Neural networks for simulation of transpiration and photosynthesis in a semi-closed greenhouse
The trend in greenhouse automation systems is to listen to plant signals while trying to minimize the ecological footprint of the vegetable production. In this work, we present a model based on artificial neural networks (ANN), which can be used to support the measuring of plant signals and help to improve the reliability of the data collected. This is important to support control strategies that can link the plants to the actuators. This project was carried out in the framework of the German ZINEG Project, which aims to reduce the energy input needed in the German greenhouse industry. The ZINEG facility, located at the Humboldt-Universität zu Berlin, comprises two similar greenhouses, one operated as closed greenhouse and the other as a standard reference greenhouse, both with a tomato crop grown on rockwool in a closed hydroponic system. Transpiration and Photosynthesis rates, as well as leaf temperature, were measured every five minutes using a phytomonitoring system (BERMONIS) developed at the university. To support the measurements for a reliable control operation, a model was developed using ANN with climatic and theoretical inputs. The model was trained using data from four production years (2010 to 2013) in both aforementioned greenhouses. The trained model is capable of giving a correct output even when the phytomonitoring system fails or is taken away for calibration. The prediction of the model can then be used by two control strategies developed so far: The threshold for opening and closing the thermal screen was triggered by the photosynthesis, and the irrigation was controlled by the transpiration. The results show that the combination of models and measurements can help to introduce and spread plant-related environmental control strategies for greenhouses.
Miranda, L., Lara, B., Rocksch, T., Dannehl, D. and Schmidt, U. 2017. Neural networks for simulation of transpiration and photosynthesis in a semi-closed greenhouse. Acta Hort. (ISHS) 1170:193-200
phytomonitoring system, artificial neural networks, greenhouse models, greenhouse climate control