JP2022131956A - Prediction model building device and prediction device - Google Patents

Abstract

To provide a prediction model building device capable of precisely predicting corrosion potential of reactor structural materials of a nuclear power plant.SOLUTION: The prediction model building device (corrosion potential prediction device 100) is a device for building a prediction model for predicting the corrosion potential of structural materials for a nuclear reactor in a nuclear power plant. The prediction model building device includes a learning unit 111 that builds a prediction model 130 by training a machine learning model using a teacher's data that includes a plant state quantity that is measurable including at least one of reactor water quality, core power output, and radiation monitoring data as input data and teacher data that includes the measured corrosion potential as the correct data.SELECTED DRAWING: Figure 3

Description

TECHNICAL FIELD The present invention relates to a predictive model construction device and a predictive device for corrosion potential related to structural materials of nuclear power plants.

As nuclear power plants (also simply referred to as plants), for example, boiling water nuclear power plants (BWR: Boiling Water Reactor) and pressurized water reactors (PWR: Pressurized Water Reactor) are known. In these plants, stainless steel, nickel-based alloys, and the like are used for the water-contacting parts of the main components such as the reactor pressure vessel in order to suppress corrosion. These materials exhibit susceptibility to stress corrosion cracking (hereinafter also referred to as SCC (Stress Corrosion Cracking)) under certain conditions. Therefore, stress corrosion cracking prevention measures are applied to maintain reactor integrity. Moreover, in recent years, measures to prevent stress corrosion cracking have been applied from the viewpoint of improving the capacity utilization rate of nuclear reactors and improving economic efficiency such as extending the life of nuclear reactors.

As a countermeasure against stress corrosion cracking, we use stainless steel and nickel-based alloys designed with a composition that is highly resistant to stress corrosion cracking, and various stress relaxation techniques are applied to alleviate the residual tensile stress generated in the weld. . For example, there are surface polishing, peening, or high-frequency residual stress improvement methods. In addition to the above materials and stress corrosion cracking countermeasures, harmful ions that promote stress corrosion cracking (chloride ions, sulfate ions, chromate ions, etc.) present in the reactor water are removed from the reactor water. Purified water is managed by actively removing it using a purification system.

Furthermore, the corrosion potential (ECP, Electrochemical Corrosion Potential), which is an index of stress corrosion cracking, is being reduced. It is known that the occurrence of stress corrosion cracking is suppressed when the corrosion potential is lower than -300 to -200 mV vs. SHE. If oxidizing chemical species such as oxygen and hydrogen peroxide generated by radiolysis of water during operation of a nuclear power plant are present in the reactor water, the ECP will increase. Simultaneously with the decrease in corrosion resistance of the material and the presence of residual tensile stress, stress corrosion cracking occurs and progresses due to the increase in corrosion potential.

Therefore, as a countermeasure against stress corrosion cracking, hydrogen injection is applied in which hydrogen is added to the reactor water to return the oxidizing chemical species to water through a recombination reaction to lower the corrosion potential. Hydrogen injection is applied to all PWR plants and many BWR plants.

In recent years, in BWRs, a technology called noble metal injection has been developed, in which platinum group noble metals are added to the reactor water and adhered to the surface of stainless steel or nickel-based alloys to provide electrochemical catalytic properties related to the reaction between oxygen and hydrogen possessed by the platinum group noble metals. It is mainly applied to BWRs in the United States. By applying noble metal injection, the amount of hydrogen required to reduce the corrosion potential below the target value can be greatly reduced, and the range of corrosion potential reduction can be expanded compared to the case where it is not applied.

These conventional techniques require accurate knowledge of the corrosion potential of structural materials in order to evaluate the effects of stress corrosion cracking. Therefore, in conventional plants, corrosion potential sensors are installed in the reactor pressure vessel or in pipes connected to the reactor pressure vessel to measure the corrosion potential of structural materials.
It is known to predict the life of nuclear reactor primary system structures by predicting stress corrosion cracking and crack propagation due to corrosion fatigue (see Patent Document 1). This life prediction is performed using the measured corrosion potential and crack growth property data of the structural material. Another method for estimating the life expectancy of structural members of a nuclear power plant is known (see Patent Document 2).

The corrosion potential of the structural material is measured by the installed sensor, but the corrosion potential other than the sensor installation position can be obtained by calculation. A method of calculating the corrosion potential will be described below.
Factors affecting corrosion potential include dissolved oxygen concentration, hydrogen peroxide concentration, and dissolved hydrogen concentration, as well as flow velocity and hydraulic equivalent diameter, which contribute to the mass transfer rate of these species to the structural material surface. The flow velocity and hydraulic equivalent diameter are determined by the core flow rate and the flow area and wetted edge length at each site. Dissolved oxygen, dissolved hydrogen, and hydrogen peroxide are produced by radiolytic reactions in water, so their concentrations are affected by the dose rate of gamma rays and neutrons emitted from the reactor core. The dose rate and dose rate distribution in these pressure vessels are affected by the operating conditions of the core, ie, fuel power, axial and radial power distribution.

When hydrogen injection is applied, the concentration of dissolved oxygen, hydrogen peroxide, and dissolved hydrogen in the furnace decreases due to the increase in feedwater hydrogen concentration, and the distribution changes, so the corrosion potential distribution also changes. Furthermore, when noble metal implantation is applied, the corrosion potential is affected by the deposition amount of noble metal, which contributes to the electrochemical reaction on the surface of the structural material, and the excessive amount of hydrogen with respect to the stoichiometric ratio with respect to the amount of oxygen. important factor to give

Precious metal injection involves injecting precious metals (primarily platinum) into the reactor water during shutdown or power operation. At this time, precious metals adhere to the surfaces of structural materials and fuel rods that are in contact with the reactor water. When the injection of the precious metal is stopped, part of the precious metal adhering to the surfaces of the structural material and the fuel rods dissolves or peels off from the respective surfaces into the reactor water. Precious metals that migrate to the reactor water adhere again to the surfaces of structural materials and fuel rods depending on their concentration, but some of them are removed by the purification system.

