​(1) to (6) described in the training data creation unit 10 (see Figure 1) for each fuel, and the rate of change of the thermal limit value is output to the output layer 23 via the intermediate layer 22. Note that for (1) control rod manipulation amount and (2) initial control rod position, the values ​​of the four adjacent control rods are input for each fuel. Also, for (5) LPRM ratio, the values ​​of the 16 adjacent LPRMs are input for each fuel.

[0017] The number of layers in the intermediate layer 22 and the number of nodes 221 in each layer are set considering the calculation time and accuracy of the thermal limit value. Here, the diagram shows the case where the intermediate layer 22 has 3 layers, but it is not limited to this. The output layer 23 outputs the rate of change of the thermal limit value, which includes the rate of change of the maximum output power density for each fuel and the rate of change of the limit power ratio. The learning model generation unit 20 optimizes the weights and biases, which are model coefficients, in such a learning model.

[0018] <Thermal limit monitoring device 30> Returning to Figure 1, the thermal limit monitoring device 30 calculates and outputs the thermal limit value through inference using the learning model generated by the learning model generation unit 20. This thermal limit monitoring device 30 includes a signal input / output control unit 31, an input unit 32, a thermal limit value calculation unit 33, a thermal limit value output unit 34, and a determination unit 35. These are as follows.

[0019] [Signal Input / Output Control Unit 31] The signal input / output control unit 31 processes input signals from various devices in the reactor facility used to control the reactor output and outputs them to the thermal limit value calculation unit 33. The devices that output input signals to the signal input / output control unit 31 are the control rod drive control device 2, the core flow monitoring device 3, the neutron flux monitoring device 4, and the core performance calculation device 5. The signal input / output control unit 31 also outputs the processed signals to the inference unit 332 of the thermal limit value calculation unit 33 as input values ​​for the learning model generated by the learning model generation unit 20. Note that this thermal limit value monitoring system 1 may also include the control rod drive control device 2, the core flow monitoring device 3, the neutron flux monitoring device 4, and the core performance calculation device 5.

[0020] Figure 3 shows the configuration of the signal input / output control unit 31 of the thermal monitoring system according to the first embodiment. As shown in Figure 3, the input signals input to the signal input / output control unit 31 from each device are as follows: The input signal from the control rod drive control device 2 is (a) the control rod position of each control rod. The input signal from the core flow monitoring device 3 is (b) the core flow rate. The input signal from the neutron flux monitoring device 4 is (c) the LPRM value detected by the neutron detectors (LPRM) placed in each part of the core. The input signals from the core performance calculation device 5 are (d) the calculation start signal from the core performance calculation device 5, (e) the thermal limit values ​​(linear power density and limiting power ratio) calculated by the core performance calculation device 5, and (f) the correction amount. Of these, (f) the correction amount is the correction amount in the core performance calculation performed by the core performance calculation device 5, for example, the correction amount for the cross-sectional area data with respect to the neutron flux.

[0021] The signal input / output control unit 31, which processes these input signals, has latch circuits 301a to 301c, 302a to 302c, arithmetic circuits 303a to 303c, a time delay circuit 304, and a comparison circuit 305, and operates as follows.

[0022] First, when the core performance calculation device 5 outputs a (d) calculation start signal to the signal input / output control unit 31, the latch circuits 301a to 301c hold the (a) control rod position, (b) core flow rate, and (c) LPRM value input at that time as their respective initial values. The latch circuits 301a to 301c also continue to output the held initial values ​​to the subsequent latch circuits 302a to 302c.

[0023] Subsequently, once the core performance calculation in the core performance calculation device 5 is completed, the core performance calculation device 5 outputs (e) thermal limit values ​​and (f) correction amounts. As a result, the signal input / output control unit 31 outputs (e) thermal limit values ​​as the new initial thermal limit values ​​and (f) correction amounts as the new (6) correction amounts to the thermal limit value calculation unit 33 (see Figure 1).

