The invention relates to a method for identifying an adhering state of a heavy load vehicle based on an extreme learning machine. The method can be applied to an adhering control system of the heavy load vehicle. An adhering state feature is extracted from the measured data and a feature vector is selected from the adhering state feature so as to perform the training of an adhering state identifying model. The method provided by the invention requires the sample having certain representativeness; after the number of the hidden nerve cells is set, the identification for various adhering states can be directly realized; the identifying model can acquire the global optimal solution without repeated iteration, so that the identifying speed is ultrahigh and the accuracy is high.