2 A Study on Temperature Prediction for Forced Ventilated Greenhouses Based on LE-SVM

Due to a strong difficult challenge of several factors coupling in influencing climate forecasting of inside greenhouse, least squares support vector machine (LS-SVM) is applied in this work to build a prediction model for inside temperature in a forced ventilated greenhouse. Three LE-SVM model are used to conducted the predictions which are self regression model of temperature inside (Ti), a multi-input model considering solar radiation, outside temperature, ventilation rate (To, So, V, T_i), and a multi-input model furthermore considering moisture content (Wi, Wo) on the base of the second one. The comparison between predicted data with real data of temperature inside greenhouse got in experiment is analyzed. And the results show that in 5 min. ahead prediction, the predictions are satisfying with E_MAPE of 2.2%, 1.8% and 3.3% by three LE-SVM models respectively.