[0024] At the same time, the (e) thermal limit value output from the core performance calculation device 5 is compared with the thermal limit value input to the time delay circuit 304 in the previous calculation cycle by the comparison circuit 305. As a result, the comparison circuit 305 detects the difference between the two and outputs a calculation completion signal to the latch circuits 302a to 303c. Upon receiving the calculation completion signal from the comparison circuit 305, the latch circuits 302a to 303c stop holding and outputting the (2) initial control rod position, (4) initial core flow rate, and initial LPRM value (LPRM_0) from the previous calculation cycle, which they had been outputting until then. Furthermore, the latch circuits 302a to 302c simultaneously output the (2) initial control rod position, (4) initial core flow rate, and initial LPRM value (LPRM_0) that the preceding latch circuits 301a to 301c have been continuously outputting.

[0025] As described above, the signal input / output control unit 31 updates each initial value and correction value simultaneously each time the core performance calculation device 5 completes its core performance calculation, and outputs the updated (2) initial control rod position, (4) initial core flow rate, (6) correction amount, and initial thermal limit value to the thermal limit value calculation unit 33 (see Figure 1).

[0026] At the same time, the calculation circuit 303a calculates the difference between (a) the control rod position output from the control rod drive control device 2 and (2) the initial control rod position output from the latch circuit 302a, and outputs (1) the control rod operation amount to the thermal limit value calculation unit 33 (see Figure 1). Similarly, the calculation circuit 303b calculates the difference between (b) the core flow rate output from the core flow rate monitoring device 3 and (4) the initial core flow rate output from the latch circuit 302b, and outputs (3) the core flow rate operation amount to the thermal limit value calculation unit 33 (see Figure 1).

[0027] Furthermore, the calculation circuit 303c calculates the ratio of the (c)LPRM value (LPRM_P) output from the neutron flux monitoring device 4 to the initial LPRM value (LPRM_0) output from the latch circuit 302c, and outputs this as the (5)LPRM ratio to the thermal limit value calculation unit 33 (see Figure 1).

[0028] [Input section 32] Returning to Figure 1, the input unit 32 is the part that inputs the model coefficients of the learning model generated by the learning model generation unit 20 to the thermal limit value calculation unit 33. This input unit 32 is an input device such as a touch panel or input display unit connected to the computer, and is connected to both the learning model generation unit 20 and the thermal limit value calculation unit 33.

[0029] Figure 4 shows the input unit 32 of the thermal monitoring system according to the first embodiment, and shows the model coefficient input screen of the input unit 32. This input screen is configured as a touch panel and allows input of model coefficients of the learning model (see Figure 2) generated by the learning model generation unit 20. Such an input unit 32 has a model coefficient input unit 321, and can be used to input data from a file, for example, in a specified CSV format. The model coefficient input unit 321 is an interface for inputting model coefficients from a file, and the model coefficients can be input by selecting a file stored on an external storage device such as a USB memory from a pull-down menu and tapping the input execution button 322.

[0030] The input unit 32 also includes a node selection unit 323, a weight coefficient display unit 324, and a bias display unit 325, allowing for confirmation and modification of the entered model coefficients. In the node selection unit 323, a node number can be selected from a pull-down menu, and as shown in the figure, all nodes [All] can also be selected. The weight coefficient display unit 324 and the bias display unit 325 are configured to display the weight coefficient and bias value for each input signal to each node. To change the weight coefficient and bias value, tap the corresponding display field in the weight coefficient display unit 324 or bias display unit 325, and then enter the value using the numeric keypad buttons 327. To save the entered value, tap the input execution button 326 below the weight coefficient display unit 324 or bias display unit 325.

[0031] The input unit 32 may also display an input screen for instructing the start and end of automatic thermal limit monitoring, which is not shown in the illustration here.

[0032] [Thermal limit value calculation unit 33] Returning to Figure 1, the thermal limit calculation unit 33 calculates the thermal limit using a learning model that takes the output signal from the signal input / output control unit 31 as its input value. This thermal limit calculation unit 33 includes a learning model storage unit 331 and an inference unit 332.

