Correlation-Aware Scenario Reduction for Wind-Power and Load Uncertainties via Density Denoising and Temporally Ordered Clustering

The increasing penetration of wind-power introduces pronounced uncertainty and variability into modern renewable energy systems, while electric load exhibits strong temporal continuity and seasonal evolution. Directly using full-year chronological wind–load data in planning or operational studies may lead to excessive computational burden, whereas conventional scenario reduction methods may distort load temporal patterns, weaken wind–load dependence, or be affected by isolated low-density samples. To address these issues, this paper proposes a correlation-aware scenario reduction method for wind-power and load uncertainties by integrating density-based denoising and temporally ordered clustering. First, the original wind-power and load profiles are normalized and processed using density-based spatial clustering of applications with noise (DBSCAN), so that isolated low-density samples can be identified before scenario extraction. Then, a temporally ordered clustering model is developed for load profiles, in which adjacent chronological samples are grouped into continuous load segments to preserve the intrinsic temporal evolution of demand. Based on the obtained load segments, conditional K-means clustering is further performed on the corresponding wind-power profiles, thereby capturing the wind-power distribution characteristics under different load states. Finally, representative joint wind–load scenarios are reconstructed with probability weights. Case studies based on annual wind-power and load data demonstrate that the proposed method reduces the original annual data into ten probability-weighted joint wind–load scenarios constructed from seven temporally ordered load states, while retaining the main statistical characteristics, temporal consistency, and multi-level wind–load dependence evaluated at the aggregate, intra-day, nonlinear, and short-lag levels. Compared with conventional clustering-based scenario reduction methods, the proposed approach provides more representative typical scenarios for uncertainty modeling in renewable energy system studies.