Many believe one way to feed the world is to move our crops indoors where they’ll be shielded from the vagaries of weather, pests, and disease. Yet this type of farming requires large amounts of energy to keep plants artificially lit and ventilated. Can we escape this bind?
Researchers on a new study believe we can, at least partly, with the help of artificial intelligence. Their Nature Food study shows AI can finetune indoor growing conditions to slash energy use by a third.
For the study, researchers developed a computational model of a typical ‘PFAL’, a ‘plant factory with artificial lighting’. The model incorporates information on indoor environmental factors including temperature and humidity levels, CO2 concentration (which plants need to photosynthesize) and simulates how changes affect plant photosynthesis and growth. The model was then combined with an AI framework, which was trained on the data to recognize the optimal growing conditions for the indoor plants. Part of its role was to determine whether growth rates could be maintained with lower energy use.
The researchers used lettuce as their test-crop, and they simulated PFAL operations under 10 different geographical locations, ranging from Reykjavík, to Dubai, because although PFALS are relatively contained, energy use can still be influenced by temperatures outside. The researchers wanted to make sure their study reflected this variability.
That makes it more striking that the experiments revealed across all ten simulations that the AI was able to reduce energy use without altering crop growth. Most of this was achieved by reducing the artificial lighting required for photosynthesis, and the ventilation needed to manage humidity and heat. Overall, these optimized systems led to a striking average energy saving of 32.34% compared to control experiments; at minimum, the AI reduced energy use by 23.6%.
Because the AI was able to observe and regulate according to the unique interplay between plants and environmental factors in individual locations, it effectively found the sweet spot in each one—for example where dimming the lights for periods of time would not reduce the lettuce yield. Ventilation presents a challenge in this regard, because it is closely linked to local conditions. In hotter locations especially, ventilation is essential to cool down crops, but at higher rates it disturbs the concentration of CO2 that drives photosynthesis. However the AI found a sweet spot in these locations too, by reducing ventilation rates during the light period when plants need to photosynthesize, and amping it up in the dark period when CO2 concentrations are less significant.
While ventilation requirements in hot places did lead to higher energy use, overall the savings were still significant in all locations. AI regulation managed to reduce energy use from between 9.5-10.5 kilowatt hours per kilo of lettuce in a regular PFAL, to 6.42 kilowatt hours in cooler climates, and 7.26 kilowatt hours in warmer climates.
There are questions to answer about whether these energy savings would still apply in vast, warehouse-style indoor farms—and how well these findings will translate to real-world farms in the first place, the researchers note. But their study at least suggests that agricultural innovation needn’t come with such a huge environmental trade-off: we have tools to help us innovate out of that bind.
You et. al. “Artificial intelligence can regulate light and climate systems to reduce energy use in plant factories and support sustainable food production.” Nature Food. 2024.
Image: © Anthropocene magazine
