Neural Networks Models for Temperature and CO2 Control
In this study, we are interested in regulating two important variables inside of the greenhouse: temperature and CO2 enrichment for two cabins in a experimental greenhouse at the Humboldt University of Berlin. Predicting the behavior of these two variables and photosynthesis will allow us to turn on and off the controls such as heating system, vents opening or CO2 enrichment at the right time, in order to save energy and keep the plants inside of the comfort zone. Artificial Neural Networks (ANN) were used because of their ability to capture the non linear relationships governing the changes in the greenhouse environment. Temperature was predicted 5 and 10 min ahead of the sensor signal, with MSE errors between measured and predicted values of 0.088 and 0.029, respectively. The CO2 predicted from the model was used as an input in the photosynthesis model. In this last model, seven variables were used and the predictions were highly precise with a MSE errors of 0.0563 and 0.0974 for photosynthesis 5 and 10 min ahead, respectively. A sensitivity analysis was performed in the photosynthesis model showing that relative humidity is an important variable for CO2 levels and for the photosynthesis process. The predicting models will allow to achieve our final goal which is to replace the sensors and give predictive information for a higher control quality in an open loop control system.
Cited References Fulltext