Fast DTW and Fuzzy Clustering for Scenario Generation in Power System Planning Problems

TL;DR

This paper introduces a novel approach combining fast dynamic time warping and fuzzy clustering to generate representative scenarios for power system planning, effectively capturing dependencies between renewable generation and load.

Contribution

It proposes a new method using FDTW and FCM++ clustering to improve scenario generation by preserving key statistical dependencies in power system planning.

Findings

Enhanced scenario representation of load and renewable generation dependencies

Improved computational efficiency over traditional methods

Validated approach on a U.S. power network case study

Abstract

Power system planning problems become computationally intractable if one accounts for all uncertain operating scenarios. Consequently, one selects a subset of scenarios that are representative of likely/extreme operating conditions, e.g. heavy summer, heavy winter, light summer, and so on. However, such an approach may not be able to accurately capture the dependencies that exist between renewable generation (RG) and system load in RG-rich power systems. This paper proposes the use of fast dynamic time warping (FDTW) and fuzzy c-means++ (FCM++) clustering to account for key statistical properties of load and RG for scenario generation for power system planning problems. Case studies using a U.S. power network, and comparison with existing scenario generation techniques demonstrate the benefits of the proposed approach.