Abstract
A common solution to mitigate the complexity of power system studies is time aggregation. This is to replace the actual data set for all time intervals with representative time periods. Previous research confirms that when energy storage systems are involved in the study, preserving the overall shape of the original data is crucial. This paper proposes a new time aggregation framework to incorporate a shape-based distance to jointly extract representative periods of wind and demand data. The duration curve of the net demand is used as a data-based validation index to compare the performance of the proposed method against other techniques. Also, a 3-bus case study that includes a wind resource, an energy storage system, and two conventional generators is designed. Four model-based validation indices are defined and applied for performance measurement, including the annual operation cost of the system, the annual wind curtailment in the system, the energy throughput of the storage facility, and the daily average of the state of the charge of the energy storage for each 365 days of the year.
Highlights
T HE complexities of practical power system planning problems are often rooted in either a large network size, the presence of multiple uncertainty sources in the operation layer, a long planning horizon, or a combination of these three
X and z share a similar shape, it may not be meaningful to let x and z be grouped together if they are far from each other according to their Euclidean distance
The proposed framework mitigates the chance of grouping sequences that are not close to each other according to their Euclidean distance by identifying the local neighbors through a consensus clustering process
Summary
T HE complexities of practical power system planning problems are often rooted in either a large network size, the presence of multiple uncertainty sources in the operation layer, a long planning horizon, or a combination of these three. Since the clustering is being conducted directly on the output space, there is no doubt that the aggregated output is closer to the actual value of that particular output Such approach cannot guarantee to provide a meaningful representation on the behavior of different parts of a network and different outcomes such as wind curtailment or the behavior of the energy storage where the shape of patterns matters [5]. This paper proposes a novel shape-based clustering framework for time aggregation that aims to preserve the shape of both wind power and system demand data, and preserves their joint behavior. The main contribution of this paper is to propose a new time aggregation framework that (i) uses dependent DTW distance to group days with similar joint behavior; and, (ii) incorporates consensus clustering and mutual information to mitigates the chance of misclustering; and enhances the quality of clusters.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
