Enhancing predictive accuracy of inside temperature and humidity in an agricultural greenhouse using data-driven modeling with artificial neural networks

This work presents an original and innovative approach by combining greenhouse cooling with artificial intelligence models to ensure food security. The study is divided into two parts: Experimental and theoretical. In the first part, a cooling system was implemented in a tunnel-type agricultural greenhouse and compared to a control system. The cooling system consists of multiple fans powered by two solar panels. Data was collected using an acquisition system (Arduino) over approximately one month. This data, along with external data, was utilized to predict the internal temperature of the greenhouse using back-propagation neural network models. The results obtained demonstrate the reliability of the model based on all tests, as evidenced by the coefficient of determination (R 2 ) and the mean square error (MSE) for the prediction.


INTRODUCTION
Fresh water is scarce in arid climates.They also have low humidity, high potential evapotranspiration, and high temperatures.So, we've needed protected agriculture to grow crops.The rising population brings concerns about food security.It also brings recognition of dwindling freshwater resources.The adoption of protected agriculture is rising in arid regions, which make up about 30% of the world's landmass.They support roughly 20% of the global population (Al-Sulaiman, 2002).In these areas, dry climates get little rain each year.They also face high rates of potential evapotranspiration.Arid steppe climates get less rain than the potential evapotranspiration.Arid desert climates get very little or no rain (Aziz et al., 2018).Under such circumstances, mitigating thermal stress and managing evapotranspiration are critical considerations.
In dry areas, raising humidity in greenhouses can cut crop water loss.This is especially true in deserts with high heat and very low humidity (Doğramacı and Aydın, 2020).Nonetheless, in many regions across the Middle East and North Africa, temperatures can soar to 46°C.Access to fresh water is also restricted.So, providing the right growing conditions there is very hard.Research shows that people often use active cooling systems like misting and fan and pad ventilation to reduce heat in arid climates (Aziz et al., 2018).These systems, plus evaporative cooling and shading, can cut the greenhouse air temperature by up to 8°C.But, misting systems might not cool as much as fan and pad systems.But they do provide more even conditions in the greenhouse (Helmy et al., 2013).But, it's crucial to regulate active ventilation to pre-vent excessive crop transpiration.
Although these cooling systems prove effective, they also consume a great deal of water.Evaporative cooling in greenhouses can use up to 67% of the total water demand (Kassem et al., 2005).To tackle this challenge, passive ventilation is often used first to cool.
When passive ventilation is not enough, people use evaporative cooling.Passive greenhouse technology emphasizes taller greenhouses.They do this to reduce peak air temperatures (Koca et al., 1991;Lalmi et al., 2021).Still, as greenhouses get bigger, active vents and cooling systems work less well.Adding retractable roofs and adjustable openings is essential.They let passive and active This study aims to conduct an experiment.It will investigate the heat dynamics of two tunnel greenhouses.One has a cooling system, and the other does not.The greenhouse lacking a cooling system will act as a control, facilitating comparison to check the impact of the cooling system.The setup is at the Unit of Applied Research in Renewable Energies (URAER) in Ghardaïa.Ghardaïa lies at a latitude of 32.37° north and a longitude of 3.77° west.This location is good for studying greenhouse thermal performance.The study has four sections.The first section provides a detailed description of the experimental site.It includes its geography and other key details.This section helps establish a clear understanding of the experimental conditions.In the second section, researchers set up the experiment, took measurements, and analyzed the results.This phase involved monitoring and recording temperatures inside the greenhouses.This allowed for a full evaluation of their thermal behavior.The third section of the study involved the use of a feed-forward backpropagation neural network.We've used this neural network to predict the inside temperature of the greenhouse, based on the collected data.This analysis can help predict how well the cooling system works.It can show how the system affects the greenhouse's temperature.The study aims to improve our understanding of tunnel greenhouses.It will do this by conducting experiments and using advanced data analysis.The green-houses will either have or not have cooling systems.This research can help make greenhouse designs more efficient and sustainable.This will benefit the agricultural industry.
The final section presents the results obtained and their interpretations, with its conclusion.We used the feed-forward backpropagation neural network to predict greenhouse temperatures.This analysis can offer insights into cooling system effectiveness.The study aims to improve understanding of greenhouse thermal behavior.This research can help develop better greenhouse designs.The final section discusses the results and conclusions.