[0033] Of these, the learning model storage unit 331 stores the learning model (see Figure 2) generated by the learning model generation unit 20. This learning model is input from the input unit 32.

[0034] The inference unit 332 also receives output signals from the signal input / output control unit 31. As explained using Figure 3, the output signals from the signal input / output control unit 31 are (1) control rod manipulation amount, (2) initial control rod position, (3) core flow manipulation amount, (4) initial core flow rate, (5) LPRM ratio, (6) correction amount, and initial thermal limit value. The inference unit 332 inputs (1) to (6) of these signals as input values ​​to the learning model stored in the learning model storage unit 331, and obtains the output value of the learning model as the rate of change of the thermal limit value (see Figure 2). The inference unit 332 multiplies the obtained rate of change of the thermal limit value by the initial thermal limit value obtained as an output signal from the signal input / output control unit 31 and outputs the current thermal limit value to the thermal limit value output unit 34.

[0035] Furthermore, the thermal limit calculation unit 33 processes the (6) correction amount input from the core performance calculation device 5 via the signal input / output control unit 31 so that it is the same for each fuel arranged symmetrically within the core, and uses it as input value for machine learning. Alternatively, the correction amount for fuels arranged adjacent to each LPRM detector within the core at approximately the same height may be used as input value for machine learning, while the correction amounts for other height positions may be provided by interpolation and extrapolation.

[0036] [Thermal limit value output unit 34] The thermal limit value output unit 34 outputs the thermal limit value calculated by the thermal limit value calculation unit 33. This thermal limit value output unit 34 may have, for example, a display device that outputs the thermal limit value calculated by the thermal limit value calculation unit 33 by display.

[0037] [Judgment section 35] The determination unit 35 determines the start and end of monitoring of the thermal limit value in the thermal limit value monitoring device 30.

[0038] <Method for monitoring thermal limits> Next, the thermal limit monitoring method using the thermal limit monitoring system 1 described above will be explained based on the flowcharts in Figures 5 and 6. The steps of the thermal limit monitoring method shown in these flowcharts are the steps of automatic thermal limit monitoring performed by the programs of each part of the thermal limit monitoring system 1 described above. Figure 5 is a flowchart showing the pre-processing steps in the thermal limit monitoring method according to the first embodiment. First, the pre-processing performed in the thermal limit monitoring method will be explained in the order shown in the flowchart of Figure 5, with reference to Figure 1 and other necessary diagrams.

[0039] [Step S11] In step S11, the training data creation unit 10 generates training data for the learning model to be generated by the learning model generation unit 20. At this time, as explained earlier, the training data creation unit 10 creates training data including correction amounts by performing a three-dimensional simulation of the reactor core performance calculation with the same correction amounts as those used in the reactor core performance calculation.

[0040] [Step S12] In step S12, the learning model generation unit 20 generates a learning model (see Figure 2) for all fuels (fuel assemblies) using correction amounts as input values, by learning with the training data created in step S11.

[0041] [Step S13] In step S13, the operator inputs the model coefficients of the learning model generated in step S12 from the input unit 32 to the thermal limit monitoring device 30. If the learning model generation unit 20 and the thermal limit monitoring device 30 are connected via a network or are configured on the same computer, step S13 may be a procedure in which the thermal limit monitoring device 30 directly acquires the learning model from the learning model generation unit 20.

[0042] After performing the preprocessing as described above, the procedure shown in Figure 6 is carried out. Figure 6 is a flowchart showing the procedure for the thermal limit value monitoring method according to the first embodiment. Next, the procedure for the thermal limit value monitoring method will be explained in the order shown in the flowchart of Figure 6, with reference to Figures 1 and 3.

[0043] [Step S101] In step S101, the determination unit 35 determines whether or not monitoring of the thermal limit value has started. In this case, the determination unit 35 determines that monitoring has started (YES) if, for example, the operator instructs the start of monitoring of the thermal limit value by operating the input unit 32, or when a predetermined monitoring start cycle has been reached. If it determines that monitoring has started (YES), the unit proceeds to the next step S102.

