Abstract: Based on the meteorological observation data inside and outside solar greenhouse from 2013 to 2014 in Shenyang,variation characteristics and forecasting models of the minimum temperature (T_min) in greenhouse were studied using methods of a correlation analysis and a stepwise regression analysis.The results show that T_min inside the greenhouse is significantly correlated with the previous day's maximum and minimum temperature outside the greenhouse,T_min outside the greenhouse of the day,and the previous day's maximum and minimum temperature inside the greenhouse.T_min forecasting models for the greenhouse are different depending on different weather types of the four seasons.Forecasting accuracy of the greenhouse T_min is the highest in winter,followed in spring and autumn,and the forecasting accuracy in summer is relatively worse.The differences between observed and simulated greenhouse T_min less than 3℃ in winter,spring,autumn,and summer account for 91%,85%,81% and 79%,respectively of the total samples for test.Forecasting accuracy of greenhouse T_min is the highest in snowy rain days,followed in overcast days and sunny days,and the forecasting accuracy in cloudy days is relatively worse.The model accuracies for the simulated greenhouse T_min less than 3℃ in snowy rain days,overcast days,sunny days,cloudy days are 90%,87%,83% and 77%,respectively.The results suggest the performance of the greenhouse T_min forecasting model is satisfied and the models have their practical applicability,which provides a reference for the greenhouse T_min forecast in Shenyang.

Key words: Solar greenhouse, The minimum temperature, Weather types, Correlation analysis, Stepwise regression analysis