Distribution Characteristics and Prediction of Temperature and Relative Humidity in a South China Greenhouse

Abstract

South China has a climate characteristic of high temperature and high humidity, and the temperature and relative humidity inside a Venlo greenhouse are higher than those in the atmosphere. This paper studied the effect of ventilation conditions on the spatial and temporal distribution of temperature and relative humidity in a Venlo greenhouse. Two ventilation conditions, with and without a fan-pad system, were studied. A GA + BP neural network was applied to predict the temperature and relative humidity in fan-pad ventilation in the greenhouse. The results show that the temperature in the Venlo greenhouse ranged from 15.8 °C to 48.5 °C, and the relative humidity ranged from 24.9% to 100% during the tomato-planting cycle. The percentage of days when the temperature exceeded 35 °C was 67.3%, and the percentage of days when the average relative humidity exceeded 70% was 83.7%. The maximum temperature differences between the three heights under NV (Natural Ventilation) and FPV (Fan-pad Ventilation) conditions were 3.4 °C and 4.5 °C, respectively. The maximum relative humidity differences between the three heights under NV and FPV conditions were 8.4% and 21.7%, respectively. The maximum temperature difference in the longitudinal section under the FPV conditions was 3.2 °C, while the relative humidity was 11.4%. The cooling efficiency of the fan-pad system ranged from 16.6% to 70.2%. The non-uniform coefficients of the temperature under the FPV conditions were higher than those under the NV conditions, while the nonuniform coefficients of the relative humidity were the highest during the day. The R², MAE, MAPE and RMSE of the temperature-testing model were 0.91, 0.94, 0.11, and 1.33, respectively, while those of relative humidity model were 0.93, 2.83, 0.10, and 3.86, respectively. The results provide a reference for the design and management of Venlo greenhouses in South China.

1. Introduction

Tomatoes are wildly popular for rich lycopene, a carotenoid which is important for its health-related properties [1]. China is one of the main tomato-growing regions in Asia, producing 67,630,000 t of tomatoes in 2021, around one-third of the world’s total yield [2]. The climate has a strong influence on the yield and quality of tomatoes, so many tomatoes are currently grown in greenhouses to reduce the impact on the environment.

Greenhouses were created to improve agricultural production for the growing population and urbanization, tackling the challenges faced by food production and agriculture due to the limited arable land [3]. Greenhouses provide an artificial environment for intensive crop production, and the temperature, humidity, solar radiation, and CO₂ therein can be controlled, protecting the crops from cold weather, storms, and insect pests [4]. China owns more than 90% of the world’s total greenhouse facilities [5], but due to China’s vast territory, the climate and environment in each region are different, resulting in diverse environmental control strategies. Greenhouses in North China need heating and humidification operations because of the low temperature and low humidity in winter, while greenhouses in South China need cooling and dehumidification operations because of the high temperature and high humidity in summer. Many studies have focused on solar energy and thermal insulation in greenhouses in North China, but there is a lack of studies on alleviating hot and humid environments in greenhouses in South China [6].

Temperature and humidity are significant factors in crop growth. Temperature is mainly affected by solar radiation [7], and humidity is mainly affected by the transpiration and evaporation of irrigation water. Lohani et al. proposed that high temperatures significantly impact sexual reproduction in crop plants, affecting seed sets and yields, highlighting the need for heat-tolerant genotypes and high temperature-resilient crops [8]. Additionally, Delgado-Vargas et al. found that high temperature reduces fruit firmness and size, and affects fruit composition [9]. Another factor that affects plant growth and yield is the concentration of water vapor in the air, and over 80% for long periods, especially at night, causes serious problems in crops [10]. Cámara-Zapata et al. suggested that high humidity around crop plants encourages plant diseases and physiological disorders [11]. Temperature and humidity problems are usually solved by ventilation, which includes natural ventilation and mechanical ventilation.

Natural ventilation depends on the wind pressure caused by outdoor wind and the thermal pressure caused by the difference between indoor and outdoor air temperatures; this promotes air flow and, thus, allows for indoor and outdoor air exchange without the use of mechanical systems, thereby playing an effective role in cooling greenhouses [4]. The effect of natural ventilation is far less than the effect of mechanical ventilation. Fan pad ventilation is one of the more effective cooling methods, and its theoretical cooling capacity can reach 8–12 °C [12]. While fan pad systems can slightly increase the relative humidity in a greenhouse, this increase is much lower than that of a sprinkler/fogging system [13]. Researchers have developed a number of mathematical models to simulate the cooling process of pad-fan cooling (PFC) systems, and identified the cooling factors and their mechanisms [14]. However, there is currently a lack of research on the distribution patterns of temperature and humidity under different ventilation methods in South China greenhouses.

