The research greenhouse used was built in the Faculty of Agriculture of Tarbiat Modares University, Tehran, Iran. The geographic coordinates of the greenhouse are 51°10’ E and 35°44’ N at an elevation of 1369 m. The region has a minimum temperature of -12°C, a maximum temperature of 43°C, and the maximum wind velocity is 90 km/h. The specifications and components of the greenhouse are shown in Table S1 [suppl] and Fig. S1 [suppl].

Fig. 1 shows the toolbars used in the greenhouse. The required equipment was installed according to the calculations made regarding the thermal load calculation, cooling and ventilation. For the research greenhouse, with an area of 40 m² (a space with a volume of 100 m³), the cooling and thermal loads of cucumber were considered. The capacity of the heating system to meet the heating needs of the greenhouse on the coldest night of the year was calculated to be 11.62 kW. The system used to supply the required heat includes a wall gas boiler with an input heating capacity of 26 kW, a thermal efficiency of 91%, a maximum heating circuit water temperature of 80°C, and a hot water unit heater, with a heating capacity of 18.46 to 24.03 kW. The unit heater involved three modes: off, slow (HS), and fast (HF). Gas boiler temperature, which was automatically controlled by a stepper motor, varied between 40°C and 80°C. The input temperature values of the unit heater were T₁=40, T₂=60, and T₃=80.

The cooling system in the research greenhouse under investigation was a cooling pad system. The parameters calculated in the design of the system included: exchanged air volume, that is, air flow intensity in m³ min^-1 for each unit of the greenhouse area, available and desirable greenhouse temperature, the locations of the air handlers, and the distances between them and the pads.

Ventilation is the process of replacing indoor air with outdoor air and is done for the three main purposes: 1) temperature control, 2) humidity control (reduction), and 3) stabilization and adjustment of the CO₂ required by the plant. Maximum ventilation is required in summer when ventilation should generally be done every minute. In the winter season, the above amount reduces to a maximum of 20-30% of greenhouse air volume per minute. A fan with an on and off mode was used for the research greenhouse ventilation system.

In this research, the purpose of using the greenhouse misting system is cooling and increasing the humidity of the greenhouse. The system consists of two rows of foggers, installed between the cultivated rows. Each row involves three nozzles at 2-m intervals. A PR44 electro pump with a power of 0.5 hp, a pumping speed of 2800 rpm, and a flow rate of 35 L min^-1, and a PC-19A automatic water flow pump switch were used to start the pump for water consumption (Arbel et al., 2003).

The amount of available CO₂ in the air is 0.03%, that is about 300 ppm, which is enough for photosynthesis. This value reaches its minimum in winter when the greenhouse is blocked from the outside environment. The optimal range for the amount of carbon dioxide is 500 to 800 ppm. If necessary, the boiler exhaust system directs the heating gas to the greenhouse using a servo motor.

The sensors used in the construction of the system include temperature and humidity, a sensor to measure the light inside and outside the greenhouse, and another sensor to measure carbon dioxide (Fig. 1). According to Fig. 1, sensors T₁ to T4 measure the temperature inside the greenhouse, T5 outside, H1 to H4 measure the humidity inside the greenhouse, and H5 measures outside. DHT22 sensors were used for temperature and humidity. The DHT22 sensor, also known as the AM2302, is one of the most popular temperature and humidity sensors. Humidity measurement was from 0% to 100%, humidity resolution from 2% to 5%, and humidity response time is 2 sec. Temperature measurement was from -40°C to 125°C, with a resolution of ±0.5°C and a response time of 2 sec.

The K30 sensor (Sense Air, Sweden) was used to measure the amount of CO₂ in the air. The measurement range was from 0 to 10000 ppm, the measurement accuracy was ±30 ppm ±3% of the reading, the operation temperature from 0°C to 50°C, the relative humidity from 5% to 95%, and the analog output was linear from 0 to 5 volts. In the present study, the CO₂ sensor was placed inside a box along with a fan for uniform airflow.

The GY-302 light sensor module was used to measure the light intensity inside and outside the greenhouse. GY-302 light data output, ranged from 0 lx to 65535 lx. The L_in and L_out sensors were installed inside and outside the greenhouse to measure solar radiation (Fig. S1 [suppl]). The sensor was calibrated and evaluated before use.