MATERIALS AND METHODS
Our experiment took place at the Unit of Applied Research on Renewable Energies in Ghardaïa, Algeria, from 15-03-2023 to 15-06-2023.
We studied temperature and humidity gradients in a greenhouse.We used a Fan-Pad cooling system under Ghardaia's conditions.

Experimental Site description
Before describing the region of our research work, it should be noted that about 77% of the Algerian surface area is arid and semi-arid regions.The characteristics of the region of Ghardaïa, (Figure 1), are: Latitude and 32°36 N.
Altitude of 469 m above sea level.

Rate of sunny days per year: 77%
Annual average daily global solar irradiation is about 7 kWh/m 2 .
It experiences a desert climate characterized by hot and dry summers, and mild winters.The monthly temperature in Ghardaia can vary throughout the year.During the summer months, from June to August, Ghardaia experiences high temperatures.Average temperatures during this period can range from 35°C to 42°C (95°F to 107.6°F), with occasional spikes reaching even higher.Intense heat and minimal rainfall characterize these months (Shaik et al., 2020).In contrast, the winter months, from December to February, are mild in Ghardaia.Average temperatures during this period range from 14°C to 20°C (57.2°F to 68°F).The temperatures are cooler than in the summer.But, Ghardaia still has pleasant weather in the winter, (Figure 2).

Design of the greenhouse
To conduct this study, we used two plastic tunnel greens.They were identical in size (200cm long, 100cm wide, 80cm tall) and design.
Researchers designed one greenhouse as the test.They used the other as the control.Using plastic packaging, we've built the greenhouses.They also used plastic tubes.This makes them easy to modify and adapt.The choice of materials made the needed changes easier.The study required modifications and adjustments.

Cooling system design
We determined the cooling system's parameters using psychrometric calculations.We based these on data from inside and outside the greenhouse.This study aimed to test the Fan-Pad system's effectiveness at keeping good plant growth conditions.It also aimed to map these conditions in the greenhouse.The study's results will help improve the design and operation of greenhouse cooling systems in similar climates.They will enhance the productivity and quality of crops grown in these places.In the experiment, they installed shading screens in the greenhouse.The screens control how much sunlight enters.Additionally, we've closed the side and top windows of the greenhouse.This prevented any outside air from entering or leaving.We created a lab to isolate the Fan-Pad cooling system's effects.It focuses on the temperature and humidity gradients in the greenhouse.

Components of the fan-pad cooling system
A direct evaporative cooler cools air: It does this by evaporating water.This makes it a simple and effective device.It operates based on the principle that when water evaporates, it absorbs heat from the surrounding air, resulting in a cooling effect.Researchers have applied the fan-pad evaporative cooling system in the experimental greenhouse.Figure 4 shows a general view of the fan-pad cooling system.
It contains the following elements: Filter Media: A highly wettable porous material made from palm leaf, has a thickness, and they are kept moist by water continuously dr ipping onto its upper edges.

ARTICLE | OPEN ACCESS
Indian Journal of Engineering 21, e6ije1681 (2024) 6 of 15 An evaporative cooling pad, also known as a wet pad, is key to evaporative cooling systems.The pad is 80 cm by 4 cm by 40 cm, as seen in Figure 5(a).It's placed on the air inlet at the east end, 10 cm above the ground.
Blower Fan: The described system includes two large blower fans, each operating at a voltage of 12V and consuming 50Kw of power, Figure 5(b).The housing contains and uses these fans to draw in warm air from the outside.Afterward, the fans push the cooled and humidified air into the greenhouse.
The water distribution system incorporates two perforated tubes positioned above the filter media, as shown in Figure 5(c).These tubes are essential for dispersing water across the entire surface of the pad.This careful distribution mechanism ensures that the filter media stays saturated well.This level helps efficient cooling.The system ensures even water distribution.This maximizes the cooling process's effectiveness, enhancing its capabilities.
The water collection tube sits at the bottom of the evaporative cooling pad.Its main job is to collect and channel extra water, as well as it's impurities or sediments.These impurities and sediments pass through the pad during evaporation.This tube collects water and prevents it from pooling or causing damage to the pad or surrounding areas.A pump directs the water collected in the pipe to the tank for recycling back into the system for further use.from the pad into the air, absorbing heat from the warm air and cooling it.The cooled air is then blown into the greenhouse, providing a refreshing and cooler environment inside, (Figure 6).

