Focusing on the self-heating roast potato and beef dishes, this paper studied the effect of temperature on the storage properties by using accelerated shelf life method, the chemical dynamics model, Q10 model and BP neural network model is established to determine the shelf life prediction model based on the physical and chemical indicators. Under the conditions of 40 ℃, 50 ℃ and 60 ℃, the pH of food basically decreased first, then increased and finally decreased again, and the pH of food changed more dramatically with the increase of temperature. Under the three temperatures storage conditions, there was a gradual decrease and then a sharp decrease in Lvalue, a rapid increase and then a gradual decrease in the rate of a value, and b value reflected a relatively stable trend of slow rise. The reaction rate of lipid oxidation and proteolysis is accelerated, which leads to the increase of malondialdehyde and volatile basic nitrogen content. Compared with the chemical dynamics model and Q10 model, the BP neural network model based on prior conditions is most suitable for the shelf life prediction of self-heating food. The final error converges to 7.16×10−5, and the shelf life of potato and beef package at 25 ℃ is 430 days by using this model. Temperature has a significant effect on the quality indexes of roast potato and beef dishes, and the optimal shelf life prediction model is BP neural network model considering prior conditions.