Based on low temperature storage and control of concentrations of oxygen and carbon dioxide, technology of controlled atmosphere storage affects the maturity and withering progress of fruit and vegetable, so as to improve the fresh-keeping effect of controlled atmosphere storage. Gray system prediction theory is used to establish a prediction model of controlled atmosphere storage environmental parameters (including temperature, humidity, CO₂, O₂, etc.) by analyzing the variation rules of them. For the parameters of large error and not complying with the model prediction accuracy required, the residuals of which is regarded as raw data to be processed by the symbol, to establish the GM (1,1) model error compensation, and to effectively reduce the prediction error rate. The experimental results show that the model has a higher degree of prediction accuracy, the accurate and timely forecasts of which can be used to adjust the controlled atmosphere storage environmental parameters, sequently to improve fresh-keeping effect.