The uncertainty of wind power significantly affects various aspects of power systems, including capacity allocation, long-term and short-term scheduling, and electricity market trading. Through clustering methods, the multidimensional features of wind power historical data can be partitioned, transforming the uncertainty of wind power into a problem of deterministic typical scenarios under different probability conditions. By reducing the original wind power scenarios, clustering not only improves the efficiency of power system operation calculations but also enhances the accuracy of typical scenario fitting. This study utilizes an improved K-means +  + algorithm based on the elbow method and Davies-Bouldin (DB) index for generating typical wind power scenarios. The elbow curve and Davies-Bouldin index are used to search for the optimal number of clusters, and a heuristic algorithm is employed for selecting initial cluster centers. This approach addresses the difficulties faced by traditional K-means algorithms, such as the need to determine the number of clusters in advance and manually set cluster centers. Subsequently, using historical wind power data from a real region with substantial wind power capacity as the research subject, typical wind power scenarios are generated, providing output curves for these scenarios.