Profit Optimization of Greenhouse Tomato Production Using Yield Prediction, Energy Cost Simulation and Machine-learning

In greenhouse tomato production, maintaining optimal environmental conditions—such as temperature and CO2 concentration—has traditionally been considered essential for maximizing crop yield. However, this study demonstrates that the most profitable strategy does not always align with biologically “optimal” conditions. Instead, the proposed framework occasionally selects mildly stressful environments that reduce operational costs and ultimately increase net profit. We developed a simple, profit-oriented optimization framework that integrates a yield prediction model, energy cost simulation, and Bayesian optimization via the Metropolis-Hastings algorithm. The framework was calibrated using data from a research greenhouse in Tsukuba equipped with advanced environmental control systems, including heat pumps. Simulation results showed that optimized control strategies could more than double profitability. Practical validation was conducted during the autumn–winter 2024 season at agricultural research stations in Gifu and Aichi prefectures, where heat pumps were not installed. Weekly environmental settings—including CO2 concentration, heating, and ventilation—were optimized to maximize profit rather than yield alone. Field experiments confirmed up to a 1.5-fold increase in profit, even in greenhouses without advanced equipment. Importantly, both Aichi and Gifu achieved increased yields without incurring losses, primarily due to effective CO2 enrichment. In Aichi, although heat stress reduced overall productivity, the conventional plot also suffered yield loss, making the optimized strategy appear more favorable in relative terms. Simulated profit reductions due to heat stress were similar in both plots, indicating that the optimization mitigated—but did not eliminate—environmental impacts. The framework selected temperature settings near the threshold where yield reduction begins (daily mean ≈23°C), balancing reduced operational costs with slight yield suppression. These findings suggest that profit-oriented environmental control, which may involve a tolerable temperature, can be more effective than traditional yield-maximization approaches, especially under economic and operational constraints.