Abstract

Global food security faces unprecedented challenges from climate change and population growth, demanding transformative approaches to sustainable nutrition. This study demonstrates how artificial intelligence revolutionizes distributed food production through optimized microalgae cultivation that simultaneously addresses food security and climate mitigation. A mechanistic model for Arthrospira platensis cultivation was developed and integrated with multi-agent reinforcement learning (MARL) control in photobioreactor systems across diverse global climates. AI-driven control achieved 14% lower energy consumption while capturing 78% more CO2 compared to conventional methods, making microalgae cultivation a net carbon sink. The MARL system demonstrated robust performance across eight climate zones and four seasons, enabling year-round sustainable protein production with enhanced climate resilience. This research establishes a scalable pathway for decentralized, climate-positive food systems contributing to multiple Sustainable Development Goals. The integration of AI with biological systems offers a replicable model for sustainable intensification of food production while actively removing atmospheric carbon, providing a critical tool for achieving global food security within planetary boundaries.

Spirulina is a protein-rich microalga that grows quickly and can be cultivated without farmland, making it an attractive source of alternative protein. Yet producing it efficiently remains challenging, particularly as temperature and sunlight vary across regions and seasons.

A research team, led by Professor Hankwon Lim of the Graduate School of Carbon Neutrality at UNIST, working with Professor Riezqa Andika of the University of Indonesia, has developed an AI-based system that adjusts spirulina cultivation to changing environmental conditions. In simulations using climate data from different regions, the system reduced energy use while accelerating growth and increasing carbon dioxide uptake.

Spirulina cultivation depends on several closely connected factors, including light, temperature, nutrients, carbon dioxide, and acidity. Increasing the supply of carbon dioxide, for example, can promote growth but may also make the culture too acidic. More artificial light can support photosynthesis, but at the cost of greater electricity use. Finding the right balance becomes more difficult as outdoor conditions change.

The new system uses multi-agent reinforcement learning (MARL), in which five AI agents manage these cultivation conditions together. Each controls one factor—light, temperature, nutrients, carbon dioxide, or acidity—and adjusts lighting, heating and cooling, and material inputs as conditions change.

To develop the system, the researchers first created a virtual cultivation environment that models spirulina growth under different combinations of environmental and biological conditions. The AI then learned through repeated simulations to favor decisions that increased spirulina production while reducing energy consumption.

Unlike conventional systems that rely on predetermined rules, the AI can adjust its response according to the overall state of the culture. This allows it to account for interactions among growing conditions, including effects that may not appear immediately.

The team evaluated the system using weather data covering four seasons in eight cities across different climate zones. Compared with conventional control methods, the AI system reached the target spirulina concentration 59 hours earlier and reduced the average energy required to produce one kilogram of dried spirulina by 14%.

The benefits varied by location and season. Energy consumption fell by as much as 20% in Ulsan and 21% in Beijing, while carbon dioxide uptake increased by an average of 78%. The proportion of supplied CO₂ absorbed by the spirulina also rose from 44% to 49%.

“Beyond accelerating microalgae growth, this study provides a way to adapt cultivation to different climates while using less energy and increasing carbon dioxide uptake,” said Ahmad Syauqi, first author of the study. “The results show how AI-based control could help improve both the productivity and sustainability of microalgae cultivation.”

“The system has been evaluated so far in a simulated environment,” said Professor Lim. “Our next step is to validate it in cultivation facilities equipped with sensors and control systems, with the goal of developing the technology for practical use in agriculture and aquaculture.”

The findings of this research have been published online in  Bioresource Technology on June 19, 2026. The research was supported by the Carbon Neutrality Demonstration and Research Center of UNIST, the National Research Foundation of Korea (NRF), and the Ministry of Science and ICT (MSIT).

Journal Reference

Ahmad Syauqi, Akmal Ramdani, Hankwon Lim,  et al. , “Sustainable distributed microalgae cultivation for food production through artificial intelligence-driven photobioreactor control,”     Bioresour. Technol.,  (2026).