[0044] [Step S102] In step S102, the signal input / output control unit 31 acquires input signals for calculating thermal limits from each device installed in the reactor equipment. The input signals acquired here, as explained with reference to Figure 3, are (a) control rod position, (b) core flow rate, (c) LPRM value, (d) calculation start signal, (e) thermal limit value, and (f) correction amount. Of these, (d) the calculation start signal is a signal transmitted by the core performance calculation device 5 when it starts core performance calculations at a specified interval (e.g., every hour) or on demand. Also, (e) the thermal limit value and (f) the correction amount are signals transmitted by the core performance calculation device 5 when it starts and when it finishes core performance calculations. These signals are input to the signal input / output control unit 31 after waiting for transmission from the core performance calculation device 5. The signal input / output control unit 31 also outputs the acquired input signals to the thermal limit value calculation unit 33.

[0045] [Step S103] In step S103, the signal input / output control unit 31 proceeds to step S104 if (d) a calculation start signal is present (YES) among the input signals acquired in step S102. On the other hand, if (d) a calculation start signal is not present (NO), the unit proceeds to step S105.

[0046] [Step S104] In step S104, the signal input / output control unit 31 holds the (a) control rod position, (b) core flow rate, and (c) LPRM value, which were acquired simultaneously with the (d) calculation start signal in step S102, as initial values. Here, as explained earlier using Figure 3, the latch circuits 301a to 301c of the signal input / output control unit 31 hold these initial values. As explained earlier, the latch circuits 301a to 301c continue to output the held initial values ​​to the subsequent latch circuits 302a to 302c.

[0047] [Step S105] In step S105, the signal input / output control unit 31 determines whether the core performance calculation has been completed. As explained earlier using Figure 3, the signal input / output control unit 31 determines that the core performance calculation has been completed (YES) if a calculation completion signal is output from the comparison circuit 305 of the thermal limit value calculation unit 33 due to (e) the input of thermal limit values ​​from the core performance calculation device 5, and proceeds to step S106. Otherwise, it determines that the core performance calculation has not been completed (NO) and proceeds to step S107.

[0048] [Step S106] In step S106, the signal input / output control unit 31 outputs the initial values ​​held in step S104 to the thermal limit value calculation unit 33, updating the initial values. Here, as explained earlier using Figure 3, when the comparison circuit 305 of the thermal limit value calculation unit 33 outputs a calculation completion signal, the subsequent latch circuits 302a to 302c output the initial values ​​that the latch circuits 301a to 301c had been continuously outputting. As a result, the initial values ​​for calculating the thermal limit value are updated.

[0049] [Step S107] In step S107, the inference unit 332 of the thermal limit value calculation unit 33 inputs the input signal from the signal input / output control unit 31 as an input value to the learning model stored in the learning model storage unit 331, and calculates the output value of the learning model as the rate of change of the thermal limit value (see Figure 2). Furthermore, the inference unit 332 multiplies the rate of change of the thermal limit value by the initial thermal limit value obtained as the output signal from the signal input / output control unit 31 to calculate the current thermal limit value.

[0050] [Step S108] In step S108, the thermal limit value output unit 34 outputs the thermal limit value calculated by the thermal limit value calculation unit 33.

[0051] [Step S109] In step S109, the determination unit 35 determines whether or not to terminate the monitoring of the thermal limit value. In this case, the determination unit 35 determines to terminate monitoring (YES) when, for example, an operator instructs the termination of thermal limit value monitoring by operating the input unit 32, or when a predetermined monitoring start cycle has ended (YES). If it determines to terminate monitoring (YES), it terminates the automatic monitoring of the thermal limit value. If it determines not to terminate monitoring (NO), it returns to step S102 and repeats the subsequent steps to continue monitoring the automatic thermal limit value.

