The invention relates to a coal spontaneous combustion danger degree multi-source information fusion early warning method, and solves the problems that an existing coal spontaneous combustion early warning method is low in early warning precision, and scientific decision is hard to make to obtain the spontaneous combustion position, ignition time and hidden danger degree of coal spontaneous combustion. The method comprises the following steps: step 1, arranging a sensor; : step 2, registering the sampling time of the sensor; : step 3, acquiring the acquisition value z (k) of each sensor at different moments; : step 4, performing noise reduction processing on the acquisition value z (k) of each sensor in the step 3; : step 5, taking historical actual detection data of the sensor at the early warning point as coal spontaneous combustion characteristics, and performing normalization processing; : step 6, determining the hidden danger degree and the ignition time of coal spontaneous combustion; and : step 7, through the coal spontaneous combustion target early warning point obtained in the step 5 and the coal spontaneous combustion hidden danger degree and ignition time obtained in thestep 6, early warning of the coal spontaneous combustion danger degree is completed.