A Clustering Method for Distribution Network Load Curve Based on Fast DDTW

Utilizing DDTW distance in load curve clustering offers greater expressiveness and adaptability than the conventional Euclidean distance-based approach, particularly when handling curve data with varying lengths and shapes. This paper suggests a load clustering approach based on the combination of the K-medoids and Fast DDTW clustering methods because the DDTW distance computation is too complicated. The user load curve's distance is computed using the Fast DDTW technique after the load data has been denoised and smoothed. Lastly, a sample analysis is performed on the daily load curve of local electric power plant consumers. The findings demonstrate that the fast DDTW algorithm increases the algorithm's resistance to time axis deformation when compared to the DTW algorithm and the DDTW algorithm. Simultaneously, the fast DDTW method significantly reduces calculation times while preserving the benefits of the DTW algorithm.