An Improved DEC Algorithm for Time Series Clustering Task

In recent years, global climate issues have escalated, driving a substantial increase in the total installed capacity of new energy sources worldwide over the past five years in order to replace fossil fuels and construct a cleaner electricity system. Consequently, the inherent fluctuations and intermittency of new energy sources have increasingly manifested negative impacts on the stable operation of power systems. Against this backdrop, energy storage technologies have garnered widespread attention for their ability to effectively mitigate the uncertainty of new energy output. However, the randomness of new energy output significantly increases the scale of power grid operation scenarios. It is not efficient to design storage solutions tailored for each day, and manually generating the most representative typical scenarios from long-term data is challenging. To address these issues, this paper proposes an improved DEC algorithm for time series clustering tasks. Through validation using actual regional data, the clustering performance of this algorithm surpasses that of the K-means algorithm, the original DEC algorithm, and the PCA+K-means algorithm, enabling the generation of more representative typical scenarios. This holds significant importance in reducing the complexity of optimizing storage configuration.