It comes as no surprise to Dutch greenhouse owner Frank van Kleef that an AI-powered greenhouse outperforms the benchmark Dutch grower - after all, there's a reason he left his own greenhouse company to join the British/Dutch company, Optimal, which is pioneering AI-powered greenhouse cultivation.
In their commercial-scale greenhouse in Westdorpe, Optimal's system is already achieving higher yields and higher Brix levels while also saving energy and reducing CO2 emissions. "The computer is always consistent in its decisions," Frank explains.
"There are not enough skilled growers available to service the greenhouses that are being built around the world," says Optimal's Founder and CEO, Dave Hunter. "This problem needs to be addressed if the greenhouse industry is to expand. All greenhouses worldwide need operational knowledge, consistent results, and an efficient way to run their facilities. If this can be done at scale, then greenhouses will thrive globally and become critical to humanity's food security in the face of climate change."
Leveraging Grower Expertise
To gain a deeper understanding of the industry and the challenges it faces, Dave joined forces with Frank van Kleef. Frank has been one of the owners of Royal Pride, which became one of the leading greenhouse companies in the world that operates 70 ha of tomatoes and cucumbers. He now focuses full-time on developing the Optimal system and actively participates in the Westdorpe demonstration greenhouse. Frank's role is not that of a traditional grower but more of a supervisor who assesses processes and evaluates results in terms of quantity and quality. He does this meticulously, as the greenhouse's produce is sold through his social enterprise, Boeregoed.
Predictive Control System
Combining Frank's knowledge with AI, the Optimal team first developed a predictive control system. A digital twin of the greenhouse is continuously updated with real-time information and weather forecasts. Using advanced optimization software, the inputs to the greenhouse are re-planned each minute (e.g., heating pipe temperatures and vent positions). Optimal then applies these inputs to the greenhouse via an interface with the process computer. Dave explains, "We can predict the full impact of decisions before they are made and anticipate the effects of weather changes in a way that no human grower can do."
Frank explains the biggest advantage lies not just in optimizing the greenhouse environment but in preventing errors. "When you present the same situation to a grower ten times, they'll act the same way seven times and differently three times. That's human nature. A computer doesn't do that. It always makes the best choice." Dave adds that the system never sleeps. "It never over- or under-irrigates, and you don't need to adjust the climate because it's already been well-managed in advance. You don't need to react to increased evaporation because the system has already accounted for that. This results in a better overall plant health, a more resilient crop, and thus a better quality and yield."
Software and Operational Knowledge
However, running a greenhouse entails more than just a control system, even if it's an exceptional one. Dave explains, "What we offer our customers is the operation of their greenhouses on their behalf, ensuring optimal efficiency and reliability season after season."
To achieve this, all tasks inside and outside the facility are documented in software-based Standard Operating Procedures. Greenhouse workers and operators are trained to follow these procedures, and a smartphone app guides them throughout their workday, providing instructions on what tasks need to be performed, how to perform them, and what to do when encountering irregularities.
The system also generates labor schedules, offering insights into labor requirements and worker performance. Daily and weekly feedback sessions are conducted to assess the team's performance. Dave states, "Labour is a critical factor in greenhouse operations, with significant implications for cultivation. Standardizing it allows us to achieve optimal and predictable results."
Optimal Performance and Real-World Deployment
So, what are the results of this comprehensive system? In their demonstration greenhouse, the team has achieved impressive results, including a 13% higher yield, a 14% higher Brix level, and energy savings of 27%, along with a 20% reduction in CO2 emissions compared to a benchmark Dutch greenhouse.
The Optimal team has also conducted various experiments involving ten different tomato and cucumber varieties. "We conduct experiments here ourselves and use this data to continually improve our performance," Dave explains. "Over the last five months, we've been running a low-energy experiment to optimize performance in high-energy-cost environments, such as those experienced in Europe over the last two years."
Real-World Deployment and Future Expansion
Currently, discussions are underway with leading greenhouse owners in the United States, the UK, and other locations. Dave notes that the system easily adapts to different climatic conditions, and there are no minimum technical requirements or specific greenhouse needs to get started. Dave explains, "We can work in any greenhouse: we install additional sensors and connect them to cloud servers and are up and running in days." The team is also exploring expansion into strawberries, peppers, and leafy greens.
When asked what he's most proud of, Dave emphasizes that Optimal's success lies in the synergy between humans and machines—a complex but rewarding endeavor. "Our engineers understood that AI and predictive control models would result in better greenhouse outcomes: they view a greenhouse as a system that can be controlled through simulations and optimization. However, it's vital to align people, the system, and the hardware to work together. That's where Frank and the other experienced operators in Optimal's team have added tremendous value."