Temperature and humidity distributions are complex due to the difference in material properties and boundary conditions [15]. A great number of studies have examined the optimization of the temperature and humidity distributions in greenhouses, as they are beneficial to yield and quality improvement [16,17]. Baeza et al. studied the effect of greenhouse sidewall vents on buoyancy-driven natural ventilation, and found that the contribution of the sidewall vents is important, even for quite large greenhouses [16]. He et al. developed a 3D numerical model and found that the vent position considerably affects microclimate patterns and the distribution and behavior of the indoor temperature and humidity [17]. Lopez et al. compared the effects of a pad-fan system and a fog system in a greenhouse, and they found that the temperature and relative humidity were more stable and uniform when using the fan pad system [18]. However, few studies have emphasized the distribution regularities of temperature and relative humidity when using a fan pad system in South China.

Different computational approaches have been applied to temperature prediction in recent years [19]. Patil et al. built linear regression models by an auto regressive method (AR), an auto regressive moving average method (ARMAX) and a neural network auto regressive method (NNARX) for a greenhouse in Thailand, and the NNARX model performed better than other models [20]. Neural networks methods [21] support vector regression algorithm methods (SVM) [22], and other combination forecasting models [23] have been explored for temperature prediction. The GA + BP neural network timing prediction algorithm is a combination of genetic algorithm (GA) and a back propagation (BP) neural network timing prediction method. It utilizes the global search and optimization capabilities of genetic algorithms and the learning and approximation capabilities of BP neural networks to predict time series data more effectively.

This study investigated the distribution of temperature and humidity in a greenhouse with and without fan-pad systems in South China. The distribution regularities of change with time and space were proposed, and they were evaluated using equations. The GA + BP neural network was applied to predict the temperature and relative humidity in fan-pad ventilation in the greenhouse.

2. Materials and Methods

2.1. Experimental Materials

This study was conducted at the Baiyun Test Base of Guangdong Academy of Agricultural Sciences (113°25′44″ E, 23°23′30″ N), Guangzhou, which is located in South China and belongs to a subtropical monsoon climate. The average annual temperature was 21.7–23.1 °C, and the average relative humidity was 77%; the average annual rainfall was 1435.5 mm, and the average annual sunshine duration was 1897 h in 2021. The maximum and average light intensity in the greenhouse during the experiment from 6:00–18:00 was 59,499 and 26,494 Lux, respectively.

The greenhouse had a length of 30 m, a width of 32 m, and a height of 5.6 m. The Venlo greenhouse used in this experiment was equipped with a fan-pad system, a skylight, a side window, and internal and external sunshade curtains, and it was cooled with the fan-pad system. The length, width, and thickness of the fan-pad system were 16 m, 1.2 m, and 0.2 m, respectively. The greenhouse was equipped with four fans, with a 1.2 m side length and 1.1 kW of power.

Tomatoes (Yuekeda 202) were transplanted in hydroponic troughs with a length of 25 m and a width of 1.5 m after 35 days of cultivation in seedling trays. The fruit of “Yuekeda 202” is characterized by its red fruits, thin peel, crack resistance, and an average fruit weight of approximately 16 g, which was cultivated by the Institute of Facility Agriculture, Guangdong Academy of Agricultural Sciences. Tomato plants were cultivated using a nutrient solution, which was pumped from the nutrient solution reservoir to the front end of the hydroponic trough and flowed back to the reservoir at the rear end of the trough, thus forming a continuous cycle. Standard Hoagland solution was used as the nutrient solution, of which the pH concentration and the EC were between 5.5 and 6.5, and 1.0 and 1.4, respectively. The experiment was conducted from 12 March 2021 (tomato transplanting) to 23 June 2021 (tomato harvesting), over a total of 104 days. The fan pad system was operated every day from 12 March to 17 June, but without the fan pad system over the 18–23 June period. The test greenhouse and the site are shown in Figure 1.