[0052] <Effects of the First Embodiment> According to the first embodiment described above, by using the correction value from the core performance calculation by the core performance calculation device as an input value for calculating the thermal limit value by machine learning, it is possible to calculate a thermal limit value with an extremely small error compared to the thermal limit value calculated by the core performance calculation device at high speed.

[0053] Furthermore, in an automated power control system for a nuclear reactor that automates the operation of control rods and core flow rates, sufficient maintainability must be considered for the thermal limit values ​​calculated quickly by the thermal limit monitoring device, so that the precise thermal limit values ​​calculated by the core performance calculator, which takes several minutes to calculate the thermal limit values, do not reach the limit values. However, by applying this first embodiment, it has become possible to calculate thermal limit values ​​with extremely small errors compared to the thermal limit values ​​calculated by the core performance calculator at high speed, thereby eliminating the need for excessive maintainability for the thermal limit values ​​calculated by the thermal limit monitoring device. As a result, it is possible to prevent unnecessary shutdowns of the automated control system and the resulting increase in the burden on operators and extension of operating time.

[0054] ≪Second Embodiment≫ Figure 7 is a block diagram showing the configuration of the thermal limit monitoring system according to the second embodiment. The thermal limit monitoring system 1a shown in this figure differs from the thermal limit monitoring system 1 of the first embodiment described using Figure 1 in that it has a correction amount training data creation unit 40 and a correction amount learning model generation unit 50. As a result, the configuration of the training data created by the training data creation unit 10a, the learning model generated by the learning model generation unit 20a, and the signal input / output control unit 31a of the thermal limit monitoring device 30 are different. Other configurations are the same as in the first embodiment. For this reason, in the following, the explanation of configurations that overlap with the first embodiment will be omitted, and the configurations of the correction amount training data creation unit 40, the correction amount learning model generation unit 50, the training data creation unit 10a, the learning model generation unit 20a, and the signal input / output control unit 31a of the thermal limit monitoring device 30 will be described.

[0055] <Correction Amount Training Data Creation Unit 40> The correction amount training data creation unit 40 creates training data using past performance data, etc., with control rod position and core flow rate as input values ​​and correction amounts for core performance calculation as output values. As explained in the first embodiment, the correction amount for core performance calculation is, for example, the correction amount for cross-sectional area data with respect to neutron flux.

[0056] <Correction Amount Learning Model Generation Unit 50> The correction amount learning model generation unit 50 generates a correction amount learning model for each fuel assembly by machine learning using the training data created by the correction amount training data creation unit 40. Figure 8 shows an example of a correction amount learning model generated by machine learning using the thermal limit value monitoring system according to the second embodiment. The correction amount learning model shown in Figure 8 is a learning model generated by machine learning using a neural network. In this correction amount learning model, the input values ​​(control rod positions and core flow rates) shown in the correction amount training data creation unit 40 (see Figure 7) are input from the input layer 51, and the correction values ​​for core performance calculation are output to the output layer 53 via the intermediate layer 52. The input values ​​input here are the control rod positions of the four control rods adjacent to the fuel to which the correction amount is to be applied, and the core flow rate.

[0057] The number of layers in the intermediate layer 52 and the number of nodes 521 in each layer are set considering the calculation time and calculation accuracy of the correction values ​​for the core performance calculation. Here, the case where the intermediate layer 52 has 3 layers is illustrated, but it is not limited to this. The correction amount learning model generation unit 50 optimizes the weights and vials, which are model coefficients, in such a learning model.

[0058] <Training data creation unit 10a> Returning to Figure 7, the training data creation unit 10a is an offline simulator, and its function of creating training data for the learning model generated by the learning model generation unit 20a, which will be described next, is the same as that of the training data creation unit 10 in the first embodiment (see Figure 1). Furthermore, this training data creation unit 10a is the same as the offline simulator of the training data creation unit 10 described in the first embodiment. However, the correction amount for the cross-sectional area data of each fuel provided to this training data creation unit 10a is the correction amount obtained by the correction amount learning model generation unit 50. This reduces the number of training data generated by the training data creation unit 10a and shortens the subsequent learning time.

