Aiming at the problem that the interaction mechanism between large agricultural machinery human factors and black soil is unclear, therefore the field test of 60 mu maize sowing operation was carried out in the cooperative of Baiquan County in Northeast China. Based on the measured data, the driver-tractor-soil model was established, analyzed and used for predicting. Firstly, the driver’s fatigue, vehicle speed, soil compaction were collected through field experiments. Secondly, the Levenberg-Marquardt neural network model of driver fatigue–tractor speed–soil system was established, and the change trend of driver fatigue on soil compaction was predicted by orthogonal test. The threshold of driver human fatigue for black soil protection was obtained: the fatigue of lumber, neck, arm and leg was 2, 1, 2 and 3. Finally, through the driver’s ideal fatigue, it is predicted that the change rate of soil compaction at 10 cm and 20 cm depth of unmanned tractor under cooperative production mode is 57.38% and 38.39%, respectively, which provides a certain design basis for guiding the intelligent human factor system under the background of tractor conservation tillage.