Reinforcement Learning-Based Adaptive Control of a Pool-Type Heating Nuclear Reactor

In recent years, people have conducted research on the application of artificial intelligence technology in nuclear reactors. For low-temperature heating nuclear reactors, adaptive control of the outlet temperature of the heating circuit is very important. This article proposes a deep reinforcement learning based adaptive control method for swimming pool heating reactors, which serves as the basic research for autonomous control of heating reactors. Firstly, a physical model of the heating reactor during operation was constructed, which was then transformed into a reinforcement learning model. The TD3 algorithm was used to solve the control problem. The experimental results show that deep reinforcement learning has the potential to realize automatic control operation, and the system control effect for multi-disturbance and multi-coupling output is better than that of traditional PID control. The algorithm proposed in this paper shows very significant control performance in the application of nuclear reactors.