[0059] As a result, the training data creation unit 10a creates training data with (1) control rod manipulation amount, (2) initial control rod position, (3) core flow rate manipulation amount, (4) initial core flow rate, and (5) LPRM ratio as input values, and the rate of change of the thermal limit value as the output value.

[0060] <Learning Model Generation Unit 20a> The learning model generation unit 20a generates a learning model for each fuel assembly of several minutes of fuel using machine learning with the training data created by the training data creation unit 10a, similar to the learning model generation unit 20 of the first embodiment (see Figure 1). Figure 9 shows an example of a learning model generated by machine learning using the thermal limit value monitoring system 1a according to the second embodiment. As shown in Figure 9, this learning model generation unit 20a generates a learning model using machine learning with a neural network, similar to the learning model generation unit 20 described in the first embodiment. However, this learning model generation unit 20a uses (1) to (5) shown in the training data creation unit 10a as input values, and does not input the correction amount for core performance calculation. Instead, the rate of change of the maximum output power density for each fuel and the rate of change of the limit power ratio are output as the rate of change of the thermal limit value.

[0061] <Thermal limit monitoring device 30> Returning to Figure 7, the thermal limit monitoring device 30 calculates and outputs the thermal limit value by inference using the learning model generated by the learning model generation unit 20. The only difference between this thermal limit monitoring device 30 and the first implementation value is the configuration of the signal input / output control unit 31a; the other configurations are the same, so only the signal input / output control unit 31a will be described here.

[0062] [Signal Input / Output Control Unit 31a] Figure 10 shows the configuration of the signal input / output control unit 31a of the thermal monitoring system according to the second embodiment. The difference between the signal input / output control unit 31a shown in this figure and the signal input / output control unit 31a of the first embodiment is that it does not acquire correction amounts from the core performance calculation device 5, but only acquires (d) the calculation start signal and (e) the thermal limit value; the other configurations are the same.

[0063] <Method for monitoring thermal limits> Next, the thermal limit monitoring method using the thermal limit monitoring system 1a described above will be explained based on Figure 11 and the flowchart in Figure 6 used in the description of the first embodiment above. The steps of the thermal limit monitoring method shown in these flowcharts are the steps of automatic thermal limit monitoring performed by the programs of each part of the thermal limit monitoring system 1a described above. Figure 11 is a flowchart showing the pre-processing steps in the thermal limit monitoring method according to the second embodiment. First, the pre-processing performed in the thermal limit monitoring method will be explained in the order shown in the flowchart of Figure 11, with reference to Figure 7 and other necessary figures.

[0064] [Step S21] In step S21, the correction amount training data creation unit 40 generates training data for the correction amount learning model to be generated by the correction amount learning model generation unit 50. At this time, as explained earlier, the correction amount training data creation unit 40 creates training data with the correction amount of the core performance calculation as the output value.

[0065] [Step S22] In step S22, the correction amount learning model generation unit 50 generates a correction amount learning model (see Figure 8) for all fuels (fuel assemblies) using the correction amount for core performance calculations as the output value, by learning with the training data created in step S21.

[0066] [Step S23] In step S23, the training data creation unit 10a generates training data for the learning model to be generated by the learning model generation unit 20a. At this time, the training data creation unit 10a creates the training data using the correction amount obtained by the correction amount learning model generation unit 50.

[0067] [Step S24] In step S24, the learning model generation unit 20a generates a learning model (see Figure 9) for all fuels (fuel assemblies) by learning using the training data created in step S23.

[0068] [Step S25] In step S25, the operator inputs the model coefficients of the learning model generated in step S24 from the input unit 32 to the thermal limit monitoring device 30. If the learning model generation unit 20a and the thermal limit monitoring device 30 are connected via a network or are configured on the same computer, step S25 may be a procedure in which the thermal limit monitoring device 30 directly acquires the learning model from the learning model generation unit 20a.

