MODELING AND PREDICTING CATALYTIC NAPHTHA REFORMINGPROCESS VARIABLES USING GMDH NETWORK

In this study, the group method of data handling (GMDH) networks is applied for estimating themomentous process variables of a commercial scale heavy naphtha catalytic reforming unit (CRU). Theproposed model can predict the research octane number (RON) and yield of gasoline by using a grandpolynomial correlation which is a function of days on stream (DOS), liquid hourly space velocity (LHSV),H2 to hydrocarbon ratio (H2/HC), inlet temperature of reactors and weight average bed temperature (WABT).To do such a task, ninety eight data are obtained from the target plant during a life cycle (about 877 days).Then, the GMDH network uses 70% of these data points for self-training whilst using the remained ones forthe validation step. The results showed that this model can precisely estimate the gasoline product propertiesduring the cycle life. Moreover, it is confirmed that the proposed model is capable of predicting RON andyield of gasoline with the average absolute deviation (AAD%) of 0.406% and 0.655%, respectively.Moreover, the root means square error (RMSE %) of the mentioned parameters are 0.507% and 0.888%,respectively.

Keywords:

Modeling , GMDH , Neural network , Catalytic reforming , RON , Yield