2.2. Experimental Methods

Integrated temperature and relative humidity sensors (Elitech RC-4, −30 °C–60 °C ± 0.1 °C; 0% to 99% RH ± 3% RH, Jiangsu Jinchuang Electric Co., Ltd., Xuzhou, China) were installed in the greenhouse one day before transplanting the tomato seedlings. They were used to record temperature and relative humidity data in the greenhouse; the indoor layout of the integrated sensors is shown in Figure 2. Half of the greenhouse was selected for testing, since the greenhouses were symmetrical. To evaluate the distribution characteristics of the temperature and relative humidity, 5 horizontal sections, 10 longitudinal sections, and 3 height sections in half the greenhouse were selected for monitoring, totaling 150 measurement positions. The horizontal sections were labeled A-E; the longitudinal sections were labeled 1–10; and the vertical sections were labeled H, M, and L. Two additional integrated temperature and humidity sensors were installed outside the greenhouse to measure the outside temperature. The sensor recorded data every 15 min. In the NV condition and FPV condition, the skylight and side windows were operated between 17:00 and 8:00 the following day, and the inner and outer sunshades were operated between 10:00 and 15:00. In the FPV condition, the fan-pad system was operated between 8:00 and 17:00.

2.3. Statistical Analysis

In this study, the time-sequence characteristics of temperature and humidity in the greenhouse over a tomato growth cycle of 104 days were studied. The spatial and temporal distribution characteristics of the day-long temperature and humidity of two conditions were analyzed. Natural ventilation tests were conducted on 21 June, which was the hottest day of the year, and ventilation tests were conducted on 17 June. Changes in indoor temperature and relative humidity under natural ventilation and fan-pad ventilation conditions were studied to reveal the cooling characteristics of the greenhouse in South China. In order to ensure that the external temperature and humidity under the two conditions did not affect the internal environment, the outdoor temperature and humidity when conducting a test of natural ventilation and fan-pad ventilation were taken to carry out a significant difference analysis using a one-way analysis of variance, respectively. The results showed that the p values of temperature and humidity were 0.24 and 0.11, respectively, both greater than 0.05, indicating that the difference between the temperature and relative humidity on the two days was not significant.

To evaluate the spatial distributions of temperature and relative humidity in greenhouses, the non-uniformity coefficient was introduced as an evaluation index [24]. The coefficient of non-uniformity can be expressed by the following:

$$S={\displaystyle \sum _i^n\left|\left(ti-tn\right)/tn\right|}$$

(1)

where, t_i is the temperature (relative humidity) of the i measuring point, °C (%), and t_n is the average temperature (relative humidity) of n measuring points, °C (%). Twenty-seven measuring points (numbered “1”, “5”, “10” in “A”, “C”, “E” columns shown in Figure 1) were selected to evaluate the uniformity of temperature and relative humidity.

The cooling efficiency (η, %), which represents the cooling capacity of the fan-pad system, was determined as the ratio between the drop in air temperature after passing through the wet pad and the maximum drop under air saturation conditions [25].

$$\eta (\%)={\displaystyle \frac{{{\displaystyle \mathrm{T}}}_{\mathrm{odb}}-{{\displaystyle \mathrm{T}}}_{\mathrm{idb}}}{{{\displaystyle \mathrm{T}}}_{\mathrm{odb}}-{{\displaystyle \mathrm{T}}}_{\mathrm{owb}}}}$$

(2)

where, T_odb is the outside dry-bulb temperature of the air entering the cooling pad (°C), T_idb is the dry-bulb temperature of the air leaving the cooling pad (°C), and T_owb is the outside wet-bulb temperature of the air entering the cooling pad. T_odb and T_idb were measured directly using a temperature sensor; among them, T_odb were measured by the sensors numbered T_i (i = 1, 2). T_idb, representing the average temperature of 4 positions numbered P_m (m = 1, 2, 3, and 4), measured the temperature of the air flowing out of the cooling pad. T_owb was calculated using the empirical models developed by Jaramillo et al. [26].

2.4. GA + BP Algorithm Settings

In order to obtain the temperature and relative humidity in the greenhouse for advance control, the GA + BP (Genetic algorithm + Back Propagation Neural Network) algorithm was applied to predict the temperature and humidity in the greenhouse. GA is a method to find the optimal solution by simulating the natural evolution process. The algorithm transforms the problem-solving process into a genetic operation process similar to chromosome selection, crossover, and mutation in biological evolution, and can quickly obtain a better optimal solution. BP has the characteristics of precise optimization, while the genetic algorithm has strong macroscopic search ability and good global optimization performance. Therefore, the combination of genetic algorithm and bp network can achieve the purpose of global optimization, speed, and efficiency by first using the genetic algorithm to search for optimization during training, narrowing the search range, and then using the bp network to solve accurately.