Inside the greenhouse is a Climate Control Center which contains a control board system, computer (data storage), irrigation timer, voltage stabilizer (to prevent power fluctuations), electrical panel control system, and ups (to store electricity in case of shutdown) (Fig. S2 [suppl]). All sensors data were saved every five seconds in text and Excel files (Fig. S3 [suppl]). A part of the display was dedicated to the light adjustment point to identify day and night. In addition, the coded program had the capability of automatic and manual fuzzy implementation.

The control system board consists of the power supply board, sensor input board, relay board, and main processor board (Fig. S4 [suppl]). First, it was coded in C++ and then simulated and built with Proteus.

To establish a control strategy for each of the control variables, a mathematical model of changes of that variable in relation to other variables should be created (Ursem et al., 2002). So far, many mathematical models for climate greenhouse have been proposed. With the help of these models, it is possible to identify the variables affecting the control factors and simulate or control the greenhouse environment. The general model of greenhouse indoor temperature and humidity equations (1-4), obtained by Lafont & Balmat (2002, 2004), was used for the objective or cost function in the greenhouse.

where T_ai: indoor air temperature (ºC); T_ao: outdoor air temperature (ºC); T_v: ceiling angle degree (ºC); C_h: thermal power system (kW); R_g: global radiation (kW m^-2); X_ai: absolute humidity inside the greenhouse (g kg^-1); T_ao: absolute humidity outside the greenhouse (g kg^-1); a_i: model parameters for temperature; b_i: model parameters for temperature; α₅, _β5: unmeasured disturbances; ∆X_ai: lack of humidity (g kg^-1); X_sat: saturated humidity (g kg^-1); H_ai: relative humidity inside the greenhouse (g kg^-1); H_ao: relative humidity outside the greenhouse (g kg^-1).

To validate the model, the control system was placed in the classical control mode and perturbation conditions were considered.

The greenhouse environmental conditions, including temperature, humidity, and CO₂ were controlled. The design of a fuzzy controller starts with the selection of linguistic variables, process state, and input and output variables. The next step is to choose a set of linguistic rules and the type of fuzzy inference process. The rules are set once, and the fuzzy set and output value must be specified after the inference, along with the development of a non-fuzzy strategy (Fig. S5 [suppl]).

Some of the most important topics in fuzzy control are the choice of rules, the fuzzy inference method, membership functions, number of input and output fuzzy sets and their degree of overlapping, implication, connection operations, and defuzzification method (Hellendoorn & Thomas, 1993; Banakar & Azeem, 2011). Fuzzification is the process of transforming a crisp input value into a fuzzy value that uses the information in the knowledge base. Various types of functions such as Gaussian, triangular, and trapezoidal MFs are the most commonly used in the fuzzification process. These types of MFs can easily implemented by embedded controllers. In order to fine-tune the performance of a FLC, these parameters, or the shape of the MFs, can be adapted (Kayacan & Khanesar, 2016). In this study, triangular membership functions were defined for input variables and output variables.

In the first step of the fuzzy control design with hydroponic cucumber culture, the information related to temperature, i.e. desired, maximum, and minimum temperatures, was extracted using the plant growth chart (Janoudi et al., 1993). The defuzzification method is the classical center of gravity method with the Mamdani fuzzy inference system.

Cucumber temperature and humidity during the day: The optimal temperature for cucumber growth during the day was 25°C. According to the growth rate curves, the fuzzy ranges of plant temperature during the day were defined as follows: very low (<15°C, temperature very cold, TVCOLD), low (15-20°C, temperature cold, TCOLD), rather low (20-24°C, temperature cool, TCOOL), desirable (24-26°C, temperature good, TGOOD), rather high (26-30°C, temperature slightly hot, TSH), high (30-35°C, temperature hot, TH), and very high (> 35°C, temperature very hot, TVH).

Temperature required by the cucumber during the night: The optimal temperature during the nighy was 20°C. According to the growth rate curves, the fuzzy ranges of plant temperature during the night were defined as follows: very low (< 10°C, TVCOLD), low (10-15°C, TCOLD), rather low (15-19°C, TCOOL), desirable (19-21°C, TGOOD), rather high (21-25°C, TSH), high (25-30°C, TH), and very high (> 30°C, TVH).

Humidity required by the cucumber: The best humidity was 75% and its range included as follow: very low (< 60%, humidity very low, HVL), low (60%-70%, humidity low, HL), desirable (70%-80%, humidity good, HGOOD), high (80%-90%, humidity high, HH), and very high (> 90%, humidity very high, HVH).

