Intelligent Frost Prediction and Active Protection Cyber-physical Systems in the Agricultural Sector

Zhou, Ian

Intelligent Frost Prediction and Active Protection Cyber-physical Systems in the Agricultural Sector

Zhou, Ian

Permalink

Publication Type:: Thesis
Issue Date:: 2023

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (18.93 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Zhou, Ian
dc.date.accessioned	2023-09-19T01:46:55Z
dc.date.available	2023-09-19T01:46:55Z
dc.date.issued	2023
dc.identifier.uri	http://hdl.handle.net/10453/172189
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	Frost damage in broadacre cropping and horticulture (including viticulture) results in substantial economic losses to producers and may also disrupt associated product value chains. Frost risk windows are changing in timing, frequency, and duration. Faced with the increasing cost of mitigation infrastructure and competition for resources (e.g., water and energy), multi-peril insurance, and the need for supply chain certainty, producers are under pressure to innovate in order to manage and mitigate risk. Frost protection systems are cyber-physical systems comprising sensors (event detection), intelligence (prediction), and actuators (active protection methods). These systems are an important tool and cost factor in fighting against frost. Therefore, there is a need to improve the efficiency and effectiveness of frost protection cyber-physical systems. This study adopts the Internet of Things 2.0 architectures and emphasizes the dimensions of machine learning intelligence, scalability, interoperability, and user-friendliness in frost protection systems. This research also improves on the limitations of existing frost protection systems and the prediction methods that control the systems. The limitations are historical data dependence, low prediction temporal resolution, non-real-time response, and low fault tolerance. In response to the limitations of low prediction temporal resolution and non-real-time response, the existing 12-24 hours prediction methods are extended by artificial neural networks and recurrent neural networks for near real-time frost prediction. A minute-wise regression model to predict the next hour minimum temperature is proposed. The minute-wise model further highlights the system dependence on local historical data and sensors. To decouple this dependence, a spatial interpolation-based frost prediction system is implemented. This model-based system only requires data from existing weather stations to predict frost at any new sites. Combining the results of the previous two models, a cyber-physical system framework is proposed to improve the system fault tolerance. This framework is a modular design with the local data-based system as the primary predictor and the model-based system as both a secondary predictor and system stopping mechanism. The result shows improvements in operational cost compared to traditional methods. The final contribution is the improvement of energy efficiency on edge-based prediction. A spatially generalized model with a smaller prediction window is constructed and deployed on a LoRa transmission node. The proposed system not only improves energy efficiency, it also reduces the false positive rate.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/172189/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2023 Ian Zhou
dc.rights	au.edu.uts.lib/cph
dc.title	Intelligent Frost Prediction and Active Protection Cyber-physical Systems in the Agricultural Sector	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

Frost damage in broadacre cropping and horticulture (including viticulture) results in substantial economic losses to producers and may also disrupt associated product value chains. Frost risk windows are changing in timing, frequency, and duration. Faced with the increasing cost of mitigation infrastructure and competition for resources (e.g., water and energy), multi-peril insurance, and the need for supply chain certainty, producers are under pressure to innovate in order to manage and mitigate risk. Frost protection systems are cyber-physical systems comprising sensors (event detection), intelligence (prediction), and actuators (active protection methods). These systems are an important tool and cost factor in fighting against frost. Therefore, there is a need to improve the efficiency and effectiveness of frost protection cyber-physical systems. This study adopts the Internet of Things 2.0 architectures and emphasizes the dimensions of machine learning intelligence, scalability, interoperability, and user-friendliness in frost protection systems. This research also improves on the limitations of existing frost protection systems and the prediction methods that control the systems. The limitations are historical data dependence, low prediction temporal resolution, non-real-time response, and low fault tolerance. In response to the limitations of low prediction temporal resolution and non-real-time response, the existing 12-24 hours prediction methods are extended by artificial neural networks and recurrent neural networks for near real-time frost prediction. A minute-wise regression model to predict the next hour minimum temperature is proposed. The minute-wise model further highlights the system dependence on local historical data and sensors. To decouple this dependence, a spatial interpolation-based frost prediction system is implemented. This model-based system only requires data from existing weather stations to predict frost at any new sites. Combining the results of the previous two models, a cyber-physical system framework is proposed to improve the system fault tolerance. This framework is a modular design with the local data-based system as the primary predictor and the model-based system as both a secondary predictor and system stopping mechanism. The result shows improvements in operational cost compared to traditional methods. The final contribution is the improvement of energy efficiency on edge-based prediction. A spatially generalized model with a smaller prediction window is constructed and deployed on a LoRa transmission node. The proposed system not only improves energy efficiency, it also reduces the false positive rate.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/172189