Three mathematical models for the dynamic behavior of CO2 inside a greenhouse with the climatic conditions of Central Mexico were analyzed: The first was based on the physical model proposed by Tap, the second was an auto regressive model with external input (ARX) with structure of multiple inputs and single output (MISO), and the third was an artificial neural network model (ANN) with feed-forward type structure and a Levenberq-Marquardt back propagation training. The external CO2, solar radiation, external temperature, wind speed and humidity were used as input for ARX and ANN models. In the Tap model, also the windward and leeward of the greenhouse windows were considered. The results show that the neural networks had a better fit with R2 = 0.91 and mean square error (RMSE) of 16.08, whereas the ARX models had an R2 = 0.77 with an RMSE of 25.5, and the physical model had an R2 = 0.72 with an RMSE of 32.11.
Key words: CO2, greenhouse, ARX model, physical model, neural networks.
Copyright © 2022 Author(s) retain the copyright of this article.
This article is published under the terms of the Creative Commons Attribution License 4.0