CNN-Assisted Growth Monitoring and Stress Management of Cucumber in Semi-Transparent PV Greenhouses for Agrivoltaics

Engineering notes

Engineering notes will be added by the Power for Solar editorial team.

Chinese explanation / 中文解读

中文解读待补充：本站会优先为光伏效率、钙钛矿太阳能电池、储能技术、太阳能热利用、BIPV、并网技术等高价值论文补充中文说明。

Original abstract

Currently operating commercial Photovoltaics (PV) systems combined with agricultural production (Agrivoltaics) have the ability to harvest renewable energy and crop products simultaneously. In Controlled Environment Agriculture (CEA), using Semi-Transparent Photovoltaics (ST-PV) and the ability to micro-manage climate and shadowing are advantageous for the production of high-value crops like cucumber. This study targeted the start of the cultivation of cucumber seedlings under evolving CEA-PV systems coupled with the incorporation of greenhouse experiments and computer-vision driven phenotyping to build an analytical and control framework for cucumber cultivation in an evolving CEA-PV system. This method implemented the control of cucumber plant shading and irrigation in real time using monitored plant vigor. Shading and irrigation control relied on monitored plant vigor assessed using a U-Net++ implementation for canopy segmentation, an EfficientNet-B3 implementation for stress differentiation, and a CNN regressor for growth trait evaluation. Uniform environmental and fertigation conditions were achieved in the greenhouse to study the effects of four shading conditions (0%, 20%, 40%, 60%) on cucumber growth. Simulated, yet realistic results depicted cucumber yield as relatively stable (within ± 4% of full yield) with 20% shading and a 15-20% reduction in evaporation and yield loss as a result of 10-14% to 40-60% shading. The CNN-based models proved to be highly reliable (segmentation IoU 0.91, stress-class F1 0.92, LAI regression R²≈0.93), which enabled precise and extensive monitoring in a non-invasive manner. The greenhouse’s annual PV Yield of 1,550 to 1,750kWh/kWp shows the greenhouse has a higher photo-demand than the PV output, therefore providing a net energy surplus. The results suggest it is possible to combine a cucumber crop with a controlled environment agrovoltaic system with moderate shading to maximize cucumber yield. Also, aided control by Artificial Intelligence (AI) through the closed-loop control system has been proven to optimize the water-use efficiency and yield stability.

Links and sources

• PDF from original source

Comments