Deep Learning-Enabled Temperature Simulation of a Greenhouse Tunnel
Date
2023
Journal Title
Journal ISSN
Volume Title
Publisher
IWACP
Abstract
Agriculture is poised to suffer greatly from the effects of climate change. Prediction models, using deep
learning, have been developed that can simulate and predict conditions in open field farming to combat the
climate variability from climate change. However, deep learning used in precision agriculture, specifically
greenhouse tunnels, is under-researched despite also being affected by this variability. Utilising tunnel data
collected over 42 days, two hybrid deep learning models were designed. Specifically, a hybrid of
convolutional neural network (CNN) and Long Short-Term Memory (LSTM), and a hybrid of CNN and
Bidirectional LSTM (BLSTM). The models are designed to forecast the internal temperature of the tunnel to
support its management. The cooling wet wall state, solar irradiance, inside and outside temperature of the
tunnel are input variables to the developed deep-learning models. Two scenarios are discussed with the
results, the first scenario includes all the external variables as input, while the second scenario only considers
the internal temperature as input. Results show a performance improvement of 48% and 14% computation
time for the CNN-LSTM compared to the CNN-BLSTM model for the two scenarios, respectively. In terms of
the measured loss metrics, both models had varied performance and model fitness, with an average mean
square error of 0.025 across the models and scenarios.
Description
The original publication is available at: http://www.advancesincleanerproduction.net
Keywords
Citation
Booysen, MJ. et al. 2023. Deep Learning-Enabled Temperature Simulation of a Greenhouse Tunnel. Industrial Engineering. 8 pages