In order to obtain the corrosion potential of the structural material surface at any location, the concentration of dissolved oxygen, hydrogen peroxide, First, determine the concentration and dissolved hydrogen concentration. Next, using these as inputs, the corrosion potential is obtained by calculating the flow velocity and electrochemical reaction at each site. Furthermore, under the noble metal injection condition, the electrochemical reaction of hydrogen is catalyzed by the noble metal adhering to the structural material, so the corrosion potential is calculated by incorporating the amount of adhering noble metal into the electrochemical reaction calculation. Patent Literature 3 shows the relationship between the amount of deposited noble metal and the corrosion potential, etc., regarding the calculation of the corrosion potential when noble metal injection is applied.

In order to suppress stress corrosion cracking, it is necessary to maintain the corrosion potential below a value of -300 to -200 mV vs. SHE. It is necessary to confirm the decrease by actual measurement or calculation. For the calculation, the power plan of the core, the planned feedwater hydrogen injection amount, and the water quality around the structural materials of each part to be evaluated, that is, the concentrations of oxygen, hydrogen peroxide, and hydrogen, are required. Also, the temperature and mass transfer coefficient are required, and if noble metal implantation is applied, the amount of noble metal deposited on the surface of the structural material is required.

Of these, the planned power output of the core and the planned amount of feedwater hydrogen injection are the values that can be obtained in chronological order, but the water quality and the amount of precious metal deposits in the parts to be protected from stress corrosion cracking in the reactor are different during operation. Actual measurements in time series cannot be obtained. Therefore, it is necessary to obtain it by calculation, and it is obtained by a physical and chemical simulation model (see Patent Document 3). In Patent Document 3, the corrosion potential at the time of precious metal injection is analyzed using the current water quality conditions as an input, and the value of the corrosion potential is obtained as the effect of suppressing stress corrosion cracking. At this time, the amount of deposited precious metal is also estimated by the model and input manually. By comparing the prediction result of the noble metal deposit amount with the actually measured deposit amount of the coupon analyzed outside the furnace, the estimated value of the precious metal deposit amount is corrected to improve the accuracy.

Japanese Patent Application Laid-Open No. 2006-010428 JP-A-06-034786 Japanese Patent Application Laid-Open No. 2003-222697

The technique described in Patent Literature 3 optimizes the model parameters of a model set based on a physical model or a chemical model and adjusts the contribution thereof, and the model needs to be prepared in advance. Therefore, the correlation between the state quantity of the prediction target and the state quantity used as the input must be described as a mathematical formula.

In the technique described in Patent Document 3, since the true value of the amount of precious metal adhering to the surface of piping and structural materials in the furnace is unknown, the amount of adhesion to the surface of piping and structural materials in the furnace and the amount of adhesion to coupons outside the furnace It is necessary to obtain the correlation of This is difficult in terms of cost, and it may be difficult to obtain a good relationship technically due to the difference in flow conditions and water quality inside and outside the reactor. It is desirable to manage the deposit amount of noble metal in the furnace based on the information in the furnace. One approach is to simultaneously monitor changes in deposition by monitoring changes in corrosion potential in the furnace. However, the corrosion potential changes not only due to changes in the amount of precious metal deposited, but also when water quality changes due to changes in the operating conditions of the plant. rice field. In addition, since the measured values only indicate the current state, there is a problem that future predictions such as model calculations cannot be made.

The present invention has been made in view of such a background, and an object of the present invention is to provide a highly accurate prediction model construction device and prediction device for corrosion potentials related to structural materials of nuclear power plants.

In order to solve the above-described problems, a prediction model construction device is a prediction model construction device for constructing a prediction model for predicting the corrosion potential of a structural material of a nuclear reactor in a nuclear power plant, the reactor water of the nuclear reactor A machine learning model is trained on supervised data including as input data measurable plant state quantities including at least one of water quality, core power, and radiation monitoring data, and including the corrosion potential, which is an actually measured value, as correct data. It has a learning unit that builds a prediction model.

Further, in order to solve the above-mentioned problems, the prediction device is a prediction device for predicting the corrosion potential of structural materials of a nuclear reactor in a nuclear power plant, and comprises reactor water quality, core power, and radiation monitoring data of the nuclear reactor. Using a prediction model built by having a machine learning model learn supervised data containing as input data a measurable plant state quantity including at least one of the above, and including the corrosion potential, which is a measured value, as correct data, the actual measurement a prediction unit for calculating the corrosion potential as a prediction result by using the plant state quantity calculated as an input.

Advantageous Effects of Invention According to the present invention, it is possible to provide a highly accurate prediction model construction device and a prediction device for corrosion potentials related to structural materials of nuclear power plants. Problems, configurations, and effects other than those described above will be clarified by the following description of the embodiments.

1 is an overall system configuration diagram of a nuclear power plant according to this embodiment; FIG. FIG. 4 is a diagram for explaining determinants of corrosion potential according to the present embodiment; 1 is a functional configuration diagram of a corrosion potential prediction device according to this embodiment; FIG. FIG. 4 is a diagram for explaining the configuration of teacher data of a prediction model in learning processing according to the embodiment; FIG. 4 is a diagram for explaining input data (explanatory variables) and output data (objective variables, correct data) of a prediction model included in teacher data according to the present embodiment; 4 is a flowchart of learning processing executed by a learning unit according to the embodiment; It is a figure for demonstrating the structure of the input data of the prediction model in the prediction process which concerns on this embodiment. It is a figure for demonstrating the input data and output data (prediction result) of the prediction model in the prediction process which concerns on this embodiment. 4 is a flowchart of prediction processing executed by a prediction unit according to the embodiment; FIG. 11 is a diagram for explaining the configuration of teacher data of a prediction model in learning processing according to a modification of the embodiment; It is a figure for demonstrating the structure of the input data of a prediction model, and output data (prediction result) in the prediction process which concerns on the modification of this embodiment. FIG. 11 is a diagram for explaining the configuration of teacher data of a prediction model in learning processing according to a modification of the embodiment; It is a figure for demonstrating the structure of the input data of a prediction model in the prediction process which concerns on the modification of this embodiment, and output data. 4 is a graph for explaining prediction of corrosion potential according to the present embodiment;

