The work aims to conduct a comparative study on the errors of different forecast methods, so as to select the method with higher accuracy and promote the scientific decision-making of relevant departments. Fifteen indicators were selected from the four dimensions of agricultural supply, socio-economic level, cold chain logistics security, size of the population and consumption capacity to construct the indicator system of influencing factors, and a gray correlation analysis was carried out between each influencing factor and cold chain logistics demand. The GM(1, 1) prediction model, GM(1, 6) prediction model and principal component-multiple regression linear prediction model were used to forecast cold chain logistics demand. The prediction errors of the GM(1, 1) prediction model, GM(1, 6) prediction model and principal component-multiple regression linear prediction model were 2.97%, 1.70% and 2.53%. The GM(1, 6) prediction model has high prediction accuracy, which is suitable for short and medium term cold chain logistics demand forecast and has high application value.