In order to adapt to requirements of fast response in smart grid, the accurate response of power load forecasting, as one of the key tasks in smart grid, plays a vital role in safety, stability and economy of power system operation. Firstly, the characteristics and classifications of power load forecasting are introduced. Secondly, the influencing factors of power load forecasting are analyzed and the basic steps and performance evaluation indexes of power load forecasting are introduced. Then the research of power load forecasting is divided into three categories: traditional forecasting method, machine learning forecasting method and deep learning forecasting method. Finally, the work done is summarized and the future development direction of power load forecasting is prospected.