Heat exchanger control: Using deep-Q network

Abstract

To achieve stable operation of district heating systems and improve the performance of heat exchange station control systems, this paper explores the application of intelligent control algorithms for controlling district heating systems. Initially, a heat exchanger neural network model is established. Secondly, a mass regulation controller based on a deep reinforcement learning algorithm is designed to accurately track the desired outlet temperature of the secondary side set point. Finally, the performance of the controller in terms of tracking response and disturbance suppression is verified using an example of a district heating system. The experimental results demonstrate that the controller exhibits minimal overshoot, with the outlet temperature of the secondary side deviating by less than 1 % when the controller setting value changes. Additionally, during disturbance suppression response, the deviation of the outlet temperature of the secondary side remains within ±2 °C. Compared to other controllers, the designed controller exhibits superior dynamic performance across various operating points and under different disturbances. Additionally, it outperforms traditional control methods in terms of stability and disturbance rejection capabilities.