≪Overview of Nuclear Power Plant≫
FIG. 1 is an overall system configuration diagram of a nuclear power plant P100 according to this embodiment. A schematic configuration of a nuclear power plant P100 (for example, a boiling water nuclear power plant) to which the corrosion potential prediction device for nuclear reactor structural materials according to the present embodiment is applied will be described with reference to FIG.
The nuclear power plant P100 includes a reactor P1, a turbine P3, a condenser P4, a reactor cleanup system, a water supply system, and the like. A reactor P1 installed in a reactor containment vessel P11 has a reactor pressure vessel P12 containing a core P13. A cylindrical core shroud P15 installed in the reactor pressure vessel P12 surrounds the core P13.

A plurality of fuel assemblies (not shown) are loaded in the core P13. Each fuel assembly contains a plurality of fuel rods filled with a plurality of fuel pellets made of nuclear fuel material. An annular downcomer P17 is formed between the inner surface of the reactor pressure vessel P12 and the outer surface of the core shroud P15. A plurality of jet pumps P21 are installed at the bottom of the reactor pressure vessel P12. The outlet of the jet pump P21 is arranged below the downcomer P17 and communicates with the lower space (lower plenum (not shown)) of the reactor pressure vessel P12.

The feedwater system includes a feedwater pipe P10 connecting the condenser P4 and the reactor pressure vessel P12, a condensate pump P5, a condensate purification device P6, a feedwater pump P7, a low-pressure feedwater heater P8, and a high-pressure feedwater heater P9. are installed in this order from the condenser P4 toward the reactor pressure vessel P12. A hydrogen injection device P16 is connected to the water supply pipe P10 by a hydrogen injection pipe P18 between the condensate purification device P6 and the water supply pump P7. An on-off valve P19 is provided on the hydrogen injection pipe P18. A precious metal injection device P31 is connected to the water supply pipe P10 by a precious metal injection pipe P32. An on-off valve P33 is provided in the precious metal injection pipe P32.

The reactor cleanup system includes a cleanup system pipe P20 (carbon steel member) connecting the reactor pressure vessel P12 and the feed water pipe P10, a cleanup system isolation valve P23, a cleanup system pump P24, a regenerative heat exchanger P25, and a non-regenerative heat exchanger. A vessel P26 and a reactor water purifier P27 are installed in this order.

Cooling water (reactor water) existing in the downcomer P17 in the reactor pressure vessel P12 is sucked into the driving fluid of the jet pump P21 and led to the lower plenum below the reactor core P13. Reactor water is supplied to the core P13 from the lower plenum and heated by heat generated by nuclear fission of the nuclear fuel material contained in the fuel rods of the fuel assemblies. A portion of the heated reactor water becomes steam. This steam is led from the reactor pressure vessel P12 through the main steam line P2 to the turbine P3 to rotate the turbine P3. A generator (not shown) coupled to turbine P3 is rotated to generate electrical power. Steam discharged from turbine P3 is condensed into water in condenser P4.

This water is supplied as feed water into the reactor pressure vessel P12 through the feed water pipe P10. The feed water flowing through the feed water pipe P10 is pressurized by the condensate pump P5, impurities are removed by the condensate purifier P6, further pressurized by the feed water pump P7, and heated by the low pressure feed water heater P8 and the high pressure feed water heater P9. be. Bleed steam extracted from the main steam pipe P2 and the turbine P3 through the bleed pipe P14 is supplied to the low-pressure feed water heater P8 and the high-pressure feed water heater P9, respectively, and serves as a heat source for feed water.

Since the reactor water in the reactor pressure vessel P12 contains metal corrosion products contained in the feed water and products generated by corrosion of structural materials in the reactor pressure vessel P12, a certain proportion of the reactor water Purified by a purification system. The reactor water in the reactor pressure vessel P12 is supplied to the regenerative heat exchanger P25 and the non-regenerative heat exchanger P26 through the cleanup system pipe P20 branched from the residual heat removal system pipe (not shown) by driving the cleanup system pump P24. and cooled to about 50°C by these heat exchangers. Metal corrosion products contained in the reactor water are removed by the cooled reactor water passing through the reactor water purifier P27, and after being heated in the regenerative heat exchanger P25, it joins the feed water flowing through the feed water pipe P10. and supplied to the reactor pressure vessel P12.

When shutting down reactor P1, all control rods (not shown) are inserted into the core. By inserting all control rods, the nuclear fission reaction of the nuclear fuel material is stopped and the operation of the reactor P1 is stopped. The heat remaining in the equipment in the core P13 and the reactor pressure vessel P12 is removed by the evaporation of the reactor water. Therefore, when the reactor water temperature drops to about 150° C., the components in the core P13 and the reactor pressure vessel P12 are cooled using the residual heat removal system.

≪Hydrogen injection and precious metal injection≫
Hydrogen injection or a combination of hydrogen injection and noble metal injection is carried out from the viewpoint of suppressing stress corrosion cracking of nuclear reactor structural materials. Hydrogen is injected from the hydrogen injection device P16 through the hydrogen injection pipe P18 into the water supply by opening the on-off valve P19. Hydrogen injected into the feedwater is jetted out from a feedwater sparger (not shown) above the downcomer P17 into the reactor water and mixed with the reactor water. Since the dose rate of gamma rays in the downcomer P17 region is moderate and promotes the recombination reaction, the added hydrogen recombines with oxygen and hydrogen peroxide in the reactor water when passing through the downcomer P17 to produce water. That is, the corrosion environment is alleviated by consuming the dissolved oxygen and the hydrogen added with the hydrogen peroxide, which are the environmental factors of stress corrosion cracking. As a result, the corrosion potential of the structural material that comes into contact with the reactor water is lowered, thereby suppressing the occurrence and progress of stress corrosion cracking.

