In order to study the demand of cold chain logistics of agricultural products of urban residents in Jiangsu Province， this paper takes the data of Jiangsu statistical yearbook from 2009 to 2017 as the sample， selects the influencing factors such as fresh vegetables， pork， fish and shrimp， fresh eggs， fresh milk， melon and fruit， gradually reduces the data range， and uses the grey GM (1，1) model and MATLAB 2017b software to predict the demand of cold chain in the next five years， The accuracy of the prediction model is 97.62%， and the prediction accuracy is 98.61% through the optimization of Markov chain. The experimental results show that the optimization of the gray Markov model can provide data reference for the corresponding departments through the prediction results， in order to promote the supply and demand balance of the cold chain of agricultural products to provide theoretical support.