| 초록 |
In recent years, hybrid neural network approaches, which combine mechanistic models and neural networks, have received considerable interest. These approaches are potentially very efficient for obtaining more accurate predictions of process dynamics by combining mechanistic and neural models in such a way that the neural network model properly accounts for unknown and nonlinear parts of the mechanistic model. In this work, such an approach was applied in the modeling of a full-scale coke wastewater treatment process. Initially, process data analysis was performed upon actual operational data using Principal Component Analysis. Secondly, a simplified mechanistic model and neural network model were developed based on specific process knowledge and operational data of the coke wastewater treatment process, respectively. Finally, a neural network was incorporated into the mechanistic model in both parallel and serial configurations. Simulation results showed that the parallel hybrid modeling approach achievedmuch more accurate predictions with good extrapolation properties than the other modeling approaches even in the case of a process upset caused, for example, by the shock loading of a toxic compound. These results indicate that the parallel hybrid neural modeling approach is a useful tool for the accurate and cost-effective modeling of biochemical processes, in the absence of a reasonably accurate process model. |
