Integrated thermal-fluid and statistical modeling for fault prediction in fixed-bed dry bottom coal gasifiers

Abstract

In this study, an integrated thermal-fluid and statistical modeling approach is developed for fixed-bed dry bottom coal gasifier fault prediction. This research is aimed to address the problem of current gasifier fault prediction models relying on data driven statistical techniques that do not incorporate thermal-fluid modeling. Both gasifier thermal-fluid and statistical modeling approaches are established. However, in the reviewed literature, there is no evidence of integrated thermal-fluid and statistical modeling approaches for gasifier fault prediction. A steady-state thermal-fluid model of the fixed-bed dry bottom coal gasifier is developed in this study. This model predicts mass flow rates for the five gases that mainly comprise syngas. As such, five fundamental thermodynamic equations are selected and applied in the model. These are three mole balances, an energy balance, and a chemical reaction equilibrium condition. Software for the Engineering Equation Solver (EES) is utilized to develop the thermal-fluid model. It is implemented so that abnormal gasifier operating conditions, where negative flow rates are predicted, can be isolated from typical data. When compared with a reference gasifier observation, the average accuracy of this model’s predictions is comparable with that of a fixed-bed gasifier model available in literature. Statistical modeling is undertaken using the R Statistical Software (R). To predict where the gasifier’s operation transitions from normal to abnormal conditions, statistical classification models are applied to the thermal-fluid model’s outputs. A classification model is selected that predicts a bounded region wherein the gasifier’s operations are normal. The existence of gasifier faults is discussed with respect to groups of predicted abnormal observations. Observations are presented on a plot of the selected classification model, and groups are identified, with a statistical clustering technique, according to the gasifier’s inlet flow rates. The thermal-fluid model’s validity is established through an analysis of the changes evident between these groups. When compared with conventional data driven fault prediction for fixed-bed dry bottom coal gasifiers, the selected classification model possesses a higher accuracy.