South Korea’s Rural Development Administration is testing AI-TEMPI at nine farms, following an internal greenhouse trial. Its 16 October release describes a system trained on experienced growers’ environmental records and control settings, aimed at smaller operations and growers with less experience.[1]
Temperature control is a continuing response to changing physical conditions. UMass Extension explains that natural greenhouse ventilation depends on wind and temperature differences. Warm air escaping through roof vents must be replaced by incoming air; the building’s openings and orientation affect that exchange. A setting that produces adequate cooling under one set of conditions can therefore behave differently when the wind or outside temperature changes.[2]
Expertise includes more than a preferred temperature. A useful record pairs a decision with the circumstances that prompted it. That pairing is what another grower would need to judge whether a previous adjustment applies now.
A learned recommendation meets the physical greenhouse
AI-TEMPI calculates ventilation and heating setpoints from current conditions; an existing environmental controller operates the equipment. Large changes are applied gradually. Recommendations outside the experienced growers’ temperature range trigger a notice so the farmer can decide. The farm pilots began in August.[1]
What the harvest comparison measures
RDA reports 10.5kg of tomatoes per plant in its 2024–May 2025 internal trial, 98% of the expert-managed comparison. It provides no farm-pilot yield, energy-saving result or statistical uncertainty. The technology was transferred to a domestic company in 2025; the release gives no retail price.[1]
Harvest weight is a concrete outcome, but it is only one part of a growing operation. Two greenhouses could produce similar quantities while requiring different amounts of labour or energy. Without those measurements, the percentage cannot answer whether the system pays for itself. That is an accounting distinction, rather than a reason to disregard a measured crop.
Physical maintenance also remains relevant. UMass advises cleaning and lubricating motorised vent controllers and checking that vents close tightly before the heating season, because poor closure loses heat. It recommends using a wind sensor to protect vents in strong winds. Software recommendations operate within these mechanical constraints.[2]
RDA plans further development and wider farm application through a commercialisation support programme running from July 2026 to June 2027.[1]
Sources & further reading
- 인공지능이 중소형 스마트온실 온도 관리한다 [AI manages temperature in small and medium smart greenhouses] Rural Development Administration, South Korea · 16 Oct 2026 · Accessed 28 Oct 2026
- Ventilation for Greenhouses University of Massachusetts Amherst Extension · Accessed 28 Oct 2026