Carbon dioxide required by the cucumber: The optimum CO₂ value was between 400 to 1000 ppm and its ranges were defined as: low (< 350 ppm, CL), desirable (350-1000 ppm, CGOOD), high (> 1000 ppm, CH).

A fuzzy rule consists of two parts: the antecedent and the consequence. System inputs are environmental conditions that enter the fuzzy rule base after fuzzification. The if part of a rule describes the conditions for which it is defined, and the then part describes the response provided by the fuzzy system in those conditions. Considering the number of inputs, the number of membership functions, and the constraints associated with the greenhouse, the fuzzy basis contains 210 rules for day and night. Since the temperatures required by the plant during the day and night were different, the fuzzy control rules were varied. The obtained rules in this research were obtained given in the cucumber growth chart. An example of the fuzzy control rule appears below:

Example: If (Temperature is TVH) and (Humidity is HVH) and (CO₂ is CL) then (FAN is ON) (PAD is ON) (HEATER is OFF) (PACKGE is T₃) (FOGSYSTEM is OFF) (CO₂INJECTION is ON).

GA maintains a set of candidate solutions called population and iteratively refines them by reproduction, crossover, and mutation operators (Lammari et al., 2020). In each generation, the list of responses are evaluated by the objective function and the best result are reported. Linguistic parameters of input temperature and humidity were generated and entered into the new fuzzy system and determined its output values. The time required to reach the points set by Eqs. (1) to (4) was calculated as an objective or fitness function. The process continued until the desired population size or minimum temperature or humidity error was reached. Finally, the best parameters of the membership function were introduced by the GA, which showed the lowest error of the control parameter regarding the desired values. The GA appears in Table 1.

Cucumber (Cucumis sativus L.) is one of the most significant greenhouse crops in Iran. The development of new methods such as soilless culture to increase its quality can play an important role in the productivity of greenhouses involved in its production. The hybrid greenhouse cucumber seed variety Gavrish Karim F1, which was used in this research, is suitable for soil cultivation and hydroponics. Coco peat and perlite were used for hydroponic cultivation of cucumber pots (60:40% of coco peat:perlite, v:v), where 36 pots were placed in three rows. The rows were one meter apart, and the adjacent pots on the same row were half a meter away. Nutrient solution drip systems were used in hydroponics for growing, where the nutrient was held in a separate tank, and applied as pre-programed by the control system.

Fig. 2a shows the changes in temperature inside and outside the greenhouse on 20 June 2019 for the system before optimization. The average temperatures inside and outside the greenhouse were 28.2°C and 35.8°C during the day and 22.2°C and 30°C during the night, respectively. According to the chart, large differences can be seen between the day and the night. To evaluate the heating and cooling system, it is necessary to record the temperature inside and outside the greenhouse. The greater the temperature difference, the greater the heat losses, so the greenhouse consumes more energy. Also, to optimize energy, it is necessary to have the temperature inside and outside the greenhouse.

Fig. 2b shows the changes in temperature inside and outside the greenhouse on 23 June 2019 for the system after optimization. The average temperatures inside and outside the greenhouse were obtained as 25.7°C and 34.2°C during the day, and 20.8°C and 28°C during the night, respectively. Comparison of Figs. 2a and 2b showed that the optimized mode includes the minimum difference between the greenhouse temperature and the desired temperature.

The performance specifications of a control system are often expressed in terms of the transient response to a step input, because the input can be easily generated, and is sufficient for a significant effect. The transient response of the system to a unit step input depends on the initial conditions. To facilitate comparison of the transient responses of different systems, the standard initial conditions are usually used where the system is initially at rest, and the output and all its derivatives are zero. In that case, the specifications of the response can be easily compared. In practice, the transient response of a control system often undergoes vibration damping before reaching a steady state (Ogata & Yang, 2010).

Figs. 2c & 2d show the performance of the conventional and optimized fuzzy control systems in response to step input. Considering that the optimal temperature of the greenhouse during the day (25°C) was different from the temperature at night (20°C), the step input was considered equal to the temperature difference (5°C). The performance characteristics of control systems in terms of transient response to step input are shown in Table 2. The optimized controller tracked the set point faster than the conventional controller.

