New modeling and control strategies for reducing disease impact in greenhouses

Laburpena

crop fungal diseases in greenhouses through the use of modeling and control approaches. To complete this task, three independent studies have been developed, namely, a greenhouse climatic model, a greenhouse fungal disease estimator, and a greenhouse climate control algorithm. Finally, combining these models and algorithms, a hierarchical optimization control strategy is proposed to maintain the optimum temperature for cucumber production, but providing priority to the control of disease apparencies. Compared with the current disease management schemes in literature, this new strategy maximizes the accumulated temperature under the condition of reducing diseases, contributing so to reduce the application of chemical pesticides. For the climatic greenhouse modelling, a transient greenhouse model was developed in 2019 and 2020, which uses a mechanism method to estimate the temperature and humidity in typical Chinese solar greenhouses. A novel and easy-to-use wall temperature estimation method based on the energy balance was adopted for the environment model rather than using boundary temperature measurements. In this way, the number of model inputs is considerably reduced, and the proposed model is able to predict future greenhouse climate conditions by using only the weather forecast. The model validation was performed in two different greenhouses (each with different sizes and physical parameters, such as the greenhouse volume, the roof and wall areas, the wall materials and so on) on three typical days in 2019 and 2020, and over four consecutive weeks in different seasons during 2016 and 2019. Promising results were obtained and the model performed well in different operating modes; these included having the vents completely closed, opening the vents, and completely closing the vents in the cold season with an additional thermal insulation blanket covering. The validation results demonstrated that the proposed model can be widely adapted to different sizes of typical Chinese solar greenhouses, as well as to different weather conditions. Thus, the developed model is a flexible and valuable tool that can be used for greenhouse climate simulation, temperature and humidity control, and as a decision-making support system to help in the management of solar greenhouses. To improve the natural ventilation model for the previous greenhouse climate model, a regression-trees natural ventilation model was developed using results from one thousand samples by Computational Fluid Dynamics (CFD) calculations. This model perfectly deals with the combined effect of wind pressure and thermal gradients. This regression-trees natural ventilation model was embedded in the greenhouse climate model and was validated for a 7-day simulation study with promising results. For disease predictions, taking cucumber downy mildew as an example, a new approach was proposed by combining the mechanism greenhouse climate model and a disease model for the forecast of diseases occurrence in greenhouses. The method was evaluated in NPADB (National Precision Agriculture Demonstration Base), Beijing, China using data collected from transplanting to the primary infection occurred in the greenhouse, in the spring season of 2021. First, the dynamic climatic model is used to predict the greenhouse indoor climate 72 hours ahead. Then, this prediction is used as input to the disease model in order to detect disease occurrence in advance. The predictions for the greenhouse downy mildew were compared using real-time measured data for two months. After several false positive reports, one positive report by both methods fitted the first observation in the greenhouse on April 24, 2021. Thus, a relevant contribution was developed in this topic where the early warning cucumber downy mildew was obtained via coupling climate and disease models, where only transient inputs from weather forecast are required. Regarding the control contributions, first, a selective event-based control approach was proposed to regulate the temperature and the humidity by using the natural ventilation as unique actuator. A temperature PI controller was studied with an event-based approach. Different values with δ = [0, 0.1, 0.2, 0.5, 1] relating to the event occurrence were tested. The results show that δ=0.5 is the optimum value, which significantly reduces the number of vent movements by 43.8%, while only increasing the temperature error by 1.13%. Secondly, comparative studies of humidity controllers (tracing relative humidity, TRH and tracing absolute humidity, TAH) are conducted independently. The results show that TRH performs ideal when the RH set-point is not high. However, the controller lost robustness when RH is over 70%. Comparatively, TAH keeps reasonable robustness with all the RH set-points, but it lacks sensitivity so that the accuracy is lower than TRH method when RH is lower than 60%. Finally, a selective temperature control strategy with a humidity priority control scheme was demonstrated through a simulation study. This control strategy constantly keeps the relative humidity below 80% while controlling the temperature to the set-point, which not only prevents high humidity damaging the crops, but also greatly avoids the loss of energy. For climate control in greenhouses, increasing yield and preventing fungal diseases are contradictory processes, because fungal pathogens and hosts are necessarily stay in the same niche. Therefore, a new hierarchical optimization control strategy was proposed to maintain the optimum diurnal and nocturnal temperature for cucumber production, but giving priority to the disease control. The hierarchical optimization control scheme provides the optimal temperature set-point in each transient step. In the lower layer, the previously developed event-based PID controller keeps the optimum temperature for cucumber production. In the upper layer, an optimizer provides a suggested set-point looking to avoid the ongoing infection. For this, the disease infection model (given by the combination of the greenhouse climate model together the disease model previously discussed) is simulated by a three-day prediction using weather forecast. The new setpoint is calculated by a cost function, which ensures the minimum integration of absolute error between the current set-point and the greenhouse temperature. This novel study is of great significance for precise control of greenhouse fungal diseases. Based on these results, further studies were explored to improve control efficiency. Classical closed-loop control takes the temperature in the center of the greenhouse as the current temperature. This temperature comes from the greenhouse model output or the sensor placed in the center of the greenhouse. However, there are remarkable nonuniformities in leaf microclimate within the canopy in a greenhouse, with implications for variable heat and mass exchange, and the heterogeneity distribution of greenhouse climate. The future closed-loop control may calculate the optimal feedback temperature according to the temperature distribution in each transient step. It may be costly to install many sensors in the greenhouse. So, a practical solution is to develop a distribution model based on the current weather condition and greenhouse structure. CFD technique is one of the powerful tools to achieve this goal. According to the previous reason, the following CFD studies were conducted relating to greenhouse climate models. Two-dimensional and three-dimensional transient CFD models were developed for the temperature and humidity distribution in the greenhouse and the LWD (Leaf Wetness Duration) distribution leading to disease infection. The RMSE of the temperature, relative humidity and leaf temperature during the two nights were 1.24 °C, 3.31%, and 1.32 °C, respectively. The leaf condensation results were manually observed for comparison with the simulated results. Leaf condensation always occurred first in the area near the semi-transparent roof, both in the observations and the simulation. The LWD was simulated by considering the duration of the simulated leaf condensation at each point. The evaluation was conducted on 216 pairs of samples. The True Negative Rate (TNR), True Positive Rate (TPR), and Accuracy (ACC) were 1, 0.66, and 0.89, respectively. This study serves as a reference for an early warning model of disease based on the temporal and spatial distribution of leaf condensation. However, CFD simulation is time-consuming, and the current computing power is insufficient to provide transient feedback for the feedback controller. In the near future, the feedback control based on the transient distribution of temperature and humidity will greatly improve the control efficiency. To conclude, this summary ends by depicting the structure of this document, which has been divided into three parts according to those described in the University of Almería regulation for Ph.D. theses presented in the compendium modality: i. Chapter 1 describes the framework of the thesis and introduces the main methodologies used. In addition, this chapter describes the development structure of the thesis and indicates the publications dealing with each of the topics covered. ii. Chapter 2 presents the scientific publications that support the work done. iii. Chapter 3 summarizes the conclusions derived from the different publications as well as the recommendations for future work.