[Objective] To address the challenges of unclear wind and solar power output characterization，low accuracy of traditional clustering strategies，and poor performance in feature modeling due to the complex environment of the Sandy，Gobi，and Desert （SGD） region，this study proposes a multidimensional dynamic clustering strategy for analyzing regional wind and solar power output characteristics. The aim is to achieve precise clustering and quantitative description of wind and solar resources，providing theoretical support and data foundations for multi-source coordinated dispatch and optimization. [Methods] First，a multidimensional clustering index system is constructed，incorporating similarity，complementarity，and grid integration adaptability. Next，a clustering method based on Bidirectional Agglomerative Hierarchical Clustering （BAHC） is proposed，with Bayesian Optimization （BO） introduced to dynamically adjust index weights，enabling accurate clustering of wind and solar resources in complex environments. Finally，a "model-scenario-indicator" linkage analysis framework is established to quantitatively evaluate output characteristics and adaptability. [Results] Simulations demonstrated that the proposed multidimensional index system comprehensively characterized the wind and solar power output features in the SGD region. The BAHC-based clustering method outperformed traditional algorithms in terms of silhouette coefficient and CH index，substantially enhancing the adaptability and accuracy of clustering results. The kernel density estimation（KDE） model effectively captured the distribution patterns of wind and solar power outputs in most SGD regions，while the Weibull distribution is suitable for auxiliary description of low-output risk clusters. [Conclusions] The proposed multidimensional dynamic clustering strategy significantly enhances the accuracy and adaptability of wind and solar power output characteristic analysis in the SGD region.