Dissertation Title: Energy Management in Plant Factories Using Advanced Energy Materials and Artificial Intelligence
Date: 2025/05/27 - 2025/05/27
Dissertation Title: Energy Management in Plant Factories Using Advanced Energy Materials and Artificial Intelligence
Speaker: Wenyi Cai, Ph.D. candidate at UM-SJTU Joint Institute
Time: 2:00 PM - 3:30 PM, Tuesday, May 27, 2025 (Beijing Time)
Location: Room 415, Longbin Building
Abstract
Plant factories with artificial light (PFALs) based on vertical farming have the advantage of producing high-quality, consistent plants in a controlled and efficient approach, making them increasingly popular technologies for resource-efficient plant production, especially in urban areas or arid regions. Compared to open-field farming and greenhouse farming systems, PFALs have the advantages of high crop yields, efficient use of water and nutrients, reduced requirements for pesticides, and year-round continuous production. However, the energy consumption and the resulting running cost are extremely high, which restricts the development of PFALs. Several challenges hinder the energy management technology and development of PFALs. Firstly, the research direction for energy-saving technologies in PFALs is not clear, a consistent metric for measuring different energy-saving technologies is required, and a detailed analysis of the feature importance of each energy-consuming parameter in PFALs is required to guide further energy-saving technology exploration on the specific energy-consuming parameter and equipment. Secondly, according to the feature importance analysis, stable, cost-effective, and highly efficient energy-saving methods are required for energy management in PFALs. Lastly, the field application of energy-saving PFALs in arid regions is difficult to implement because of water scarcity. As a result, the objective of this dissertation is to address these challenges and finally demonstrate the contributions through a small-scale practical application.
In this dissertation, the first challenge is addressed by proposing a metric, energy consumption per unit yield, to quantitatively evaluate different energy-saving methods and energy-consuming parameters, and a simulation model integrated with EnergyPlus and the XGBoost algorithm has been proposed in Chapter 2 to derive the feature importance of each energy-consuming parameter. Consequently, the photosynthetic photon flux density of the lighting system is the most energy-consuming parameter based on the metric analysis, which is required to be reduced without yield loss for further energy management in PFALs.
According to the most energy-consuming parameter analysis, a significant energy-saving method in PFALs through reflective materials integration is proposed in Chapter 3, and the energy-saving capability is quantitatively evaluated with a simulation model based on the Monte-Carlo ray tracing method and computational fluid dynamics simulation. The energy consumption per unit yield (fresh weight) can be reduced by 36% while the growth performance of plants can be enhanced. A stable and low-cost reflective film is fabricated with high overall reflectivity, where the energy-saving effectiveness is demonstrated with optical simulation analysis and the corresponding computational fluid dynamics simulation. The corresponding field experiments of three different types of lettuce under different light intensities are conducted, and a reduction of light energy consumption is demonstrated while the dry weight and fresh weight still increase. An additional benefit is that the growth performance becomes better due to the more uniform lighting with the integration of reflective materials. Consequently, a simple yet stable strategy is proposed to reduce the energy consumption per unit yield in PFALs for energy management.
A few-shot agentic assessment framework integrated with the Segment Anything Model and Large Language Models is proposed in Chapter 4, and the energy-saving capability of this framework has been quantitatively analyzed. This energy-saving method selects better seedlings for transplanting at the plant seedling stage. Under the condition of reducing the light intensity, the energy consumption of production can be reduced by more than 20 %. This method shows reliable accuracy with around 80 % and remarkable interpretability compared to other approaches while reducing the data requirements and finally increasing the plant yields. The Segment Anything Model can generate pixel-wise segmentation masks for different plant instances across different growth phases, and the few-shot agentic framework is constructed to automatically derive the assessment rating on the extent of expert knowledge. After that, the transplanting group is selected based on the framework. Consequently, Chapter 4 presents a rapid yet reliable plant growth information measuring system and can greatly reduce the energy consumption per unit yield in PFALs for energy management.
Finally, a small-scale practical application based on the energy-saving methods and a supplementary water purification assistance system is proposed to address the field application problems in arid regions. The results figure out that the water purification system based on the interfacial evaporation method can meet the water source requirements of PFALs, and the small-scale practical application based on the energy-saving methods in this dissertation can reduce the energy consumption by more than 40 % while the growth performance of plants can be enhanced. The water distillation system is based on the interfacial evaporation method. It has the potential to relieve freshwater scarcity in arid regions based on an oxidized copper foam cube dip-coated with CuS nanoparticles and agarose gel. A high freshwater collection rate is demonstrated with heat transfer analysis and nutrient waste evaporation experiments. Finally, a small-scale cultivation tray integrated with a supplementary water purification system, based on reflective material integration and a few-shot agentic assessment framework, is demonstrated for an energy-saving PFALs application.
This dissertation presents progress in the development on energy management of PFALs, which can achieve energy management application in arid regions and can provide guidance for the future development of PFALs. A consistent metric is proposed for quantitatively evaluating energy-saving effects, and then an energy-saving parameter feature importance analysis is demonstrated. According to the parameter importance, two reliable energy-saving methods combined with the supplementary water purification system are integrated with PFALs, which represents the field application objective of this dissertation. Based on this dissertation, it is possible to analyze the energy consumption of PFALs and derive the corresponding energy management method. This advancement is of immense significance for the future development of energy consumption optimization strategies and daily plant production of PFALs in arid regions.
Biography
Wenyi Cai is a Ph.D. candidate at the University of Michigan-Shanghai Jiao Tong University Joint Institute, supervised by Prof. Hua Bao. Before that, he graduated from University of Michigan-Shanghai Jiao Tong University Joint Institute, with a bachelor’s degree in Electrical and Computer Engineering from 2016 to 2020. His research focuses on the energy management in plant factories using advanced energy materials and artificial intelligence.