This study is based on the genetic algorithm and the neural network model provided in the neural network toolbox in the MATLAB. The algorithms used in this study for temperature prediction were programmed in MATLAB 2018a on a 2nd Gen Intel(R) Core(TM) i9-12900 computer (Dell Technologies, Xiamen, China). The samples were collected once per hour over two months from March 2021 to May 2021. A total of 1000 data sets were applied, of which the first 700 were used for model training, and the remaining 300 for model testing. The steps of the GA + BP neural network algorithm are as follows [27]:

Step 1: Set the genetic algorithm parameters and perform the population initialization according to the coding method.

Step 2: Establish the neural network in the initial state.

Step 3: Provide a set of samples to the input layer of the network, train the network, and calculate the output error of the neural network.

Step 4: Determine whether all the samples are trained and, if not, return to step 3; if this is satisfied, the total error of the obtained network is used as the fitness function of the genetic algorithm as described, and the value of the fitness function of each individual is calculated.

Step 5: Calculate the crossover probability and mutation probability, and then use the selection, crossover, and mutation genetic operation methods to manipulate the current population to generate a new population genetically.

Step 6: Determine whether the maximum number of evolutionary generations is reached and, if satisfied, move to step 7; if not, build a neural network with the current population and return to step 2.

Step 7: End of genetics. Calculate the fitness function value of each individual in the current population and consider the maximum value as the optimal individual. Decode this individual according to the decoding method to obtain the optimal network structure, initial weights, threshold, and learning rate, and build the optimal neural network.

Step 8: Provide a set of input samples to the input layer of the neural network, train the network, and calculate the output error.

Step 9: Determine whether all the samples are trained; if not, select the following learning sample to provide to the network and return to step 8; if satisfied, move to step 10.

Step 10: Calculate the total error according to the error formula and judge whether the total error satisfies; if it does, end the training; if not, move to step 11.

Step 11: If the network reaches the predetermined number of training, the training is terminated; otherwise, the process returns to step 8 to continue training.

The indexes R², MAE (Mean Absolute Error), MAPE (Mean Absolute Percentage Error), and RMSE (Mean Absolute Error) were used to evaluate the forecasting capacity of the GA + BP models. The indexes were calculated from the following equations:

$${\mathrm{R}}^2={\displaystyle \frac{{\left[{\displaystyle {\sum }_{i=1}^n\left(X_i-\overline{X}\right)}\left(Y_i-\overline{Y}\right)\right]}^2}{{\displaystyle {\sum }_{i=1}^n{\left(X_i-\overline{X}\right)}^2}{\displaystyle {\sum }_{i=1}^n{\left(Y_i-\overline{Y}\right)}^2}}}$$

(3)

$$\mathrm{MAE}={\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n\left|Y_i-X_i\right|}$$

(4)

$$\mathrm{MAPE}={\displaystyle \frac{1}{n}}{\displaystyle \sum _i^n\frac{|X_i-Y_i|}{X_i}}$$

(5)

$$\mathrm{RMSE}=\sqrt{{\displaystyle \frac{1}{n}}{\displaystyle \sum _{i=1}^n{\left(Y_i-X_i\right)}^2}}$$

(6)

where X_i and Y_i represent the measured and predicted at the ith time step, respectively; $\overline{X_i}$ and $\overline{Y_i}$ represent the corresponding mean values; n is the number of data. The closer R² is to 1, the better models perform; MAE, MAPE, and RMSE all range from 0 (perfect fit) to ∞ (the worst fit).

3. Results

3.1. Temporal Variation in Temperature and Relative Humidity

Figure 3 shows the temporal variation in the temperature and humidity inside the greenhouse from March to June 2021. It can be seen in Figure 3a that the temperature in the greenhouse ranged from 15.81 °C to 48.46 °C, from the beginning to the end of the experiment. The percentage of days when the temperature exceeded 35 °C, 40 °C, and 45 °C was 67.3%, 12.5%, and 5.8%, respectively, and the percentage of days when the average temperature exceeded 35 °C was 3.8%.

It can be seen in Figure 3b that the highest relative humidity in the greenhouse was 100%, while the lowest was 24.87%. The relative humidity of most greenhouses is between 40% and 100% [28]. The percentage of days when the relative humidity was below 40% was 10.6%. The percentage of days when the highest relative humidity exceeded 70% and 90% was 100% and 77.9%, respectively, and the percentage of days when the average relative humidity exceeded 70% and 90% was 83.7% and 9.6%, respectively.

