Load forecasting is an important basis for guiding planning, safe and economic operations of power system. The traditional load forecasting generally refers to the prediction of the total regional load, which cannot reflect the underlying bus load level and cannot meet the requirements of grid lean management. Therefore, bus load forecasting is the key way to solve this problem. However, the number of bus in the system is large, the load base is small, the characteristics are different, and the volatility is strong, which brings difficulties to the bus load forecasting work. In this paper, the bus load forecasting model is studied. According to the actual grid situation, the concept and forecasting idea of load distribution factor are proposed. Considering the validity of historical data and selecting similar days by using daily feature quantity and trend similarity, a variable weight combination forecasting method based on information entropy is proposed to improve the prediction load accuracy of the whole system. Combined with the system load forecasting results and the load distribution factor, the predicted results of each bus are finally obtained. The example verification is carried out by using a certain regional grid load. The results show that the prediction model established in this paper has good prediction accuracy and stability.