With advancements in new energy generation and electric vehicle technologies, uncertainties in power systems have significantly increased. Accurately identifying key input variables affecting system output helps enhance operation control and reduce the scale of solving stochastic problems. Traditional local sensitivity analysis, based on partial derivatives of output to input, fails to account for system nonlinearity, all input values, and variable correlations. A Global Sensitivity Analysis (GSA) method for new energy power systems' load margin has been proposed to identify key uncertainty factors. First, a probabilistic model for load margin was established. Second, GSA theory for independent input variables was introduced. Then, using conditional variance theory and Nataf transformation, a GSA method considering correlations was developed to quantify the impact of wind and solar power fluctuations. Finally, simulations calculated first-order and total sensitivity indices for wind and solar power, identifying key inputs and comparing source and load impacts under strong and weak correlations, validating the method's effectiveness.