guest ::
| Home > Publications database > Modelling of Energy Systems with Seasonal Storage and System State dependent Boundary Conditions using Time Series Aggregation and Segmentation |
| Home > Publications database > Modelling of Energy Systems with Seasonal Storage and System State dependent Boundary Conditions using Time Series Aggregation and Segmentation |
| Contribution to a conference proceedings | FZJ-2022-00983 |
; ;
2021
Please use a persistent id in citations: http://hdl.handle.net/2128/30524
Abstract: The optimization of planning is one of the challenging tasks for the optimal control of energy systems with seasonal storage. An optimization quickly becomes computationally intractable due to a high temporal resolution and a long time horizonneeded for seasonal energy storage. Time series aggregation, in combination with additional time coupling constraints, can be used to reduce the size of the optimization problem drastically. However, some constraints of an energy system aredirectly dependent on the current system state and cannot be modeled as part of a typical period. To preserve the computational advantages of time series aggregation with extra constraints for storage units while modeling a set of constraints with a full temporal resolution, we propose a method that uses a mapping between intra-period, inter-period, and full-resolution variables. Furthermore, we propose a separation of the year into different regions during the clustering. This leads to a decoupling of different regions of the year and therefore increases the flexibility of the optimization.In a case study, we adopt the approach for an energy system with a dynamic hydrogen pipeline and a liquid organic hydrogen carrier (LOHC) storage system with a hot pressure swing reactor. By using full-resolution variables and a separationof the year in 3 different regions, we were able to reduce the computational time by 78% while maintaining an accuracy of 3% compared to an optimization with the full-time resolution.The separation of the year into 3 regions lead to a consistent improvement in accuracy of up to 29.4% and a run time decrease of up to 82% compared to a clustering of the whole year in typical periods. Furthermore, a separation of the yearinto 3 regions extended the feasibility of the optimization problem to very low numbers of typical periods.
|
The record appears in these collections: |
PDF
Fulltext by OpenAccess repository