Abstract: The paper aims to forecast the lowest temperature in the sunlight greenhouse effectively, reduce impact of chilling damage on facility agriculture production. Based on the data of sunlight greenhouse microclimate and meteorological observation outside the greenhouse, 8 meteorological factors which affected the lowest temperature in the sunlight greenhouse were selected and diagnosed through correlation and statistic test. Then a model for forecasting the lowest temperature of sunlight greenhouse was established by using of principal component regression method. The accuracy of model was verified by comparing model simulated values and actual values of the lowest temperature inside greenhouse. The results showed that collinearity existed among X₁, X₂, X₄, X₆, X₇ and X₈. The former 3 principal components could stand for 8 variables, and the regression equation passed the significance test (α=0.01). The regression coefficients properties of principal component regression equation were consistent with the results of correlation analysis, which made the unreasonable symbols of regression coefficients in the least square estimation reasonable. Model precision analysis showed that the R² was 0.81-0.89 and the RMSE was 0.90-1.16℃ under different weather conditions; the R² was 0.82-0.89 and the RMSE was 0.94-1.13℃ at different times.