The temperature increases during the day due to the sunlight, and the plants can reach their optimal conditions only if the control system works properly. The maximum temperature during the day in the greenhouse is of great importance, because as the temperature increases, evaporation and transpiration in the plant increases, and the plant undergoes thermal stress, which affects the plant’s performance. As seen in Fig. 3a, the recorded average temperature of the greenhouse in the fuzzy control system was higher in the pre-optimization period than in the post-optimization period. The results showed that the average temperature of the greenhouse during the days of cucumber cultivation was 34.25°C in the conventional fuzzy control system and 31.17°C in the optimized system.

As the air outside the greenhouse cools during the night, the temperature inside greenhouse decreases and only if the heating system works properly, the plants can reach their optimal conditions. The minimum temperature during the night in the greenhouse is very important because the plant undergoes cold stress when the temperature decreases, which is very important during the cold nights of the year. Fig. 3b shows the average recorded temperature inside and outside the greenhouse during the nights of cucumber cultivation. The results showed that the average temperature of the greenhouse during the night was 23.22°C in the conventional fuzzy control system and 21.96°C in the optimized system.

Fig. 2e shows the humidity changes inside and outside the greenhouse on 20 June 2019 for the system before optimization. The coefficients of daily and night changes of humidity variation in that state were obtained as 11.1% and 10.46%, respectively. A big difference in humidity was observed between the inside of the greenhouse and the outside. According to the proposed fuzzy rules, the misting system was turn on when the temperature in the greenhouse was too high, or the humidity was too low. Due to the extremely hot summer weather, the misting system was also turned on to help the fan and pad systems. The average humidity inside the greenhouse during the day and night was 51.35% and 68.57%, respectively, and the average outdoor air humidity was 22.7% during the day and 30.94% at night.

Fig. 2f shows the humidity changes inside and outside the greenhouse on June 23 2019 for the system after optimization. The coefficients of daily and nightly changes of humidity, in that state, were 19.82% and 26.62%, respectively. A big difference in humidity was observed between the inside of the greenhouse and the outside. The average humidity inside the greenhouse during the day and night was 60.18% and 65.7%, respectively, and the average humidity outside was 24.77% during the day and 26.3% at night.

The temperature increases during the day due to sunlight, and the cooling system (including fan and pad) works for the plant to reach the desired conditions. When the fan is on, the humidity is removed from the greenhouse, which is compensated by the fogger. In addition, the control system becomes complex because the control parameters influence each other, in which case prioritization must be applied between them. In the greenhouse, the priority is to regulate the temperature, the humidity is the second, and the amount of carbon dioxide is the last. Fig. 3c shows the average humidity recorded inside the greenhouse during the days of cucumber cultivation. The results showed that the greenhouse humidity recorded in the conventional fuzzy control system was lower than the humidity obtained in the period after the optimization of the system. The average humidity recorded inside the greenhouse in the conventional fuzzy control system was 39.4%, which reached 60.22% in the optimized system with GA. During the night, due to lack of photosynthesis and reduction of evaporation and transpiration in the plant, the humidity of the greenhouse increases. As can be seen in Fig. 3d, the greenhouse humidity recorded in the fuzzy control system in the period before optimization was lower than the greenhouse humidity obtained after. The average humidity recorded inside the greenhouse was 56.5% in the conventional fuzzy control system and 74.59% in the optimized system with GA.

