The invention discloses a temperature prediction method for an earthen archaeological site based on the characteristics of solar terms, and mainly solves the problems that an earthen archaeological site temperature prediction result obtained in the prior art is bigger in relative error and is low in prediction precision. The method comprises the implementation steps: 1, extracting the time characteristics of temperature data of the earthen archaeological site; 2, extracting the characteristics of solar terms in the temperature data of the earthen archaeological site; 3, obtaining the general characteristics of the temperature data through the time characteristics and the characteristics of solar terms; 4, introducing a neural network, and carrying out the learning and training through thegeneral characteristics of the temperature data and an earthen archaeological site temperature data sample to obtain a neural network model for the temperature prediction of the earthen archaeologicalsite; 5, predicting the temperature of the earthen archaeological site through the obtained neural network model, and obtaining a temperature prediction result matrix of the earthen archaeological site. The method achieves the prediction of the temperature of the earthen archaeological site on the basis of yearly, daily and hour, minute and second characteristics through a membership function andthe characteristics of the solar terms, thereby reducing the relative errors caused during the temperature prediction, and improving the prediction precision.