The invention discloses a confidence embedded
model predictive control method, a medium and equipment, and belongs to the technical field of suspension control. According to the method, a kinetic equation of a suspension
system is established, a data-mechanism dual-drive
uncertainty quantification method is adopted, original operation data of a
maglev train and suspension state information measured on line are utilized, and dynamic characteristic learning modeling is conducted on the suspension
system under the constraint of the kinetic equation; state quantity distribution of suspension air gaps, vertical speed,
electromagnet current and the like at future moments is predicted, and prediction mean values and prediction standard deviations of corresponding states are obtained. In
model prediction control optimization, a nominal trajectory of the suspension
system is constructed by a predicted state mean value, and a penalty term proportional to prediction uncertainty is added into a cost function; meanwhile, confidence coefficient constraint is constructed based on the predicted standard deviation, and dynamic constraint tightening of key physical quantities such as suspension air gaps, vertical speed and coil current is achieved. According to the invention, real-time sensing and self-adaptive closed-
loop control of the operation risk of the suspension system are realized.