3.2. Temporal Characteristics of Greenhouse Environment under Different Ventilation Conditions

3.2.1. Temporal Characteristics of Temperature

Figure 4 shows the temporal distribution of the temperature of the greenhouse under different ventilation conditions. The average temperature of the three heights (H, M, and L) was calculated by taking the mean of the five columns (A~E) of temperature data, which characterize the temperature changes inside the greenhouse over time under the NV conditions (Figure 4a). It can be seen from the figure that the highest temperature reached 47.4 °C, while the lowest temperature was 26.0 °C. The temperature inside the greenhouse started to increase after sunrise at 6:00 AM, it peaked at around 5:00 PM, and then began to decline. The suitable temperature range for most crops is between 30 and 35 °C, so temperatures above 35 °C can be considered as stress conditions for plant growth. Without the cooling systems, the duration of stress was 11.25 h (from 8:00 AM to 7:15 PM), of which the duration of high-temperature stress (45 °C) was 3.75 h (from 12:45 PM to 4:30 PM). The maximum temperature difference between the three heights occurred at 10:30 AM, with a value of 3.4 °C.

Under the FPV conditions, the highest temperature was 38.2 °C, and the lowest temperature was 26.1 °C (Figure 4b). The maximum temperature difference between the three heights was 4.5 °C. After 6:00 AM, the temperature in the greenhouse began to rise, and it began to drop at about 6:00 p.m., during which time the temperature in the greenhouse was maintained within a certain range, at about 30.9 °C–37.5 °C, due to the intervention of the fan-pad system. Under the condition of the fan-pad system cooling, the L height of crops in the greenhouse was subjected to high-temperature stress for 10:00 h (8:00–18:00), while the M height was subjected to high-temperature stress for only 0.75 h (17:15–18:00), and the L height was not subjected to high-temperature stress. After the fan-pad system was turned on, the cooling effect in the greenhouse was obvious, and the maximum cooling capacity was over 10 °C. However, the H height of crops was still subjected to high-temperature stress, so precise regulation of the cooling effect of this height can be carried out in subsequent studies.

3.2.2. Temporal Characteristics of Relative Humidity

Figure 5 shows the temporal distribution of the relative humidity of the greenhouse under different ventilation conditions. The highest relative humidity was 94.8%, and the lowest relative humidity was 33.4% under the NV conditions (Figure 5a). Since the appropriate relative humidity for most crops is 40–70%, relative humidity higher than 70% can be considered as high-humidity stress, and the high-humidity stress duration under the NV conditions was 11.75 h (0:00–8:00 and 20:15–24:00). The maximum relative humidity difference between the three heights was 8.38%.

It can be seen in Figure 5b that the relative humidity in each height changes with time in a consistent manner. The maximum relative humidity difference between the three heights was 21.7%. From 0:00 to 6:00, the relative humidity of each height remained at a high value, and there was no significant difference. After 7:30, the relative humidity began to decrease and maintained a small fluctuation for a period of time. The relative humidity began to rise after 17:45 and returned to a stable state after 22:00. The stress duration of each height was different. The stress duration of the L and M heights was 24 h, and that of the H height was 13.75 h in a day. A comparison of the stress duration under the natural ventilation conditions showed that dehumidification is necessary to avoid crop yield reduction due to high-humidity stress when the fan-pad system is in operation. As can be seen in the figure, when the fan-pad system was in operation, the highest relative humidity in the greenhouse was 99.5%, and the lowest relative humidity was 55.5%, which was 22.1% higher than that under the natural ventilation conditions.

3.3. Spatial Characteristics of Greenhouse Environment under Different Ventilation Conditions

3.3.1. Spatial Characteristics of Temperature

Figure 6 shows the spatial distribution of the temperature in the greenhouse. It can be seen from the figure that the temperature differences between the heights during the time periods of 0:00–6:00 and 18:00–24:00 under the NV conditions were not significant, with a maximum temperature difference of 2 °C (Figure 6a). However, during the time period of 18:00–24:00, the temperature in the middle position of the front part of the greenhouse was lower, which may be due to its proximity to the fan-pad system and the influence of natural wind from outside, which would result in a lower temperature in this position than in other positions of the same height. The time period of 12:00–18:00 showed a significant temperature difference between the upper, middle, and lower heights, with a maximum temperature difference of 5.13 °C between the H and L heights at the right end of the front part.

It can be seen in Figure 6b that the highest temperatures in the four periods under the FPV conditions were 27.0 °C, 35.7 °C, 40.1 °C, and 30.8 °C. The lowest temperatures were 26.2 °C, 29.2 °C, 30.9 °C, and 28.5 °C, respectively. The maximum temperature differences were 0.8 °C, 6.5 °C, 9.2 °C, and 2.3 °C, respectively. The temperature difference in the 0:00–6:00 period was the smallest, and the temperature difference in the 12:00–18:00 period was the largest.

