Determining Optimal CO2 Concentration of Greenhouse Tomato Based on PSO-SVM

Abstract

<abstract> Carbon dioxide (CO₂) enrichment increases the net photosynthetic rate (Pn), which can significantly affect the quality and quantity of greenhouse tomato plants (). Hence, establishing an optimal prediction model of Pn and accurately managing CO₂ concentration are important for commercial tomato production. A photosynthesis prediction model of the tomato plant in the seedling stage is proposed by applying a support vector machine (SVM). First, the environmental parameters affecting the growth of the tomato plant, including CO₂ concentration, photosynthetic photon flux density, air temperature, relative humidity, soil temperature, and soil moisture are used as input variables of the SVM. These environmental parameters which are monitored in real time by using wireless sensor nodes. Pn of the tomato plant is used as the output variable, which was obtained by using a portable photosynthesis system. Second, a rough set theory is used to analyze the correlation and dependence between datasets and reduce data redundancy before modeling. Third, improved particle swarm optimization (PSO) is designed to optimize the parameters of the SVM model. The association between the CO₂ concentration and Pn is predicted by the established model, in which the optimal CO₂ concentration under different environmental conditions is obtained. Finally, a regulation model of CO₂ concentration is created by processing multivariate nonlinear regression of the corresponding environmental conditions. Results showed that the determination coefficient R² of the photosynthesis prediction model with minimum attributes were 0.933, the root-mean-square error was 0.881, and elapsed time was 9.18 s, whereas the corresponding parameters of the model with complete attributes was 0.915, 1.170, and 9.79 s, respectively. These results showed that the model with minimum attributes had higher accuracy and faster training speed to provide a reliable basis for determining the relationship between CO₂ concentration and Pn. The R² of the CO₂ concentration regulation model was 0.891. The results demonstrated good correlation and similarity. The conclusion provides a theoretical basis for the optimal regulation of an important raw material of tomato photosynthesis.