Abstract: Based on temperature data monitored inside the greenhouse and the corresponding surface meteorological observation data, authors established a forecasting model for the highest and lowest temperature in the sunlight greenhouse in autumn, as well as in the sunny and overcast sky separately in winter and spring, by means of stepwise regression method. The models were also tested in the application. The results showed that the absolute error (ABSE) of the highest and lowest temperatures between the simulated values and the actual values were separately 1.1 and 0.2℃ in autumn,0.8 and 0.4℃in the sunny day in winter,1.5 and 0.3℃ in the overcast sky in winter,0.3 and 0.4℃ in the sunny day in spring,1.1 and 0.2℃ in the overcast sky in spring. The root mean square error(RMSE) were separately 1.3 and 0.2℃ in autumn,1.0 and 0.5℃ in the sunny day in winter,1.7 and 0.3℃ in the overcast sky in winter,0.3 and 1.3℃ in the sunny day in spring,0.4 and 0.5℃ in the overcast sky in spring. The ABSE of the highest and lowest temperatures between the simulated values and the actual values in autumn were separately 0.8-1.1℃ and 0.3-0.4℃ and the root mean square error were 0.9-1.2℃ and 0.3-0.5℃ in the examination. This forecast model which can be used to predict the highest and lowest temperatures in future 24 hours inside the sunlight greenhouse, can provide effective decision support for timely ventilation in greenhouse to prevent high temperature as well as for heat insulation measures to prevent cold harm.

Key words: Sunlight greenhouse, The highest temperature, The lowest temperature, Forecasting model