Parameters selection in predictive online simulation

Abstract

Industrial applications with reliable predictive features are becoming increasingly important. A tracking simulator is an example of an online simulation system with great capabilities that fills the gap left by other predictive applications. In a tracking simulator, a simulation model is run in parallel with a physical process controlled by the process' control system. At the same time, a tracking mechanism is used to keep the state of the simulation model as close as possible to the real process by continually adjusting parameters of the model. The selection of these parameters impacts directly on the quality of the tracking simulation results and it is a complex task in processes with a big number of variables. This paper presents two case studies of tracking simulation where the controlled parameters are selected using different techniques. The first case study deals with a laboratory-scale hot water generation process where the parameters' selection is performed manually. The second case study deals with a combined heat and power production process with major uncertainties in the process structure. In this case, we focus on the variance decomposition method used to determine the most suitable controlled parameters. Conclusions and future work are finally presented.