Also, noble metal injection is a technique that enhances the effect of hydrogen injection by catalytic action. An aqueous solution of, for example, a compound of platinum (Pt) is injected from the precious metal injection device P31 into the feed water through the precious metal injection pipe P32 by opening the on-off valve P33. The platinum injected into the feed water joins the reactor water, and part of the platinum particles adhere to the surface of the structural material in contact with the reactor water in the form of minute platinum particles. The remainder adheres to the fuel surface. Oxygen and hydrogen peroxide in the reactor water electrochemically catalytically react with hydrogen to form water on the surface of the platinum particles. If hydrogen is present in a stoichiometric ratio (H:O= ₂ :1 for H2O) of 2 or more than oxygen (hydrogen peroxide is treated as 1/2 equivalent oxygen), then hydrogen It becomes a surplus, and the reaction between hydrogen and oxygen proceeds at a potential around -500mV vs. SHE, which is the hydrogen redox potential. For this reason, the corrosion potential of the structural material is mixed and decreases to around -500mV vs. SHE, and becomes below the target potential. This can suppress stress corrosion cracking.

A corrosion potential sensor is temporarily or permanently installed in the reactor P1 as a means of confirming the effect of hydrogen injection or noble metal injection on stress corrosion cracking. The bottom drain line P34 draws out reactor water from the bottom of the reactor pressure vessel P12 and is connected to the purification system pipe P20. A branch line is provided from this bottom drain line P34, and a corrosion potential sensor P35 is installed there to measure the corrosion potential of the water quality at the bottom of the reactor pressure vessel P12. The branch line is connected to a reactor water sampling line (not shown) or the like.

In the case of a BWR, the reactor water below the downcomer P17 flows through the recirculation system piping P30 and the purification system piping P20. Corrosion potential measurements are made by installing branch lines or flanges loaded with corrosion potential sensors. Further, in the case of ABWR, the purification system pipe P20 can be used to measure the corrosion potential in the water quality of the region above the downcomer P17. For example, a flange P36 may be installed on the recirculation pipe P30, and a corrosion potential sensor P35 may be installed to measure the corrosion potential. Alternatively, a branch line may be provided to enable the same measurement as the bottom drain.

<<Determinants of Corrosion Potential>>
FIG. 2 is a diagram for explaining determinants of the corrosion potential according to this embodiment. As shown in FIG. 2, the corrosion potential during noble metal implantation is determined by the key parameters of oxygen, hydrogen peroxide and hydrogen concentrations, platinum coverage, and mass transfer coefficients at the site of interest. The amount of deposited platinum is determined by the amount of platinum injected and the state of the material. The mass transfer coefficient is determined by the reactor water flow conditions, which are determined by the plant design. Oxygen, hydrogen peroxide, and hydrogen are generated by radiolysis of reactor water, and the amount of radiolysis is determined by the dose rate distribution in the reactor and the quality of feed water. The dose rate distribution is the distribution of gamma rays and neutrons emitted from the nuclear fuel in the core, and is determined by the design conditions such as the power and power distribution of the core and the size and structure of the reactor.

In addition to the factors shown in FIG. 2, hydrogen injection is another factor that affects the corrosion potential. Specifically, the hydrogen injection changes the water quality of the reactor water from oxidizing to reducing, reducing the concentration of oxygen and hydrogen peroxide. do. Therefore, the amount of radioactive N13 and N16 transferred to the main steam system changes. Also, iodine changes from IO ₃ ⁻ to I ⁻ . The amount and form (solubility, insolubility) of the metal elements also change as the concentration of the metal elements in the reactor water changes from oxidizing to reducing. If the dose rate distribution changes due to changes in operating conditions, the concentrations of oxygen and hydrogen peroxide will change even if the hydrogen concentration in the feedwater is the same.

In the present invention, a model (prediction model) for predicting the corrosion potential (objective variable) from the measurable plant state quantity (explanatory variable) that is the determinant of the corrosion potential is constructed, and the plant state is constructed using this prediction model. Corrosion potential is predicted from the quantity. Specifically, a prediction model is generated using machine learning technology from learning data (teacher data) including plant state quantities and actually measured or calculated corrosion potentials. The prediction model thus generated reflects the conditions and correlations of the plant state variables that affect the corrosion potential, and the corrosion potential can be predicted from the plant state variables.

≪Corrosion Potential Predictor for Structural Materials: Overall Configuration≫
A corrosion potential prediction device for predicting the corrosion potential of the structural material of the nuclear reactor in the nuclear power plant P100 according to this embodiment will be described below. The apparatus for predicting corrosion potential of reactor structural materials generates a machine learning model (prediction model) for predicting the corrosion potential of reactor structural materials, and predicts the corrosion potential using this prediction model.

FIG. 3 is a functional configuration diagram of the corrosion potential prediction device 100 according to this embodiment. Corrosion potential prediction device 100 is a computer and includes control unit 110 , storage unit 120 and input/output unit 160 . Corrosion potential prediction device 100 receives operational data such as thermal output from process computer P110 used in nuclear power plant P100 and output data from measuring instrument P120 installed in nuclear power plant P100.

The storage unit 120 includes a RAM (Random Access Memory), ROM (Read Only Memory), SSD (Solid State Drive), etc., and stores a prediction model 130, a plant state quantity database 150, and a program 121. The prediction model 130 is a learning model of machine learning, such as a neural network. The program 121 includes descriptions of learning processing (see FIG. 6) and prediction processing (see FIG. 9), which will be described later.

The plant state quantity database 150 stores plant state quantities including plant data, feed water data, and reactor water quality data of the nuclear power plant P100. The plant state quantity database 150 also stores plant state quantities such as the residence time of the fuel assemblies loaded in the core P13 (see FIG. 1) and the electrical output.

