Abstract: Photosynthesis is the basis of crop growth and metabolism. CO 2 concentration and soil moisture are the important environmental factors affecting plant’s photosynthetic rate under controlled temperature and light intensity in greenhouse. To effectively evaluate the effect on plant’s photosynthesis, reasonably elevating CO 2 concentration under different soil moisture conditions is of great significance to achieve precision regulation of CO 2 concentration. To achieve the requirements, the photosynthetic rate prediction models based on back-propagation (BP) neural network were proposed at different growth stages of tomato plants. The two-factors interaction experiment was designed, in which the CO 2 concentration was set to four different levels ((700±50) (C1), (1 000±50) (C2), (1 300±50) μmol/mol (C3), and ambient CO 2 concentration in greenhouse (450 μmol/mol, CK)) combined with three different soil moisture levels (35%~45% (low), 55%~65% (moderate), 75%~85% of saturated water content (high)). The sensor nodes of WSN were used to realize the real-time monitoring of greenhouse environmental factors, including air temperature and humidity, light intensity, CO 2 concentration and soil moisture. An LI-6400XT photosynthesis analyzer was used to measure net photosynthetic rate of tomato leaf. The environmental factors were used as input variables of models after processed by normalization, and the photosynthetic rate was taken as the output variable. The model verification test was conducted by comparing and analyzing the observed values and predicted data. The results indicated that the training determination coefficient (R 2) of photosynthesis prediction model was 0.953, and root mean square error (RMSE) was 1.019 μmol/(m 2 ·s); testing R 2 of the model was 0.925, RMSE was 1.224 μmol/( m 2 ·s ) at seedling stage. At flowering stage, the training R 2 of the model was 0.958 and RMSE was 0.939 μmol/(m 2 ·s); testing R 2 of the model was 0.920 and RMSE was 1.276 μmol/(m 2 ·s). At fruiting stage, the training R 2 of the model was 0.980 and RMSE was 0.439 μmol/(m 2 ·s); testing R 2 of the model was 0.958 and RMSE was 0.722 μmol/(m 2 ·s). It was concluded that the model based on BP neural network reached high accuracy. Furthermore, the relationship between CO 2 concentration and photosynthetic rate was described by the established BP neural network model aiming at CO 2 saturation points under different soil moisture conditions at different growth stages. The observed and predicted results showed the same trend. The results can provide a theoretical basis for quantitative regulation of CO 2 enrichments to tomato plants in greenhouse.