A Neural Network Model for Photosynthesis Prediction

A common problem in greenhouse production is the CO₂ supply inside of the greenhouse to increase crop yields by stimulating photosynthesis; However, CO₂ is one of the most expensive production inputs. Therefore it is necessary to apply CO₂ only when it is necessary in order to reduce cost. Consequently a good greenhouse control tool was necessary, so two neural network models were developed, one for CO₂ prediction and the other for photosynthesis prediction by doing this, we can know the photosynthesis tendency. If this process is increasing CO₂ supply continuous, on the contrary CO₂ stops. For the CO₂ model eight input variables were used and a 1800 data pattern. The ANN was feed with 200 different input data (June 22, 3:20 p.m to 19:55 p.m) and the MSE error between actual and predicted values was 535. The results from the CO₂ model was linked with the photosynthesis model. In this last model seven variables were used. Predictions were very good in both cases. The sensitivity analysis performed in CO₂ and photosynthesis prediction show that relative humidity is one of the most important variables affecting photosynthesis, after solar radiation, and CO₂.