The input/output unit 160 receives data from the process computer P110 and the measuring instrument P120, and stores the data in the plant state quantity database 150. FIG. The input/output unit 160 also includes a display, keyboard, and mouse (not shown), receives operations from the user of the corrosion potential prediction device 100, and displays data such as prediction results.

The control unit 110 includes a CPU (Central Processing Unit) and includes a learning unit 111 and a prediction unit 112 . The learning unit 111 generates teacher data (learning data, see FIG. 4 to be described later) from data stored in the plant state quantity database 150, performs learning processing (see FIG. 6 to be described later), and performs corrosion of reactor structural materials. Generate a predictive model 130 that predicts potential. The prediction unit 112 inputs data stored in the plant state quantity database 150 to the generated prediction model 130, and executes a prediction process (see FIG. 9 described later) for predicting the corrosion potential of the reactor structural material. Details of the processing of the learning unit 111 and the prediction unit 112 will be described later with reference to FIGS. 4 to 9. FIG.
Although the prediction model 130 itself is data and does not predict the corrosion potential, the prediction model 130 is regarded as a function part (the prediction model 130 and the prediction part 112 are integrated), and the prediction model 130 predicts the corrosion potential. It is also written as predicting.

<<Teacher Data>>
FIG. 4 is a diagram for explaining the configuration of teacher data of the prediction model 130 in the learning process according to this embodiment. FIG. 5 is a diagram for explaining input data (explanatory variables) and output data (objective variables, correct data) of the prediction model 130 included in the teacher data according to this embodiment. Here, it is assumed that a prediction model for predicting the corrosion potential after 30 days from the plant state quantity for 30 days is generated.
The data set 210 shown in FIG. 4 is a data set included in the training data with the first day as the start date, and includes the plant state quantities from the first day to the 30th day and the corrosion potential on the 60th day. These data are stored in the plant state quantity database 150 .

Measurable plant state quantities (actual measurement data) included in each of the plant state quantities on the 1st to 30th days are included as input data of the prediction model 130 . The teacher data, which is the input data of this prediction model 130, corresponds to the plant state quantity 151 shown in FIG.
The measurable plant state variables used as input data are oxygen, hydrogen, hydrogen peroxide, iron, zinc, nickel, cobalt, platinum, and rhodium contained in the reactor water quality. There are at least one of the concentrations, the core outermost peripheral power as the core power, and the radiation data of N16 as the radiation monitoring data. The input data of the predictive model 130 may include plant state quantities, which will be described later.

The corrosion potential (measured value) on the 60th day is included as output data (correct data) of the prediction model 130 . The teacher data, which is the output data of this predictive model 130, corresponds to the corrosion potential 152 shown in FIG.
The input data and output data described above form a set of teacher data for the prediction model 130 . Similarly, a set of teacher data is generated from the data set 220 starting on the second day. By repeating this, 30 sets of teaching data for the prediction model 130 are generated from the plant state quantities on the 1st to 59th days and the corrosion potential on the 60th to 89th days.

Specifically, assuming that X=1, . Core inlet temperature of reactor water, core outlet temperature of reactor water, linear flow velocity in each region in the pressure vessel, injection amount of precious metals, concentration of reactor water metals (Fe, Ni, Cr, Cu, Co), thermal power, core fuel assemblies Axial power distribution and radial power distribution, nuclear fuel assembly power, core power distribution, electrical power, radiation monitoring data (N16, N13, off-gas nuclide), feedwater hydrogen concentration, main steam system oxygen and hydrogen concentration, radioactivity concentration at reactor water sampling points (Co60, Co58, Cr51, Na24, Au199), reactor water impurity concentration (ions, such as chloride ions, sulfate ions, chromate ions, insoluble components), dissolved oxygen concentration , dissolved hydrogen concentration, hydrogen peroxide concentration, nitrogen compound concentration, iodine compound concentration, etc., or the plant state quantities 151 already stored in the process computer P110 (see FIG. 3) are included as input data for teacher data. be The output data (correct data) for this input data is the actually measured value of the corrosion potential on the X+59th day.

≪Corrosion potential prediction device: learning processing≫
FIG. 6 is a flowchart of learning processing executed by the learning unit 111 according to this embodiment.
In step S11, the learning unit 111 starts a process of repeatedly executing step S12 for each predetermined start date (1st to 30th days).

In step S12, the learning unit 111 generates one data set of teacher data. Specifically, the learning unit 111 uses the feedwater flow rate, core flow rate, main steam flow rate, recirculation flow rate, reactor water purification flow rate, reactor water core inlet temperature, reactor Core outlet temperature of water, linear flow velocity in each region in pressure vessel, precious metal injection amount, reactor water metal concentration (Fe, Ni, Cr, Cu, Co), thermal power, axial power based on power of each core fuel assembly distribution and radial power distribution, nuclear fuel assembly power, core power distribution, electrical power, radiation monitoring data (N16, N13, offgas nuclides), feedwater hydrogen concentration, main steam system oxygen/hydrogen concentration, reactor water sampling point Radioactivity concentration (Co60, Co58, Cr51, Na24, Au199), reactor water impurity concentration (ions such as chloride ions, sulfate ions, chromate ions, insoluble components), dissolved oxygen concentration, dissolved hydrogen concentration, hydrogen peroxide Concentration, nitrogen compound concentration, iodine compound concentration, and other measurable plant state quantities 151 are input data, and corrosion potential 152, which is actually measured 59 days after the start date, is output data (correct data). Generate a dataset.

In step S<b>13 , learning unit 111 generates prediction model 130 by training prediction model 130 using the teacher data generated in step S<b>12 (making prediction model 130 learn the teacher data).
By the learning process described above, the prediction model 130 for predicting the corrosion potential of the reactor structural material can be constructed. Next, the process of predicting the corrosion potential of reactor structural materials using the predictive model 130 will be described.