In the research conducted by Faouzi et al. (2017), the simulation results of a fuzzy control system in a greenhouse in a dry region showed that the highest temperature values in spring and summer were in the range of 20°C to 28°C. Robles et al. (2017) designed and implemented a low-cost system for remote monitoring and control of a greenhouse using fuzzy logic. The method included an Arduino board with a fuzzy algorithm for ambient temperature, soil moisture, relative air humidity, and greenhouse light control. The results of the fuzzy system showed that the temperature was between 28°C and 29°C, and the relative humidity was 45%. Chen et al. (2011) used the proportional integral derivative (PID) and fuzzy method to control a cherry tomato greenhouse with an optimal temperature of 26.5°C and an optimal humidity of 85%. The average daily temperatures using fuzzy and PID controllers were 27°C and 28.2°C, respectively, and the humidity was 80.2% with the fuzzy controller and 78.5% with the PID controller. Guerbaoui et al. (2013) presented a control system based on fuzzy logic to adjust climate parameters in a greenhouse using LabVIEW software to control temperature and humidity. The control system was evaluated for three days in December. The evaluation results showed that the overall performance of the fuzzy controller was satisfactory for temperature and humidity control within the specified range. The temperature outside the greenhouse varied between 8°C and 18°C. The control system maintained the temperature inside the greenhouse at around 21°C and mean relative humidity between 47% and 55%. Azaza et al. (2014) used a fuzzy control system to control temperature and humidity in a greenhouse in summer using meteorological data. The design was simulated in MATLAB/Simulink. The results of the control system showed that the response time of the system was 40 min while sampling rate was 10 min. The temperature varied between 18°C and 35°C, and humidity between 65% and 75% with fuzzy control. Using fan and pad cooling systems, Ganguly & Ghosh (2007) recorded temperature difference of 6°C between the inside and outside of the greenhouse during peak summer sunlight. Also, they recorded the maximum temperature of 29.5°C inside the greenhouse in summer in an ambient temperature of 35.85°C, a relative humidity of 50%, and a shading level of 75%. Lammari et al. (2020) optimized a proportional integral (PI) controller in greenhouse climate control by GA. They emphasized that the optimized controller had better energy saving parameters. The comparison of the present study with the optimized PI controller, shows that temperature and humidity fluctuations were much lower in the optimized fuzzy system.

Fig. 2g shows the CO₂ changes on June 20, 2019 for the system before optimization. The coefficient of variation in the amount of carbon dioxide in that state was equal to 6.2. As can be seen, there were large changes in CO₂ at night, when the carbon dioxide injection system did not work due to the complete absence of photosynthesis. Fig. 2h shows the changes in CO₂ on June 23, 2019 for the system after optimization. The coefficient of variation of CO₂ in that state was equal 2.2. The average amount of carbon dioxide was recorded as 574 ppm. There were lower changes in the amount of CO₂ than in the normal fuzzy control mode.

The optimal amount of CO₂ inside the greenhouse in this research was considered as 500 ppm, which was almost constant by the control system. The value was 510.66 ppm in the conventional fuzzy control system and 514 ppm in the optimized system.

Kläring et al. (2007) investigated the effect of CO₂ on cucumber. The obtained results indicated an increase in CO₂ density to 400 g kg^-1, which increased yield. The difference between CO₂ –supply and non-CO₂–supplied greenhouses is in photosynthesis. The efficiency of CO₂ supply was maximum in moderate light and decreased as air temperature outside the greenhouse increased due to the ventilation required in strong light. Chen et al. (2011) investigated greenhouse environmental factors and obtained the average light intensity and CO₂ concentration in the fuzzy controller during the day at 18,900 lux and 617 mg L^-1, respectively.

One of the most important goals of this research was to obtain the yield of hydroponic cucumber, which was measured weekly during the period. Fig. 4 shows the yield during the harvest period: the mean yield was 5.49 kg pot^-1 and the yield per unit area was 24.4 kg m^-2 (i.e. 244,000 kg ha^-1). Alomran and & Luki (2012) obtained yields of cucumber in greenhouse of 19.49, 5.18, and 15.07 kg m^-2 in summer, fall, and winter, respectively. Mao et al. (2003) obtained a maximum yield of 193,999 kg ha^-1 in greenhouse cucumber.

Greenhouse producers usually face many problems in the optimal use of energy used in greenhouses, and energy carriers without subsidies have doubled their economic problems. These cases mainly include the use of outdated heating systems with very low efficiency, improper distribution, unbalanced heat production on the greenhouse, and heat energy waste in different ways. In the current research, the amount of energy consumed during the period of hydroponic cucumber cultivation was calculated by daily measurements. Total electrical energy consumption during the period was 11.912 kWh. The average daily energy consumption was 0.098 kWh in the conventional fuzzy control system and 0.094 kWh in the optimized system by GA (Fig. 5a). Mohammadi & Omid (2010) in Iranian greenhouses obtained values of 2438.74 kWh and 1108.63 L ha^-1 of electricity energy and diesel fuel energy, respectively. In this research, a gas meter was used to measure the gas consumption of the gas boiler used in the heating system. Fuel energy consumption was calculated every day. Since hydroponic cucumber cultivation was done during the hot season of the year, gas consumption was very low. The total fuel energy consumed during the cultivation period was 184.6 m³ (Fig. 5b).