Abstract
Small modular reactors (SMRs) are ideal energy sources for remote areas, but they still require operators to make decisions. Due to the harsh working environment, it is difficult to resume operation of small modular reactors once an accident occurs in unmanned application scenarios. Therefore, a multi-objective operation scheme that ensures safety and takes into account the operational task requirements is particularly important. We proposed an artificial intelligence approach for multi-objective optimization and decision-making for SMRs. We used a multi-objective extension of the Asynchronous Advantage Actor-Critic (A3C) algorithm to learn a comprehensive optimization function reflecting the importance of each objective under different tasks. We also used a proxy deep neural network (DNN) simulating the reactor dynamics and speeding up the training process. We tested our approach on a simulated SMR under different scenarios, such as normal operation and fault occurrence. The results showed that our approach can perform well under different conditions and avoid local optima. The verification results showed that the operation limits are not violated in the transient process. Our approach solves a challenging problem and provides a valuable insight for the field of SMRs. It can also be applied to other complex systems that require multi-objective optimization and decision-making.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call