Design optimum PI controller by reinforcement learning for tank level in industrial process

Currently, the PID controller is widely used in many industrial applications, mainly in the oil and gas fields. The efficiency of any process system is essentially influenced by its control and its capacity to endure any adverse change. The control system’s capability is proportional to the values of its parameters and how much it is optimized. This fact led to the introduction of ways for optimizing real-world control systems for separator drums situated in the Basrah refinery and based on the PI controller. Two approaches have been employed: first, a Closed-Loop PID Autotuner; and second, Reinforcement Learning (RL). The mathematical model for the proposed system is developed and then simulated using MATLAB in the real-world conditions of the Basrah refinery, further adding some frequent disturbances to assess the proposed system’s performance. The findings indicate that when a Closed-Loop PID Autotuner is used, the performance is excellent. Whereas the RL technique is better than PI manual tuning and Closed-Loop PID Autotuner.