≪Reactor water radioactivity concentration prediction device: prediction processing≫
FIG. 7 is a diagram for explaining the configuration of input data of the prediction model 130 in the prediction processing according to this embodiment. FIG. 8 is a diagram for explaining input data and output data (prediction result) of the prediction model 130 in the prediction processing according to this embodiment.
In FIG. 7, the reference date is the execution date of the prediction process, and the prediction unit 112 predicts the corrosion potential 30 days after the reference date from the plant state quantities for the past 30 days including the reference date.

Specifically, from 29 days before the reference date to the reference date included in the plant state quantity database 150, the feed water flow rate, core flow rate, main steam flow rate, recirculation flow rate, reactor water purification flow rate, reactor water core inlet temperature, reactor water core Outlet temperature, linear flow velocity in each region in the pressure vessel, precious metal injection amount, reactor water metal concentration (Fe, Ni, Cr, Cu, Co), thermal power, axial power distribution and diameter based on the power of each core fuel assembly Directional power distribution, nuclear fuel assembly power, core power distribution, electrical power, radiation monitoring data (N16, N13, offgas nuclides), feedwater hydrogen concentration, main steam system oxygen/hydrogen concentration, radioactivity concentration at reactor water sampling points (Co60, Co58, Cr51, Na24, Au199), reactor water impurity concentration (ions such as chloride ions, sulfate ions, chromate ions, insoluble components), dissolved oxygen concentration, dissolved hydrogen concentration, hydrogen peroxide concentration, nitrogen Measurable plant state quantities (measured data) such as compound concentrations and iodine compound concentrations are included as input data for the prediction model 130 . This input data corresponds to the plant state quantity 153 shown in FIG.
The corrosion potential 30 days after the reference date is the output data (prediction result) of the prediction model 130 . This output data corresponds to the corrosion potential 154 shown in FIG.

FIG. 9 is a flowchart of prediction processing executed by the prediction unit 112 according to this embodiment.
In step S<b>21 , the prediction unit 112 generates input data for the prediction model 130 . Specifically, the prediction unit 112 predicts the feedwater flow rate, core flow rate, main steam flow rate, recirculation flow rate, reactor water cleaning flow rate, reactor water core Inlet temperature, core outlet temperature of reactor water, linear flow velocity in each region in the pressure vessel, precious metal injection amount, reactor water metal concentration (Fe, Ni, Cr, Cu, Co), thermal power, power of each core fuel assembly Axial power distribution and radial power distribution, nuclear fuel assembly power, core power distribution, electrical power, radiation monitoring data (N16, N13, off-gas nuclide), feedwater hydrogen concentration, main steam system oxygen/hydrogen concentration, reactor water Radioactivity concentration at sampling points (Co60, Co58, Cr51, Na24, Au199), reactor water impurity concentration (ions, such as chloride ions, sulfate ions, chromate ions, insoluble components), dissolved oxygen concentration, dissolved hydrogen concentration , hydrogen peroxide concentration, nitrogen compound concentration, iodine compound concentration, and other plant state quantities that can be actually measured to generate input data for the prediction model 130 .
In step S<b>22 , the prediction unit 112 predicts the corrosion potential by calculating the corrosion potential 154 that is the output for the input data generated in step S<b>21 using the prediction model 130 .

If the reference date is now, the corrosion potential 154 can be predicted from the plant state quantity 153, which is the actual value (actually measured value) for 30 days. Also, if the reference date is in the past, the forecast can be verified. Specifically, the prediction accuracy can be evaluated by calculating the corrosion potential 154 from the plant state quantity 153, which is the past actual value, and comparing it with the actual value of the corrosion potential. For example, the prediction accuracy can be evaluated by comparing the corrosion potential when the reference date is 30 days ago and the corrosion potential of today.

≪Characteristics of Corrosion Potential Prediction Device≫
The corrosion potential prediction device 100 constructs a prediction model 130 for predicting the corrosion potential from the measurable plant state quantities, and uses the prediction model 130 to predict the corrosion potential from the plant state quantities. Specifically, the prediction model 130 is generated using machine learning technology from teacher data (learning data) including plant state quantities and actually measured or calculated corrosion potentials. The prediction model 130 generated in this way reflects the conditions and correlations of the plant state quantities that affect the corrosion potential, and can predict the corrosion potential from the plant state quantities.

≪Modification: Period of input/output data to prediction model≫
In the above-described embodiment, the data input to the prediction model 130 (see FIGS. 5 and 8) was for 30 days, but is not limited to this. may be data based on Further, although the corrosion potential 30 days after the reference date is predicted, the corrosion potential 90 days later may be predicted. Alternatively, the number of days to be predicted is not limited to one day, and may be set as a period such as 1 day after the reference date to 60 days after the reference date. By creating teacher data and building a prediction model 130 according to the desired period of input data and the interval between the reference date and the date for predicting the corrosion potential (prediction date), the corrosion potential prediction device 100 can respond Corrosion potential can be predicted based on the period of input data and the interval between the reference date and the forecast date.

In the above-described embodiment, the input data does not include plant state quantities between the reference date and the forecast date. Plant state quantities that are predetermined or predictable in the operation plan of the nuclear power plant may be added to the input data as plan values.
FIG. 10 is a diagram for explaining the configuration of teacher data of the prediction model 130 in the learning process according to the modified example of this embodiment. FIG. 11 is a diagram for explaining configurations of input data and output data (prediction result) of the prediction model 130 in prediction processing according to a modification of the present embodiment.
In the learning process, in addition to the plant state quantities for 30 days from the start date (see FIG. 4), the plant state quantities for 29 days are added to the input data of the teacher data. Also, the output data (correct data) shall be the actual measured values of the corrosion potential for 30 days after the 59th day from the start date.

A prediction model constructed using such teacher data can predict the corrosion potential for 30 days from the plant state quantity, which is the measured value or planned value for 59 days. In the prediction process, in addition to the plant state quantities for the past 30 days including the base date (see Fig. 7), the plant state quantities, which are planned values from the day after the base date to the day before the forecast date, are used as input data. Predict corrosion potential for 30 days from +30 days.

