Photosynthetic Rate Prediction Model Based on PSO-LSSVM for Optimization and Control of Greenhouse Environment

Due to lack of efficient measuring means, it is not possible to realize the accurate measurement for the photosynthetic rate of greenhouse crops, and it affects the reliability of optimization and control based on photosynthesis demands for greenhouse environment. Therefore, basing on Least Squares Support Vector Machine (LSSVM), a photosynthetic rate prediction model is established in this paper. Through the multi-factor experiment on tomato seedlings, the photosynthetic rate data is obtained under different temperature, CO2 concentration and photosynthetic photon flux densities (PPFD) conditions. Basing on the obtained data, the photosynthetic rate prediction model is established by training LSSVM. Aiming at the influence of the empirical values of parameters (the penalty parameter γ and kernel constant σ ) on the model prediction performance, Particle Swarm Optimization (PSO) algorithm is used for optimizing and determining the two parameters values. Compared with the commonly used BP model and SVM model, the proposed model has obvious advantage on the prediction accuracy and performance. In addition, the Maximum Relative Error (MRE) of the proposed prediction model is 0.0224, and the Root Mean Square Error (RMSE) is 0.2324, which indicates that the proposed model has the higher prediction accuracy and provides a good model basis for optimization and control of greenhouse environment.