Analyzing the influence of factors on a power load is important for power grid dispatchers to understand load characteristics and improve the accuracy of load forecasting. The traditional correlation analysis method cannot consider complex nonlinear effects, so a new method, which can find the nonlinear correlation, is proposed to analyze the correlation of factors and power load. First, this paper uses the learning ability of the gradient boosting decision tree to train a power load forecasting model. Then, based on the forecasting model, it proposes the importance of measuring the nonlinear correlation between factors and power load. Finally, the partial dependence of power load on the factors is used to calculate the nonlinear effects of influencing factor changes on the trend of load changes. The real power load is used for verification and compared using the Pearson correlation coefficient. The experimental results show that the method can effectively identify the important factors which affect power load, and can find the nonlinear relationship between various factors and the power load. This work is supported by Science and Technology Project of State Grid Corporation of China (No. 5700-202055368A-0-0-00).