In order to fully consider the combined effects of temperature and humidity variables on summer power load, an improved summer short-term load forecasting method based on temperature and humidity multi-form variables is proposed. Firstly, by analyzing the influence of summer meteorological factors on load changes, three different forms of temperature and humidity variables are constructed as model input variables. Then, the load is layered according to the change of the weekly characteristics, the prediction model based on LASSO regression is established for each layer load, and the input variables are selected by the enumeration search algorithm to optimize the prediction model. Finally, by calculating the coefficients corresponding to the residual variables, the load distribution in each period is further estimated. The example results show that the proposed method can effectively improve the prediction accuracy and robustness of the model. This work is supported by National Natural Science Foundation of China (No. 61672032) and Anhui Science and Technology Major Project (No. 18030901018).