Abstract:

Objective In the context of modern agricultural practices, regulating the indoor microclimate of Venlo-type greenhouses during the summer months through mechanical ventilation remains a significant challenge. This is primarily due to the nonlinear dynamics and strong coupling characteristics inherent in greenhouse environmental systems, where variables such as temperature, humidity, and CO₂ concentration interact in complex, interdependent ways. Traditional control strategies, which often rely heavily on empirical knowledge and manual intervention, are insufficient to cope with such dynamic conditions. These approaches tend to result in imprecise control, substantial time delays in response to environmental fluctuations, and unnecessarily high energy consumption during ventilation operations. Therefore, the study combines computational fluid dynamics (CFD) model with a multi-objective particle swarm optimization (MOPSO) algorithm to establish a joint optimization framework, aiming to address the issues of significant time delays and excessive operational energy consumption caused by vague environmental control strategies in this scenario. Methods To build a reliable simulation and optimization framework, environmental parameters such as temperature, wind speed, and CO₂ concentration were continuously collected via a network of environmental monitoring sensors installed at various positions inside the greenhouse. These real-time data served as validation benchmarks for the CFD model and supported the verification of mesh independence. In the CFD model construction, the internal structure of the Venlo-type greenhouse was precisely reconstructed, and appropriate boundary conditions were set based on empirical data. The airflow dynamics and thermal field were simulated using a finite volume-based solver. Four grid resolutions were evaluated for grid independence by comparing the variations in output metrics. The controllable parameters in the model included fan outlet wind speed and cooling pad condensation temperature. These parameters were systematically varied within predefined ranges. To evaluate the greenhouse environmental quality and energy consumption under different control conditions, three custom-defined objective functions were proposed: temperature suitability, CO₂ uniformity, and fan operating energy consumption. The MOPSO algorithm was then applied to conduct iterative optimization over the defined parameter space. At each step, the objective functions were recalculated based on CFD outputs, and Pareto-optimal solutions were identified using non-dominated sorting. After iterative optimization using the algorithm, the conflicting objectives of environmental deviation and energy consumption were balanced, leading to the optimal range for the greenhouse environmental control strategy in this scenario. Results and Discussions The experimental results showed that the environmental field simulation accuracy of the CFD model was high, with an average relative error of 5.7%. In the grid independence test, three grid types—coarse, medium, and fine—were selected. The variations in the grid divisions were 1.7% and 0.6%, respectively. After considering both computational accuracy and efficiency, the medium grid division standard was adopted for subsequent simulations. The optimization strategy proposed in this study allows for closed-loop evaluation of the environment. The algorithm set the population size to 100 particles, and within the specified range of fan outlet wind speed and cooling pad condensation temperature, each particle iterates 5 times for optimization. The position updated in each iteration was used to calculate the values of the three objective evaluation functions, followed by non-dominated comparison and adaptation of the solutions, until the optimization was complete. In the Pareto surface fitted by the output results, the fan outlet wind speed ranges from 2.8 to 5.4 m/s, and the inlet temperature ranges from 295.3 to 299.7 K. Since the evaluation functions under the environmental control strategy were all in an ideal, non-dominated state, two sets of boundary control conditions were randomly selected for simulation: operating Condition A [296 K, 3.5 m/s] and operating Condition B [299 K, 5 m/s]. Post-processing contour plots showed that both operating conditions achieve good environmental optimization uniformity. The approximate ranges for each parameter were: temperature from 300.3 to 303.9 K, wind speed from 0.7 to 2.3 m/s, and CO₂ concentration from 2.43 × 10^-5 to 3.56 × 10^-5 kmol/m³. Based on environmental uniformity optimization, operating Condition A focused on adjusting the suitable temperature for crops by lowering the cooling pad condensation temperature, but there was a relative stagnation of CO₂. Operating Condition B, by increasing the fan outlet wind speed, focused on regulating CO₂ flow and diffusion, but the gradient change of airflow near the two side walls was relatively abrupt. Conclusions This study complements the research on the systematic adjustment of greenhouse environmental parameters, while its closed-loop iterative features also enhance the simulation efficiency. The simulation results show that by arbitrarily combining the optimal solution set within the theoretical range of the strategy output, optimization of the targeted objectives can be achieved by appropriately discarding other secondary objectives, providing a reference for regulating the uniformity and economy of mechanical ventilation in greenhouses. Subsequent research can further quantify the coupling effects and weight settings of each objective function to improve the overall optimization of the functions.

Key words: greenhouse, environmental control strategy, multi-objective optimization, CFD, mechanical ventilation