CONTROL OF GASHOLDER LEVEL BY TREND PREDICTION BASED ON TIME-SERIES ANALYSIS AND PROCESS HEURISTICS

A novel method to control gasholder levels in an iron and steel company with accurate prediction of future trend is presented. Although various gasholders are used to recycle by-product gases generated during iron-making, coke-burning and steel-making process, the capacity of the gasholders are insufficient to handle large amount of the gases. To overcome this problem, tight control of the gasholder level should be conducted by predicting their anticipated changes. However, the current prediction logic cannot show satisfactory results due to the lack of characterization of relevant processes. In the proposed method, time-series modeling and heuristics of industrial operators are used to correctly reflect the process characteristics and deal with unexpected process delays. By applying the proposed method to an off-line data set, a significant reduction of discrepancy between predicted values and actual values has been observed. The method is expected to be adopted in the prediction system of POSCO. Copyright © 2002 IFAC