Abstract:

The minimum temperature prediction model inside solar greenhouse in Ganzhou of Gansu Province was established based on meteorological elements by using ridge regression analysis. The multicollinearity of predictors was diagnosed through statistic test on the basis of reasonable selecting predictors, the ridge regression analysis was used to get over the influence of multicollinearity on the model stability, and the accuracy of the prediction model was tested by comparing simulated values and measured values. The results showed that collinearity existed among the predictors, and the prediction model of the minimum temperature inside solar greenhouse based on ridge regression could overcome the influence of collinearity on the model parameters. Between the simulated values and measured values, the accuracy rate of the absolute error (≤3℃) was 98.4%, the coefficient of determination (R²) was 0.8543, the root mean square error (RMSE) was 0.7849℃, and the accuracy of the prediction model was high. The minimum temperature prediction model based on ridge regression could reasonably and effectively predict the minimum temperature inside the local solar greenhouse.

Key words: solar greenhouse, minimum temperature, ridge regression, prediction model