This study describes the development of a dynamic knowledge-based reasoning-enhanced model predictive control system (KBRECS) for in-situ bioremediation processes. The automated control system balances the complex physical, chemical, and biological processes involved in the remediation process while minimizing overall cost of the entire remediation process. The control system includes an optimization subsystem and a monitoring subsystem. The optimization subsystem consists of a simulation model supported by an optimization function which is designed to generate a series of optimal control actions. The monitoring subsystem is a knowledge-based system which is designed to monitor and adjust the online control actions. The numerical simulation model describes the fate and transport of the subsurface contaminants. The optimization function is a constrained, nonlinear function that has been implemented using a genetic algorithm (GA). Intermediate genetic algorithm individuals are indexed and stored in the knowledge base, thereby reducing search times for values to replace the unqualified schemes used by the monitoring subsystem. The system was applied to a lab experiment and compared with the control system presented in [9]. The results indicated that the knowledge based reasoning system enhanced the control system by generating an appropriate control strategy and adjusting control actions promptly. This helps to enhance efficiency in control of the in-situ bioremediation process at petroleum-contaminated groundwater systems.
Integrating multiple descriptions in simulation model design: a knowledge based approach