Figure 6
Cooling system in the experimental greenhouse In our study, we've utilized a total of seven DHT22 sensors to measure temperature and humidity.We placed three sensors in the experimental greenhouse.We put another three in the control greenhouse, which lacks a cooling system.We positioned one sensor outside.For data acquisition, we've employed an Arduino board of the Mega 2560 type along with a relay block.The wiring system involved connecting the Arduino board, memory card, and clock.Two solar panels powered the system, which consisted of a regulator (24V) and fans.You can find the specifications of these components in (Table 1).Figure 7 presents all system concepts.

Experimental results
The researchers conducted the trials during the period specified in section one.They did them at the same location as the previous trials to cool the greenhouse.We've installed a cheap evaporative system.It has pads and a chopper.It also has a water recovery and reuse system.This system saves and uses water well.The system only wasted 100 ml per minute during water evaporation.Standard engineering data recommends keeping the greenhouse temperature below 30-32°C.
Figure 8 shows a comparison.It is between the outdoor temperatures and the greenhouse temperatures.The greenhouse lacks a cooling system.The period is from June 4th to June 15th.The results show a big temperature increase in the greenhouse.Peak temperatures ranged from 18°C to 50°C, and they were from 13:00 to 14:00.This temperature rise is likely due to the influence of global warming inside the greenhouse.Similarly, when comparing temperatures inside the greenhouse to the outside, there is a big difference.
The inside can be over 10°C hotter during the day.This is clear during the hot, dry weather from days 8 to 10. Temperatures range from 25°C to 50°C.The difference was up to 4°C in the day.This variation is due to the effectiveness of the large evaporative system in the greenhouse.Its economic cost did not exceed 10,000.00DZ.
To ensure efficient water usage, we've implemented a water recovery system to replenish moisture in the pad.We measured the amount of lost water and evaporated it in the greenhouse.We used a controlled water flow.The water tank set the downward flow at 6 L/min and maintained the upward flow at 5.9 L/min.This configuration allowed for an evaporation rate of 100 mL/min.The observed evaporation rate shows a cost-effective performance.This performance is rational compared to the system.Figure 9 illustrates the temporal variation of humidity in both greenhouses.Throughout the day, we've observed a big inverse link.It's between temperature and humidity.As the temperature increases, the humidity decreases, and vice versa.These findings support past studies.They also confirm the well-known link.It is between high temperature and low humidity.The multi-layer perceptron (MLP) is a neural architecture used in machine learning.It falls under the category of artificial neural networks (ANNs).MLPs consist of many layers of connected neurons (nodes).They include an input layer, one or more hidden layers, and an output layer.These networks can approximate any quantifiable function.

CONCLUSION
cooling and ventilation systems Liao et al., (1998) integrate well.In arid climates, energy and water conservation are crucial.To use less water and energy, there's growing interest in combining passive and active ventilation.Recent studies show that misting and shading cool plants.They improve the plant environment in semi-arid regions (Mahmood and Aljubury, 2022; Malli et al., 2011).Researchers have studied alternative evaporative cooling techniques for greenhouses in arid climates.These techniques need less forced ventilation than traditional fan and pad systems (Nada et al., 2019).But, it's still necessary to develop cheap cooling methods.These methods should use materials available in arid climates.Recent research has focused on protected agriculture in semi-arid regions.It shows that misting and shading are key.They use passive evaporative cooling.It's the best way to improve the plant environment.Other studies have looked at ways to cool greenhouses in arid climates.They do it without forced ventilation.They demand less ventilation than fan and pad systems.But none have focused on low-cost cooling techniques using local materials (Sellam et al., 2022).

Figure 1
Figure 1 Ghardaïa position in Algeria maps

Figure 2
Figure 2 Annual monthly temperature evolution in 2019, 2020 and 2021

Figure 5
Figure 5 Components of Fan-Pad cooling system: a) Evaporative cooling pad, b) Fan, Palm leaf, c) Palm leaf, d) Arranging palm leaf inside an evaporative cooling pad

Figure 7
Figure 7 Schematic representation figure

Figure 8 Figure 9
Figure 8The evolution of inside greenhouse temperature (with and without a cooling system), versus outside greenhouse temperature

Figure 11
Figure 11 shows a strong correlation between measured and predicted values.The agreement is strong.The low mean squared error (MSE) values show this.They measure the average squared difference between actual data and model predictions.The small MSE values confirm that the multi-layer perceptron works well.It's so by aligning its predictions with the real data.

Figure 12
Figure 12 Evolution training, test, validation, and global error

Figure 13 Best
Figure 13 Best Validation Performance Chart

Table 1
Characterization of the two Solar Monocrystalline Panels 100W 12V