3.3.2. Spatial Characteristics of Relative Humidity

Figure 7 shows the spatial distribution of the relative humidity in the greenhouse. It can be seen from the figure that the distribution of the relative humidity in different time periods was inconsistent under the NV conditions. The differences in the relative humidity between heights were more obvious in the 0:00–6:00, 6:00–12:00, and 12:00–18:00 periods. Overall, the difference in the relative humidity in the upper, middle, and lower heights during the 18:00–24:00 period was small, but the relative humidity was high near the pad, which may be due to the natural evaporation and diffusion of water on the pad. The highest relative humidity was 86.4% in the 0:00–6:00 period, and the lowest relative humidity was 81.9%, with a difference of 4.5%. From 6:00 to 12:00, the highest relative humidity was 70.4%, and the lowest relative humidity was 51.4%, with a difference of 19.0%. The highest relative humidity was 45.8% in the 12:00–18:00 period, and the lowest relative humidity was 35.4%, with a difference of 10.4%. The highest relative humidity was 84.3% in the 18:00–24:00 period, and the lowest relative humidity was 71.9%, with a difference of 12.4%. It could be seen that the difference in the relative humidity in the 0:00–6:00 period was the smallest, and the difference in the 6:00–12:00 period was the largest.

It was found that the highest relative humidity under the FPV conditions was 100%, 92.7%, 84.5%, and 93.2%, and the lowest relative humidity was 97.2%, 67.7%, 49.0%, and 74.5%, respectively. The maximum difference in the relative humidity between each height was 2.8%, 25.0%, 35.5%, and 18.7%, respectively. The 0:00–6:00 period showed the smallest difference in relative humidity, while the 12:00–18:00 period showed the maximum difference in relative humidity.

3.4. Longitudinal Distribution of Greenhouse Environment under Different Ventilation Conditions

3.4.1. Longitudinal Distribution of Temperature

Table 1. The longitudinal distribution of temperature.

Time	Natural Ventilation (NV)		Fan-Pad Ventilation (FPV)
	Fitting Equation	R2	Fitting Equation	R2
0:00–6:00	y = −0.0021x + 29.885	0.01	y = −0.0286x + 26.658	0.69
6:00–12:00	y = −0.033x + 36.702	0.07	y = −0.2098x + 33.365	0.75
12:00–18:00	y = −0.0884x + 43.91	0.61	y = −0.3702x + 36.829	0.90
18:00–24:00	y = −0.0491x + 31.266	0.03	y = −0.09x + 30.31	0.91

shows the longitudinal distribution of the relative humidity in the greenhouse. No trend was evident in the temperature in the different periods under the NV conditions, while the temperature showed a great fit to the first-order equation under the FPV conditions, and the R² values in the different periods were 0.69, 0.75, 0.90, and 0.91. The temperature differences in the longitudinal section under the FPV conditions in the four periods were 0.2 °C, 1.9 °C, 3.2 °C, and 0.8 °C, respectively.

3.4.2. Longitudinal Distribution of Relative Humidity

Table 2. The longitudinal distribution of relative humidity.

Time	Natural Ventilation (NV)		Fan-Pad Ventilation (FPV)
	Fitting Equation	R2	Fitting Equation	R2
0:00–6:00	y = −0.0793x + 84.003	0.55	y = 0.0776x + 98.549	0.36
6:00–12:00	y = 0.0178x + 64.002	0.002	y = 0.8898x + 77.252	0.79
12:00–18:00	y = 0.0634x + 39.813	0.16	y = 1.3329x + 62.085	0.89
18:00–24:00	y = −0.0328x + 76.24	0.02	y = 0.2749x + 84.841	0.80

shows the longitudinal distribution of the relative humidity in the greenhouse. It can be seen from the table that the trends of the relative humidity in different periods under the NV conditions and in the 0:00–6:00 period under the FPV conditions were not significant. The relative humidity under the FPV conditions showed a great fit to the first-order equation in the 6:00–24:00 period, and the R² values in three periods were 0.79, 0.89, and 0.80. The relative humidity differences in the longitudinal section under the FPV conditions in the four periods were 0.1%, 7.7%, 11.4%, and 2.1%.

3.5. Uniformity of Temperature and Relative Humidity in Greenhouse

Table 3. Non-uniform coefficients of temperature and relative humidity.

