INTEGRATION OF GENETIC ALGORITHM BASED PID CONTROLLER WITH DATA-DRIVEN BASED INFERENTIAL MEASUREMENT AS AN ALTERNATIVE CONTROL OF DESULFURIZATION PROCESS IN AMMONIA PLANT

• Main Author: Taufiq Heryuano, Baninda
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/56829
• Institution: Institut Teknologi Bandung

Desulphurization is one of the important processes for the production of ammonia (NH3). The main purpose of this unit is to reduce the sulfur content in natural gas. These values are obtained by applying offline measurements in the laboratory, making the process barely directly controllable. To predict sulfur content, the development of data-driven based inferential measurements is introduced in this study in order to control the desulfurization process. Inferential measurement consisting of a Diagonal Recurrent Neural Network (DRNN) to model the relationship between immeasurable variable (sulfur content) and measurable variables (existing variables) as secondary variables, with Bayesian regularization as the data training algorithm. This scheme is then integrated with optimized PID control using the Genetic Algorithm (GA) tuning technique. The objective is to have an optimal responsse of desulfurization process by controlling it. The potential of this technique is implemented in a non-linear process model using real operational data obtained from an ammonia plant located in East Kalimantan. After a comprehensive simulation studies, the proposed strategy has been able to predict the sulfur content by approximating the results obtained from laboratory measurements with an RMSE value of 0.00284. The GA tuning control scheme has been able to control the sulfur content according to the desired set point, with an IAE value of 0.3706 where this technique has better performance than manual tuning of 0.4238.