Abstract
The increasing complexity of capacity expansion models underscores the need for time-series aggregation methods that balance accuracy with computational efficiency. While previous studies have focused on developing methods for small-scale power sector capacity expansion models, this work compares three traditional clustering approaches with a novel dynamic time warping based methodology using a single-node and a multi-node sector-coupled capacity expansion model with seasonal storage. All methods are evaluated in terms of statistical accuracy, objective function approximation, and operational performance across three weather years. The proposed methodology achieves the highest statistical accuracy and best approximates the full-hourly objective function in the single-node case. In the multi-node model, where a full-hourly benchmark is computationally infeasible, proxy-based operational metrics indicate that the proposed methodology most closely approximates the full-hourly solution among the evaluated methods. Overall, the findings suggest that the proposed method performs particularly well in complex systems and when a low number of representative periods is used.