Item		Time	0:00–6:00	6:00–12:00	12:00–18:00	18:00–24:00
	Pattern
Temperature	NV		0.146	1.163	1.448	0.446
	FPV		0.122	0.704	0.841	0.313
Relative humidity	NV		0.166	2.006	3.212	1.017
	FPV		0.226	0.997	1.592	0.532

shows the uniformity of the temperature and relative humidity under the two conditions, which indicates that larger value represents less uniformity. It can be seen from the table that the non-uniform coefficients of the temperature increased from 0:00 to 18:00 but decreased from 18:00 to 24:00. Additionally, the non-uniform coefficients of the temperature under the FPV conditions were higher than those under the NV conditions in the same period. It can be seen from the table that the trend of the non-uniform coefficients of the relative humidity were similar to those of the temperature. The non-uniform coefficients of the relative humidity under the FPV conditions were higher than those under the NV conditions in the last three periods.

3.6. Cooling Efficiency of Cooling Pad

Figure 8 shows the cooling efficiency of the fan-pad system during the day. It can be seen from the table that the cooling efficiency first increased and then decreased, which could be approximated using quadratic functions, revealing a R² of 0.87. The efficiency at different times was 17.70%, 67.89%, 70.22%, 60.73%, 36.40%, and 16.61%, and the average cooling efficiency was 44.92%. The maximum efficiency was 70.22% at 11:00, and the minimum efficiency was 16.61% at 18:30.

3.7. Prediction Results

Figure 9 shows the prediction results of temperature and relative humidity in the fan-pad ventilation condition. The training model was established for 700 samples, and the testing model was established for 300 samples. The R² of the training model and the testing model for temperature were 0.92 and 0.91, while those of relative humidity were 0.94 and 0.93 (

Table 4. Accuracy evaluation of prediction model.

Item	Samples	R2	MAE	MAPE	RMSE
Temperature	Training	0.92	0.78	0.004	1.18
	Testing	0.91	0.94	0.011	1.33
Relative humidity	Training	0.94	2.30	0.004	3.09
	Testing	0.93	2.83	0.010	3.86

). It can be seen from the table that the R² of the relative humidity model was higher than that of the temperature model, regardless of the training model or the testing model. The MAE, MAPE, and RMSE of the temperature testing model were 0.94, 0.11, and 1.33, respectively, while those of the relative humidity model were 2.83, 0.10, and 3.86, respectively. The results showed that it is possible to predict accurately the temperature and relative humidity up to an hour into the future.

4. Discussion

Greenhouses have become pivotal in modern agriculture due to their ability to control environmental conditions, extend growing seasons, and enhance crop yields. It is important to overcome the high temperature and high humidity stress inside the greenhouse under the high temperature and high humidity climate in South China. The results revealed a considerable difference between the natural ventilation and the fan-pad ventilation conditions. The results of Figure 3a indicate that there were strong high-temperature periods in the greenhouse, and it achieved a better temperature distribution after fan-pad ventilation. Efficient ways to reduce the temperature in greenhouses include the use of forced ventilation cooling and fog/mist systems [29], but these systems, except fan-pad ventilation, consume a lot of energy, driving researchers to find the ideal reduced-cost solution [30]. Besides temperature, this paper also studied the relative humidity in a tomato-planting cycle. There was generally high-humidity stress in the greenhouse (Figure 3b), which is conducive to the development of fungal and bacterial infections [31]. Therefore, dehumidification should be considered to achieve a better relative humidity environment.

Due to the influence of buoyancy, natural convection occurred in the air inside the greenhouse, resulting in significant temperature fluctuations (Figure 4a). At night, when the temperature inside the greenhouse was lower, natural convection weakened, resulting in a more stable temperature with smaller fluctuations. The temperature in the greenhouse dropped significantly after using the fan-pad system. The trend over the day was similar to that in earlier studies [32,33].

The relative humidity was higher during the night than during the day (Figure 5), so it is practicable to reduce the water vapor content inside the greenhouse via dehumidification at night. This can accommodate more low-temperature and high-humidity air from the wet pad during the day and reduce the air humidity inside the greenhouse, thereby delaying the increase in the temperature rate during the day. The reason for the more uniform relative humidity at night was the closure of the fans [33]. By comparing the relative humidity distribution of the two conditions, it was found that natural ventilation had little potential to provide the optimal relative humidity for tomatoes [30]. As can be seen in Figure 5b, the relative humidity in the greenhouse at night was high, so it must be reduced to inhibit the development of gray mold and other fungal diseases [34].

