Articles
Decision of optimal sensor location for predicting the internal environment of greenhouse using machine learning model
Article number
1426_6
Pages
39 – 42
Language
English
Abstract
The number of large-sized greenhouses has recently increased due to the automation and mechanization with the development of information and communication technologies (ICT). To set an appropriate environment of greenhouse, it is necessary to install and maintain several sensors for internal environmental monitoring and control of the air conditioning system.
This requires machine learning (ML) models to determine the optimal sensor location with minimum number of sensors and which offers good monitoring performance.
In this study, the ML models such as DFF, SVR, and LSTM were developed to predict the internal air temperature for each point in a greenhouse and the optimal sensor location was evaluated based on the best performing model.
The ML models were designed with the use of internal environmental and external weather data measured during the summer season from nine locations in the eight-span greenhouse.
The optimal ML model for greenhouse internal temperature prediction was selected, and the optimal temperature sensor location was selected using the model.
The final ML model which was developed as the prediction current temperature using optimal sensor machine learning model (PCTO-ML) has been proposed to require a minimum kind of learning features.
Among the three ML models, the LSTM model was selected as it showed the highest prediction accuracy of air temperature in the greenhouse.
To minimize the number of training features, four LSTM models (Basic LSTM, simple LSTM-1~3) were developed.
As a result, the sensor which is located in the center was recommended to optimal sensor location in case of using one sensor.
In addition, the optimal sensor location for using two and three sensors has been proposed.
This requires machine learning (ML) models to determine the optimal sensor location with minimum number of sensors and which offers good monitoring performance.
In this study, the ML models such as DFF, SVR, and LSTM were developed to predict the internal air temperature for each point in a greenhouse and the optimal sensor location was evaluated based on the best performing model.
The ML models were designed with the use of internal environmental and external weather data measured during the summer season from nine locations in the eight-span greenhouse.
The optimal ML model for greenhouse internal temperature prediction was selected, and the optimal temperature sensor location was selected using the model.
The final ML model which was developed as the prediction current temperature using optimal sensor machine learning model (PCTO-ML) has been proposed to require a minimum kind of learning features.
Among the three ML models, the LSTM model was selected as it showed the highest prediction accuracy of air temperature in the greenhouse.
To minimize the number of training features, four LSTM models (Basic LSTM, simple LSTM-1~3) were developed.
As a result, the sensor which is located in the center was recommended to optimal sensor location in case of using one sensor.
In addition, the optimal sensor location for using two and three sensors has been proposed.
Publication
Authors
S.M. Kang, I.B. Lee, Y.B. Choi, J.H. Cho, H.H. Jeong, D.I. Kim, S.H. Park
Keywords
ANN, air temperature, greenhouse, LSTM, machine learning, monitoring, optimal sensor location, SVR
Groups involved
- Division Precision Horticulture and Engineering
- Division Greenhouse and Indoor Production Horticulture
- Working Group Nettings in Horticulture (subgroup of Protected Cultivation in Mild Winter Climates)
- Working Group Light in Horticulture
- Working Group Organic Greenhouse Horticulture
- Working Group Vegetable Grafting
- Working Group Modelling Plant Growth, Environmental Control, Greenhouse Environment
- Working Group Protected Cultivation, Nettings and Screens for Mild Climates
- Working Group Computational Fluid Dynamics in Agriculture
- Working Group Design and Automation in Integrated Indoor Production Systems
- Working Group Mechanization, Digitization, Sensing and Robotics
- Working Group Greenhouse Environment and Climate Control
- Division Landscape and Urban Horticulture
- Commission Agroecology and Organic Farming Systems
- Division Vegetables, Roots and Tubers
- Working Group Vertical Farming
Online Articles (75)