Abstract: To forecast minimum temperature inside greenhouse,the minimum temperature forecast model was established based on meteorological observation data inside the solar greenhouse in winter of 2010 and 2011,by using of principal component regression.The characteristic of temperature,total cloud cover and maximum wind velocity was discussed through cloud covered coefficient method and standard of wind velocity conversion.The results showed that the simulation errors of cloud cover and wind velocity were reasonable.There was a good correlation between minimum temperature inside the solar greenhouse and microclimate elements in greenhouse last day.Moreover,similar correlation also existed between meteorological elements inside and outside greenhouse.The multiple correlation coefficient of the model was 0.857,and approved by significant testing.The average absolute errors of forecast minimum temperature inside greenhouse in different weather conditions were less than 1℃,the average relative errors were within 13%,and the RMSE was 1.1℃ during the whole winter by using principal component regression method.The results indicated that minimum temperature forecast model had quite precisely for predicting minimum temperature inside greenhouse,which could meet the forecast requirements for greenhouse microclimate.

Key words: Solar greenhouse, Temperature characteristic, Principal component regression, Low temperature forecast