Research on Scene Generation Method of Wind and Solar Active Power Output Based on k-Medoids Clustering and Generative Adversarial Networks

In recent years, wind power and photovoltaic power generation have developed rapidly, and the installed proportion of wind power and photovoltaic power will further increase in the future. Aiming at the strong uncertainty of wind power and photovoltaic power, a scene generation method of wind and solar active power output based on k-medoids clustering and generative adversarial networks is proposed to avoid problems of model convergence and gradient disappearance, meanwhile preserving the diversity of generated results. Using the active power output data of two wind power plants and six photovoltaic power plants in Western China for 365 days a year, the data samples are clustered into ten scenarios by k-mediods method, which are used as training sets under different typical scenarios. The distribution characteristics of active power output data are effectively fitted by the generative adversarial networks, and its generation effect is analyzed. The results show that the scene generation method of wind and solar active power output based on k-medoids clustering and generative adversarial networks has the advantages of fast convergence and high precision.