The invention relates to the technical field of power grid dispatching, in particular to a method for predicting 220kV main transformer short-term load curve of a power grid. The method comprises thesteps of obtaining all 110kV main transformer and 10kV line load prediction data; traversing the 220kV main transformer and topologically forming the load composition of the 220kV main transformer; predicting load prediction data of the low part of the 220kV main transformer; calculating the power supply proportion of each mounted 110kV main transformer; predicting partial load prediction data inthe 220kV main transformer; correcting partial load prediction data in the 220kV main transformer according to the power failure maintenance list; correcting load prediction data of the low part of the 220kV main transformer according to the power failure maintenance list; and generating a prediction curve of the 220kV main transformer in the future 48 hours. Data of the method come from an automatic E file, an oriental SCADA Web database and an auxiliary decision-making system. The method can be written into a program module to be placed into a computer system, main transformer load composition is obtained through rapid and automatic analysis, then historical data of all load composition elements are fitted, and a main transformer short-term load curve is generated.