Data-Driven Cluster Method for Photovoltaic Power Stations

With the gradually increasing penetration of distributed generations of renewable energy sources, such as photovoltaic and wind power, new power systems face difficulties regulating distributed generations. One practical solution is scheduling a large-scale distributed power generation in a hierarchical and clustered mode. Therefore, based on the data-driven method, this paper proposes a clustering method for photovoltaic (PV) power stations to support the power system operation work. Firstly, based on an improved k-means clustering method and silhouette coefficient, the meteorological data collected by PV power stations are clustered to generate typical weather scenarios. Secondly, the standard output data of PV power stations are calculated using empirical equations. Thirdly, considering the similarity of output data, a state evaluation method is established to cluster PV power stations. Simulations are conducted using actual PV output data, and the results show that the proposed method can improve the accuracy of clustering results.