[0069] After performing the preprocessing as described above, the procedure for the thermal limit value monitoring method according to the second embodiment is carried out. The procedure for the thermal limit value monitoring method carried out here is the same as the procedure described using the flowchart in Figure 6 in the first embodiment. However, in step S102, the signal input / output control unit 31a acquires other signals without acquiring the correction amount. As a result, in step S107, the inference unit 332 of the thermal limit value calculation unit 33 calculates the output value of the learning model as the rate of change of the thermal limit value (see Figure 9) without using the correction amount as an input value.

[0070] <Effects of the second embodiment> In the second embodiment described above, correction values ​​from core performance calculations are used as input values ​​to create training data for calculating thermal limit values ​​using machine learning. As a result, similar to the first embodiment, it is possible to calculate thermal limit values ​​with extremely small errors compared to the thermal limit values ​​calculated by the core performance calculation device at high speed. Furthermore, as a result, similar to the first embodiment, excessive maintainability of operating limits can be eliminated in the automatic power control system of a reactor with automated control rod and core flow rate operations, and unnecessary shutdowns of automatic control can be reduced.

[0071] Furthermore, according to this second embodiment, the correction quantity learning model allows for the reduction of the amount of training data to be created and the amount of training data to be learned by providing a correction quantity in advance for the creation of training data for calculating thermal limit values ​​using machine learning.

[0072] ≪Third Embodiment≫ Figure 12 is a block diagram showing the configuration of the thermal limit value monitoring system 1b according to the third embodiment. The thermal limit value monitoring system 1b of the third embodiment shown in Figure 12 is characterized by the addition of a limit signal output unit 34b in place of the thermal limit value output unit 34 in the thermal limit value monitoring device 30 of the thermal limit value monitoring system 1 of the first embodiment described using Figure 1. Since the other configurations are the same, the configuration of the limit signal output unit 34b will be described here.

[0073] <Thermal limit monitoring device 30> [Limiting signal output section 34b] The limit signal output unit 34b compares the thermal limit value of each fuel calculated by the inference unit 332 of the thermal limit value calculation unit 33 with the pre-set thermal limit value of each fuel that was stored. If the limit signal output unit 34b determines, as a result of the comparison, that the thermal limit value calculated by the inference unit 332 exceeds the set value, it outputs a stop signal for automatic power adjustment to the reactor's automatic power adjustment device 6. In this case, the limit signal output unit 34b also outputs a stop signal for control rod operation to the control rod drive control device 2.

[0074] <Effects of the Third Embodiment> According to the third embodiment described above, as explained in the first embodiment, a stop signal is transmitted to the control rod drive control device 2 and the automatic power adjustment device 6 based on a thermal limit value with an extremely small error compared to the thermal limit value calculated by the core performance calculation device. This makes it possible to realize an automatic power control system for a nuclear reactor that monitors the thermal limit value with the same accuracy as the core performance calculation device. As a result, unnecessary shutdowns of the automatic control can be reduced, and the use of the automatic power control system can be expanded.

[0075] Furthermore, this third embodiment can be combined with the second embodiment, thereby obtaining the effects of the second embodiment. In addition, this thermal limit monitoring system 1b may include a control rod drive control device 2, a core flow monitoring device 3, a neutron flux monitoring device 4, a core performance calculation device 5, and an automatic power adjustment device 6.

[0076] It should be noted that the present invention is not limited to the embodiments and modifications described above, and includes a variety of further modifications. For example, the embodiments described above are described in detail for the purpose of clearly illustrating the present invention, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace parts of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add configurations from other embodiments to the configuration of one embodiment. In addition, it is possible to add, delete, or replace parts of the configuration of each embodiment with other configurations. [Explanation of Symbols]

