Neural network-based PID gain tuning of chemical plant controller

Plant control systems are now highly automated and are used in many industries. The control performance changes with the passage of time because of deterioration of plant facilities. For this reason, human experts tune the control system to improve overall plant performance. In this study, a PID control system for the oil refining chemical plant process is discussed. In oil refining, thousands of control loops are used in plants in order to keep the product quality at the desired value and to assure the safety of plant operation. Due to the ambiguity of the interference between control loops, it is difficult to estimate the plant dynamical model accurately. Using a neuro emulator and a recurrent neural networks model (RNN model) for emulation and tuning of parameters, a PID gain tuning system for a chemical plant controller is constructed. Numerical experiments using actual plant data demonstrate the effect of the proposed method.