For example, the hydrogen injection rate, feedwater flow rate, core flow rate, recirculation system flow rate, reactor water cleaning flow rate, core power and power distribution are determined (predictable) in the operation plan or determined by the operation of the nuclear power plant. Because it is adjustable, it may be added to the input to predictive model 130 . By adding the input data of a day close to the expected date, the corrosion potential prediction device 100 can predict the corrosion potential with higher accuracy.

As another modification, the period of the input data and the period of the output data may overlap. FIG. 12 is a diagram for explaining the configuration of teacher data of the prediction model 130 in the learning process according to the modified example of this embodiment. FIG. 13 is a diagram for explaining configurations of input data and output data of the prediction model 130 in prediction processing according to the modification of the present embodiment.
In the learning process, in addition to the plant state quantities for 30 days from the start date, the plant state quantities for 29 days are added to the input data of the teacher data. The output data (correct data) are the measured values of the corrosion potential for 29 days after 30 days from the start date.

The prediction model constructed using such training data is calculated from the plant state quantity, which is the actual measured value for 30 days, and the plant state quantity, which is the planned value for 29 days. Corrosion potential can be predicted for the same period as quantity. In the prediction process, in addition to the plant state quantity for the past 30 days including the base date, the plant state quantity that is the planned value from the base date to the forecast date is used as input data, and the corrosion for 29 days after the base date Predict potential.

<<Modification: Input data>>
Although the plant state quantity to be input to the prediction model 130 in the above-described embodiment does not include the corrosion potential, it may include the corrosion potential actually measured up to the reference date. For example, in FIG. 12, the actually measured corrosion potential may be added to the plant state quantities for 30 days from the start date. The prediction model constructed using such teacher data is based on the plant state quantity and corrosion potential, which are the actual measured values for 30 days, and the plant state quantity, which is the planned value for 29 days. It is possible to predict the corrosion potential for the same period as the plant state quantity, which is a value. In the prediction process, in addition to the plant state quantity and corrosion potential for the past 30 days including the base date, the plant state quantity, which is the planned value from the base date to the forecast date, is used as input data. Predict corrosion potential for days.

≪Drafting an operation plan using a corrosion potential prediction device≫
In order to formulate an operation plan for suppressing the corrosion potential to a desired value or less, it is sufficient to repeatedly execute the prediction process while changing the plant state quantity (see the input data shown in FIG. 13), which is the planned value. For example, the plant state quantities include the precious metal concentration, zinc concentration, and iron concentration in feed water. Most of the precious metals injected into the reactor water are precipitated and attached together with iron oxides on the boiling fuel rod surfaces. The injection of precious metals into the feed water is continued for about 10 days after about two months after the start of operation of the plant, and thereafter the injection of precious metals into the feed water is stopped. The concentration of precious metals in the reactor water after stopping the injection of precious metals is almost determined by the elution from the precious metals attached to the surfaces of the fuel rods together with iron oxides during the period of the injection of precious metals. For this reason, it is considered that the concentration of precious metals in reactor water is likely to be affected by the concentration of co-precipitated iron in feed water. Zinc also acts to reduce deposition of precious metals during precious metal injection. The iron concentration and zinc concentration in the feed water can be adjusted even after the precious metal injection is stopped. It is possible to influence the water precious metal concentration.

FIG. 14 is a graph 410 for explaining prediction of corrosion potential according to this embodiment. The horizontal axis of the graph 410 is the operation date after the plant operation start date, and the vertical axis is the corrosion potential. The prediction model used for this prediction is based on the plant state quantities and corrosion potential measured from the start of operation to the reference date, and the planned plant state quantities from the day after the base date. Predict potential. From the 1st day, which is the operation start date, to the Nth day, which is the reference date, the values are actually measured values, and after the reference date, the predicted values of the corrosion potential based on the plant state quantities input as planned values are shown.

A graph 410 shows changes in corrosion potential caused by a combination of changes in reactor water quality and temporal changes in the deposition amount of precious metals when the plant is operated according to the plan after the day after the reference date. If the corrosion potential continues to rise with time, it is considered that the precious metal at the corrosion potential measurement site is reduced due to peeling and elution. The time point at which the corrosion potential is predicted to become equal to or higher than the target corrosion potential (referred to as target ECP in FIG. 14) is the precious metal re-injection time limit, and the precious metal re-injection is performed before that time. The prediction unit 112 may calculate and notify the point in time (operating day) when the predicted corrosion potential becomes equal to or higher than a preset target corrosion potential.
By repeatedly executing the prediction process while changing the planned plant state quantities such as feedwater hydrogen concentration and precious metal injection amount, the feedwater hydrogen concentration, Precious metal doses can be planned.

The deposition amount of noble metals on the surface of structural materials, which strongly affects the corrosion potential, is affected by the concentration of noble metals in the reactor water. Plant status quantities such as the amount of precious metals deposited on the surface of fuel rods, the amount of iron oxide deposited, and the amount of deposited precious metals on the surface of structural materials inside and outside the reactor cannot be measured during the operation of a nuclear power plant. It is a value that can only be measured by removing fuel rods from the reactor core P13 (see FIG. 1) or accessing structural materials inside and outside the reactor while the nuclear power plant is shut down.

The input data of the prediction model 130 in the above-described embodiment includes the amount of precious metals deposited on the fuel rods, the amount of precious metals deposited and accumulated on the surfaces of structural materials inside the reactor, and the amount of precious metals deposited and accumulated on the surfaces of structural materials outside the reactor. However, these cannot be measured during operation. Therefore, by measuring the corrosion potential in the recirculation system piping where peeling and elution are most likely to occur, in the case of ABWR, the furnace purification system piping, or the sampling line with the flow velocity equivalent to the recirculation system piping, Changes in precious metal loading can be implicitly added to the learning. For example, since the recirculation pipe has a high flow velocity of 10 m/s and the pipe system is also large, the surface shear stress and mass transfer coefficient are large, and peeling and elution proceed the fastest.

