Demand response scheduling using derivative-based dynamic surrogate models. (April 2022)

• Record Type: Journal Article
• Authors: Di Pretoro, Alessandro; Bruns, Bastian; Negny, Stéphane; Grünewald, Marcus; Riese, Julia
• Is Part Of: Computers & chemical engineering. Volume 160(2022)
• Publication Date: 2022-04
• Subjects: Data-driven modeling -- Demand-side management -- Setpoint tracking -- Polynomial regression -- Response surface methodology -- Plantwide modeling; Chemical engineering -- Data processing -- Periodicals; 660.0285
• DOI: 10.1016/j.compchemeng.2022.107711
• Languages: English
• ISSNs: 0098-1354

Abstract

Highlights: The dynamic response of complex process plants can be synthesized by a single function. The derivative-based approach provides a dynamic model of general validity. Optimal scheduling problem computational time can be reduced by orders of magnitude. Quantitative assessment of downstream dynamics impact on demand response scheduling. Computational effective surrogate models meet the need of DR for daily update. Abstract: When assessing demand response to solve optimal scheduling problems, the optimization algorithm needs to be coupled with the process model in order to quantify the behavior of the monitored variable. For negligible transients, a first approximation consists of applying the steady state correlation between input and output variables. On the contrary, when dynamics show a relevant bias with respect to the estimated steady state response, a more accurate model is required. When coupling dynamic models of entire chemical processes with the optimization algorithm, the computational effort drastically increases. In those cases, the model should be simplified without losing its accuracy. In this research work, we propose a derivative-based approach for dynamic surrogate modeling applied to an ethylene oxide production scheduling problem. Thanks to its general validity, once derived, it can be applied to any setpoint trajectory. This approach allows to reduce the computational time of the optimization algorithm by orders of magnitude, providing results with the same accuracy as the detailed process simulation.