Abstract: Accurate day-ahead load power forecasting is a crucial data foundation to ensure the safe and stable operation of the power grid and the effective functioning of the electricity market. However, the increasing randomness of grid-connected renewable energy and customer behaviors poses significant challenges on the accuracy of day-ahead bus load power forecasting. To improve the accuracy of 220 kV bus day-ahead load forecasting, a temperature-sensitive adaptive 220 kV bus day-ahead load forecasting method considering the impact of renewable power and temperature sensitivity is proposed. First, multiple load decoupling is performed on the load power measurements based on historical renewable power measurements, where meteorological observations is also combined to predict the renewable energy output day-ahead. Then, a temperature-sensitive adaptive algorithm is proposed to predict the net load day-ahead based on temperature information. Finally, the day-ahead forecasts of renewable energy output and user net load are superimposed and reconstructed to obtain the final day-ahead load power forecasts. Based on practical 220 kV data, experimental results show that, compared with traditional machine learning models, the proposed method can effectively improve the accuracy and robustness of 220 kV bus day-ahead load power forecasting.