On the other hand, if the accumulation due to adhesion on the fuel surface is slow, re-adhesion to the recirculation system will occur, and the overall separation and elution will be delayed. Fit. Therefore, by measuring the corrosion potential in the recirculation system, it is possible to appropriately determine the timing of reinjection of the noble metal.
Similarly, in an ABWR without a recirculation system, the timing of reinjection can be determined by measuring the corrosion potential in the piping of a system with a high flow rate such as a furnace cleanup system. In the case of measurements at locations outside the reactor, proper corrosion control can be achieved by installing a measurement manifold designed so that the flow velocity at the measurement location is equivalent to the mass transfer amount and shear force of the recirculation system inside the reactor. Potential can be measured and predicted to determine when to reinject. Note that the measurement point of the corrosion potential is not limited to one, and may be plural points.

<<Modification: Separation of devices that execute learning processing and prediction processing>>
It should be noted that the present invention is not limited to the above-described embodiments, and can be modified without departing from the scope of the invention. For example, the learning process and prediction process performed by the corrosion potential prediction device 100 may be performed by another device. A prediction model construction device may execute learning processing to construct a prediction model, and a prediction device that acquires the constructed prediction model may execute prediction processing.

<<Other Modifications>>
In the above-described embodiment, data for each day is used as input data, but the input data is not limited to this. For example, data for every two days or data for every six hours may be used. Moreover, the intervals of input data farther from the predicted date may be longer than the constant intervals. By lengthening the interval and reducing the number of input data, the corrosion potential prediction device 100 can speed up the learning process and the prediction process.

Although several embodiments of the present invention have been described above, these embodiments are merely examples and do not limit the technical scope of the present invention. The present invention can take various other embodiments, and various modifications such as omissions and substitutions can be made without departing from the gist of the present invention. These embodiments and modifications thereof are included in the scope and gist of the invention described in this specification and the like, and are included in the scope of the invention described in the claims and equivalents thereof.

100 Corrosion potential prediction device (prediction model construction device, prediction device)
111 learning unit 112 prediction unit 130 prediction model 151, 153 plant state quantity 152, 154 corrosion potential P1 reactor P20 purification system piping (reactor purification system piping)
P30 Recirculation system piping P100 Nuclear power plant

Claims (9)

A prediction model building device for building a prediction model for predicting the corrosion potential of a structural material of a nuclear reactor in a nuclear power plant,
Machine teacher data including as input data a measurable plant state quantity including at least one of reactor water quality, core power, and radiation monitoring data of the reactor, and including as correct data the corrosion potential, which is an actually measured value. A predictive model construction device comprising a learning unit for constructing a predictive model by causing a learning model to learn. The reactor water quality is the concentration of at least one of oxygen, hydrogen, hydrogen peroxide, iron, zinc, nickel, cobalt, platinum, and rhodium contained in the reactor water of the nuclear reactor,
The core power is the core outermost power,
The predictive model construction device according to claim 1, wherein the radiation monitoring data is N16 radiation data. the reactor water quality further includes at least one of impurity concentration, nitrogen compound concentration, and iodine compound concentration;
the core power further comprises at least one of thermal power, axial power distribution and radial power distribution based on power per core fuel bundle, nuclear fuel bundle power, core power distribution, and electrical power;
the radiation monitoring data further includes radiation data of at least one of N13, an off-gas nuclide;
The plant state variables include feedwater flow rate, core flow rate, main steam flow rate, recirculation flow rate, reactor water cleaning flow rate, reactor water core inlet temperature, reactor water core outlet temperature, linear flow velocity in each region of the pressure vessel, and precious metal injection. feedwater hydrogen concentration, main steam system oxygen concentration, main steam system hydrogen concentration, and radioactivity concentration at the reactor water sampling point. building equipment. 2. The predictive model construction device according to claim 1, wherein the corrosion potential is measured at at least one of the recirculation system piping and the furnace cleaning system piping. A prediction device for predicting the corrosion potential of a structural material of a nuclear reactor in a nuclear power plant,
Machine teacher data including as input data a measurable plant state quantity including at least one of reactor water quality, core power, and radiation monitoring data of the reactor, and including as correct data the corrosion potential, which is an actually measured value. Using a prediction model built by learning a learning model,
A prediction device comprising: a prediction unit that receives the actually measured plant state quantity or the actually measured plant state quantity and the planned value as input and calculates the corrosion potential as a prediction result. The reactor water quality is the concentration of at least one of oxygen, hydrogen, hydrogen peroxide, iron, zinc, nickel, cobalt, platinum, and rhodium contained in the reactor water of the nuclear reactor,
The core power is the core outermost power,
The prediction device according to claim 5, wherein the radiation monitoring data is N16 radiation data. the reactor water quality further includes at least one of impurity concentration, nitrogen compound concentration, and iodine compound concentration;
the core power further comprises at least one of thermal power, axial power distribution and radial power distribution based on power per core fuel bundle, nuclear fuel bundle power, core power distribution, and electrical power;
the radiation monitoring data further includes radiation data of at least one of N13, an off-gas nuclide;
The plant state variables include feedwater flow rate, core flow rate, main steam flow rate, recirculation flow rate, reactor water cleaning flow rate, reactor water core inlet temperature, reactor water core outlet temperature, linear flow velocity in each region of the pressure vessel, and precious metal injection. feedwater hydrogen concentration, main steam system oxygen concentration, main steam system hydrogen concentration, and radioactivity concentration at the reactor water sampling point. . 6. The predicting device according to claim 5, wherein the corrosion potential is measured at at least one of recirculation system piping and furnace cleaning system piping. The prediction device according to claim 5, wherein the prediction unit calculates a point in time when the corrosion potential, which is the prediction result, becomes equal to or higher than a preset target corrosion potential.

Images