The invention provides a deep learning method for different fuel characteristics. The deep learning method comprises a temperature automatic adjusting process, wherein the air speed of an air supply motor and the air speed of an air suction motor are adjusted according to the temperature deference delta T1 between the set temperature T0 and the room temperature T1 and the temperature deference delta T2 between the set temperature T0 and the temperature T2 of an air outlet; and a dust blowing automatic adjusting process, wherein the set time t1 and the set time t2 are adjusted according to comparison of the weight M of dust blown off in the dust blowing process and the set threshold value M0. According to the deep learning method for the different fuel characteristics, the air speed of the air supply motor, the air speed of the air suction motor and the dust blowing time interval can be automatically adjusted according to temperature changes, and adaptability can be generated on fuel; and meanwhile, relevant fuel data are stored to a storer to be used as the basis of using raw materials next time, and intelligent growth of the raw materials is achieved.