Power source configuration is an important stage of independent microgrid planning, which guarantees the economic and reliable operation of the microgrid system. The type and capacity of power source for the independent microgrid are affected by the factors such as load level, geographical location, wind, solar, water, and other natural resources. This paper analyzed the characteristics of natural resources in different regions. The index evaluation system of wind, solar, and hydropower resource was constructed, and the fuzzy evaluation method was proposed to evaluate the wind-solar-hydropower resource. Considering the power supply reliability and environmental friendliness, the optimization model of the microgrid power source was established with the objective of minimum comprehensive generation cost and the power and energy balance as constraints. Generative Adversarial Networks (GANs) were applied to simulate the uncertainty of wind, solar, and hydropower output and generate multiple operation scenarios of independent microgrids. The improved method for the K-Medoids clustering algorithm was studied to reduce the number of operational scenarios and enhance computational efficiency. Finally, an example was presented to illustrate the analysis processes and verify the feasibility and rationality of the proposed method. The wind-solar-hydropower resource levels of 30 provinces in China were obtained, and five representative regions were selected. The power source configuration scheme of an independent microgrid in these regions was proposed, which provided a reference for independent microgrid planning.