Predictive model construction and prediction method for radioactive metal corrosion concentration in nuclear reactor water

A prediction method of a reactor water radioactivity concentration prediction device that predicts a radioactive metal corrosion product concentration in reactor water of a nuclear reactor in a nuclear power plant,

• the prediction device stores a predictive model that is obtained by causing a machine learning model to learn learning data for supervised learning, the learning data for supervised learning including a plant state quantity that is able to be actually measured and includes at least one of a flow rate of feedwater, a metal corrosion product concentration in feedwater, a metal corrosion product concentration in reactor water, a radioactive metal corrosion product concentration in reactor water, a flow rate of reactor water cleanup system, duration of stay in reactor core in a fuel assembly, and electrical output of the nuclear reactor and a plant state quantity prediction value calculated by using a physical model that describes plant state quantities of the nuclear power plant including the flow rate of feedwater and the metal corrosion product concentration in feedwater of the reactor water in the nuclear reactor as input data and including a radioactive metal corrosion product concentration in the reactor water as output data, wherein the radioactive metal corrosion product concentration is an actual measured value, the prediction method comprising:
• the prediction device executing:
• a step of calculating the plant state quantity prediction value based on the plant state quantities by using the physical model;
• a step of inputting the calculated plant state quantity prediction value and a plant state quantity that includes at least one of the flow rate of feedwater, the metal corrosion product concentration in feedwater, the metal corrosion product concentration in reactor water, the radioactive metal corrosion product concentration in reactor water, the flow rate of reactor water cleanup system, the duration of stay in reactor core in the fuel assembly, and the electrical output of the nuclear reactor into the predictive model as input data and calculating a radioactive metal corrosion product concentration in the reactor water as output data;
• and
• a step of using the predictive model to predict radioactive corrosion product concentration on nuclear power plant structures and lower an exposure dose of work personnel during periodic inspection of the structures, wherein
• the plant state quantity to be input to the predictive model is actually measured data at an input time point, and
• the plant state quantity to be input to the predictive model further includes a planned value from the input time point to a predicted time point of the radioactive metal corrosion product concentration in reactor water.

A prediction method of a reactor water radioactivity concentration prediction device that predicts a radioactive metal corrosion product concentration in reactor water of a nuclear reactor in a nuclear power plant,

• the prediction device stores a predictive model that is obtained by causing a machine learning model to learn learning data for supervised learning, the learning data for supervised learning including a plant state quantity that is able to be actually measured and includes at least one of a flow rate of feedwater, a metal corrosion product concentration in feedwater, a metal corrosion product concentration in reactor water, a radioactive metal corrosion product concentration in reactor water, a flow rate of reactor water cleanup system, duration of stay in reactor core in a fuel assembly, and electrical output of the nuclear reactor and a plant state quantity prediction value calculated by using a physical model that describes plant state quantities of the nuclear power plant including the flow rate of feedwater and the metal corrosion product concentration in feedwater of the reactor water in the nuclear reactor as input data and including a radioactive metal corrosion product concentration in the reactor water as output data, wherein the radioactive metal corrosion product concentration is an actual measured value, the prediction method comprising:
• the prediction device executing:
• a step of calculating the plant state quantity prediction value based on the plant state quantities by using the physical model;
• a step of inputting the calculated plant state quantity prediction value and a plant state quantity that includes at least one of the flow rate of feedwater, the metal corrosion product concentration in feedwater, the metal corrosion product concentration in reactor water, the radioactive metal corrosion product concentration in reactor water, the flow rate of reactor water cleanup system, the duration of stay in reactor core in the fuel assembly, and the electrical output of the nuclear reactor into the predictive model as input data and calculating a radioactive metal corrosion product concentration in the reactor water as output data;
• and
• a step of using the predictive model to predict radioactive corrosion product concentration on nuclear power plant structures and lower an exposure dose of work personnel during periodic inspection of the structures, wherein
• the physical model is a mass balance model that uses the flow rate of feedwater and the metal corrosion product concentration in feedwater of the reactor water of the nuclear reactor and is related to a migration behavior of a metal corrosion product and a radioactive metal corrosion product in the reactor water, and
• the plant state quantity prediction value includes at least one of an amount of metal corrosion product and an amount of radioactive metal corrosion product accumulated on a fuel rod of the nuclear reactor, the amount of metal corrosion product and the amount of radioactive metal corrosion product being calculated by using the mass balance model.