[0077] 1,1a,1b…Thermal limit monitoring system 2…Control rod drive control device 3…Core flow monitoring device 4...Neutron flux monitoring device 5…Core performance calculator 6…Automatic output adjustment device 10,10a...Teacher Data Creation Department 20,20a...Learning model generation unit 30…Thermal limit monitoring device 31,31a...Signal Input / Output Control Unit 32...Input section 33... Thermal limit calculation unit 34b... Output section of the limiting signal 40…Correction Amount Training Data Creation Department 50... Correction Amount Learning Model Generation Unit

Claims

1. A thermal limit monitoring system equipped with a thermal limit monitoring device, The aforementioned thermal limit value monitoring device is The system includes a thermal limit calculation unit that calculates thermal limit values ​​using machine learning, with the correction amounts used in the calculation of thermal limit values ​​for each fuel type by the core performance calculation device as training data. Thermal limit monitoring system.

2. The aforementioned thermal limit value monitoring device The system includes a signal input / output control unit that outputs signals acquired from each device for controlling the output of the reactor, including the core performance calculation device, as input values ​​for the machine learning in the thermal limit value calculation unit. The thermal limit monitoring system according to claim 1.

3. The thermal limit value calculation unit stores the machine learning model and uses the input values ​​output from the signal input / output control unit to infer the thermal limit value from the learning model. The thermal limit monitoring system according to claim 2.

4. The signal input / output control unit updates the initial value among the input values ​​output to the thermal limit value calculation unit each time the core performance calculation device performs a core performance calculation. The thermal limit monitoring system according to claim 2.

5. The correction amount, which serves as training data for the aforementioned machine learning, is the same for each fuel arranged symmetrically within the reactor core. The thermal limit calculation unit processes the correction amount input from the core performance calculation device via the signal input / output control unit so that it is the same for each fuel arranged symmetrically within the core, and uses it as the input value for machine learning. The thermal limit monitoring system according to claim 2.

6. The thermal limit calculation unit uses only the correction amount corresponding to fuel located adjacent to the local power region monitor of the neutron flux located within the reactor core at approximately the same height as the correction amount input from the reactor core performance calculation device via the signal input / output control unit as input to the machine learning function. The thermal limit monitoring system according to claim 2.

7. The system includes a learning model generation unit that generates a learning model using the correction amounts used as training data in the calculation of thermal limit values ​​for each fuel by the aforementioned core performance calculation device, The thermal limit value calculation unit generates the input value output from the signal input / output control unit in the learning model generation unit, has the learning model infer the value, and outputs the thermal limit value. The thermal limit monitoring system according to claim 2.

8. The system includes a training data creation unit that calculates thermal limit values ​​by performing offline simulations with correction amounts set using the same algorithm as the aforementioned core performance calculator, and then creates the training data. The thermal limit monitoring system according to claim 7.

9. A correction amount learning model generation unit generates a correction amount learning model that takes the control rod position and core flow rate as input values ​​and the correction amount as the output value, The system includes a learning model generation unit that generates a learning model based on training data created using the correction quantities obtained through the training of the correction quantity learning model. The thermal limit monitoring system according to claim 1.

10. The aforementioned thermal limit value monitoring device The thermal limit calculation unit is equipped with an input unit for inputting the machine learning model. The thermal limit monitoring system according to claim 1.

11. The system has a limit signal output unit that outputs a stop signal for automatic power adjustment to the reactor's automatic power adjustment device when the thermal limit value calculated by the thermal limit value calculation unit exceeds a set value. The thermal limit monitoring system according to claim 1.

12. The system has a limit signal output unit that outputs a signal to stop the operation of the control rods to the reactor's control rod drive control device when the thermal limit value calculated by the thermal limit value calculation unit exceeds a set value. The thermal limit monitoring system according to claim 1.

13. The system includes a thermal limit calculation unit that calculates thermal limit values ​​using machine learning, with the correction amounts used in the calculation of thermal limit values ​​for each fuel type by the core performance calculation device as training data. Thermal limit monitoring device.

14. The thermal limit calculation unit calculates the thermal limit using machine learning, with the correction values ​​used in the calculation of the thermal limit for each fuel by the core performance calculation device as training data. A method for monitoring thermal limits.

Images