Research On Multi-objective Optimization Decision-making Method For Small Integrated Reactor Operation

With the development of nuclear energy in our country,the research on intelligent nuclear energy systems has become a hot topic.In recent years,the demand for research on unmanned and autonomous nuclear energy control systems is gradually increasing.Due to the reactor still need the operator to make decisions,the decisions are made by the operator who combine operation data and operation plan.The decisions made by operators in a short period of time cannot fully consider numerous operational data,and these decisions may not be the optimal ones for the operational task.Therefore,it is meaningful to research the multi-objective optimization decision-making method for small integrated reactor operation(SIR).In this paper,a small integrated reactor simulation model is used as the object of research.This research develops a multi-objective optimization A3 C algorithm based on the Asynchronous Advantage Actor-Critic algorithm.The evaluation index system for the economy,safety and flexibility of the operation process are developed to be the optimal object for MO-A3 C.The data of the small integrated reactor simulation model at different power levels for steady-state operating conditions are collected to be the training data set.The training data set is used to train a deep neural network model to obtain a decision environment agent model.The decision environment agent model is used as the reinforcement learning environment to improve the speed of the MO-A3 C algorithm’s optimization-seeking iterations.The research of multi-objective optimization decision method for small integrated reactor operation is carried out on this basis.This research takes as input the control settings of the reactor operation,the variables related to the evaluation system.The reward function is developed according to the degree of safety,economy and flexibility considered for the operation of the task.This research takes as output the steady state operating conditions of the reactor.The normal reactor operating conditions and main pump speed failure conditions under increased capacity and reduced capacity tasks are taken as tests.The optimization results are validated by decision results to verify that the operating parameters are within the operating limits during the transient transition of the reactor from initial to optimized operating conditions.The multi-objective optimization decision-making method for small integrated reactor operation designed in this article implements a data-driven and empirical knowledge-based decision making for reactor operation,which is data-driven instead of human subjective weighting for optimization.The optimization decision result of the normal reactor operating conditions and main pump speed failure conditions under increased capacity and reduced capacity tasks are in line with the operational task requirements.So this research has reference value for the future research on unmanned small integrated reactor operation systems.