These results of spatial temperature distribution (Figure 6a) are similar to those of some studies [35] but different from those of other studies [36]. There was a minimum temperature difference of 0.3 °C between the H and L heights. During the 6:00–18:00 period, the greenhouse was exposed to the maximum solar radiation, and the blockage of plant branches and leaves resulted in the occurrence of a significant temperature stratification phenomenon inside the greenhouse, showing a gradual decrease in the temperature from top to bottom. The reason for the temperature difference in the 12:00–18:00 period being larger than that in the 0:00–6:00 period (Figure 6b) may be because there was no radiation at night, and the overall temperature in greenhouses does not change much, so the temperature difference was smaller. In the afternoon, the outside radiation was higher, coupled with the blocking of the branches and leaves of each height of crops, resulting in a large difference in the temperature of each height. Studies have shown that temperature stratification is a common problem in most greenhouses, and the temperature distribution varies greatly [33,37].

The difference in relative humidity in the 6:00–12:00 period was larger than that in the 0:00–6:00 period (Figure 7a). It may be that the heat radiation absorbed during the day was balanced at night, and there was no external heat and humidity intervention, so the spatial distribution of the internal relative humidity was uniform. The relative humidity showed a significant difference in the morning because of the temperature fluctuations caused by the reception of external radiation and the opening of the fan-pad system, coupled with the air circulation blind area, crop blocking, and other factors, thus aggravating the phenomenon of the vertical distribution of relative humidity. The difference of relative humidity was larger in the 12:00–18:00 period than in the 0:00–6:00 period (Figure 7b). The reason for this may be that there was no external radiation at night, and the temperature was stable. In addition, the fan-pad system was not opened, so the water vapor inside the greenhouse was diffused through the gradient to form a more uniform humidity distribution. The fan-pad system was opened in the morning to cool down, and, in the afternoon, there was a lot of water vapor all over the greenhouse. The relative humidity difference in each height was mostly due to the large temperature difference in each height. The results are similar to those from the research conducted by Deng et al. [38].

According to the length of the experimental greenhouse, it was found that the cooling efficiency of the fan-pad system in South China was 0.06 °C/m to 0.11 °C/m (Figure 8). The fan-pad system played an important role in the cooling of the greenhouse, but the temperature difference between the front and back of the greenhouse was too large to achieve temperature uniformity, so it is necessary to install a spray device in the rear of the greenhouse for supplementary cooling. The relative humidity difference in the longitudinal section was significant during the day, and it was caused by the fan-pad system, which brought steam into the greenhouse.

The maximum cooling efficiency could reach up to 70.22% in the South China greenhouse (Figure 9). The results seem to be consistent with those of previous studies, which reported values of 80% [39], 78–92% [40], 79–86% [41], 65% [42], and 57–77% [43]. It might be an effective method for improving the efficiency of cooling the water used in the pad system, which would increase the difference between the water temperature and the outside temperature.

The GA + BP algorithm could accurately predict the temperature and relative humidity (Table 4). Such forecasts could be used by farmers as appropriate advance warnings for changes in temperatures so that they could apply preventative measures to avoid damage caused by extreme temperatures [44]. The greenhouse is a complex system, which involves outdoor temperature and humidity, solar radiation, soil temperature and humidity, and crops. Therefore, it is necessary to understand the interaction between various environmental factors and crops in order to optimize the accuracy of the prediction model.

5. Conclusions

This research studied the temperature and relative humidity distribution in a Venlo greenhouse in South China. Two ventilation conditions, with and without a fan-pad system, were applied to investigate the spatial and temporal distribution. The cooling efficiency was adapted to evaluate the performance of the fan-pad system. A GA + BP neural network was applied to predict the temperature and relative humidity in fan-pad ventilation in the greenhouse. The conclusions can be summarized as follows:

(1) The temperature in the greenhouse ranged from 15.8 °C to 48.5 °C, and the relative humidity ranged from 24.9% to 100%. The maximum temperature and relative humidity differences between the three heights under FPV conditions were 4.5 °C and 21.7%, respectively;

(2) The maximum temperature difference in the longitudinal section under the FPV conditions was 3.2 °C, while the relative humidity was 11.4%, and the cooling efficiency of the fan-pad system ranged from 16.6% to 70.2%;

(3) The R², MAE, MAPE, and RMSE of the temperature testing model were 0.91, 0.94, 0.11, and 1.33, respectively, while those of the relative humidity model were 0.93, 2.83, 0.10, and 3.86, respectively.

The results provide a reference for the better understanding of the temperature and relative humidity distribution in Venlo greenhouses with/without a fan-pad system in South China, and the greenhouse structure will be optimized in future studies.