The correlation between minimum temperature, maximum temperature, and other meteorological elements on that day and the previous day are analyzed by using the meteorological observation data inside solar greenhouse in winter of 2012 and 2013. The forecast model is established by using the principal component regression based on the significantly related elements. The results show: (1) There was significant correlation between minimum temperature and 8 meteorological elements inside and outside the greenhouse on that day and previous day, while maximum temperature and 9 meteorological elements had significant correlation, and similar correlation also existed between these elements.（2）The principal component regression extracted 3 principal component factors influencing minimum and maximum temperature inside solar greenhouse, and the forecast model of temperature passed significant testing. （3）The results of fitting test show that the average absolute errors of forecast minimum temperature in different conditions are 1 ℃ or so, while the average absolute errors of forecast maximum temperature in different conditions are 1.5 ℃ or so. The results of application test show that the average absolute errors are 1.1 ℃ and 1.5 ℃, respectively, and especially the test results of minimum temperature are better in sunny conditions, with the absolute error being 0.9 ℃; the test result of maximum temperature is better in sunless day, with the absolute errors being 1.4 ℃.