Predicting Sulfur Content of Desulfurizer using Data-Driven based Inferential Measurement: An Ammonia Plant Case

Abstract

The parameters of the sulfur content in the process gas of the desulfurizer unit is an important issue for the production process of ammonia (NH3) in the ammonia plant. Due to limited conditions and equipments at the plant, quite often that the measurement of sulfur content is still carried out indirectly (off-line) by sampling and analysis methods in the laboratory, causing significant delays. In this paper, an alternative method to predict the value of sulfur concentration will be proposed using the data-driven based inferential measurement. The sulfur concentration (as primary variable) will be predicted from the available measured data (as secondary data) using neuro-fuzzy based method. In this case, a model representing the desulfurizer unit needs to be reconstructed based on input and output relationships that affect the main variables in the ammonia plant. For verifying the applicability of the proposed method, real-time operational data from a running ammonia plant located in East Kalimantan, Indonesia will be used. After an extensive simulation studies, it is show that the sulfur concentration can be predicted quite successfully, with RMSE value of the designed inferential estimator is 0.003134.