Determination of hybrid energy needs of a fully automated greenhouse with artificial neural networks

Objective: Determining the energy need by analyzing the data collected from the greenhouse environment with artificial neural network machine learning techniques. In this way, it is to reduce energy consumption, increase energy efficiency and increase greenhouse efficiency by using renewable energy sources and support sustainable agricultural practices. Material and Methods: The study was carried out in the A&S Agriculture greenhouse in the coordinates of 37°5644.6"N 28°5039.7"E in the Aydin province of Bukharken district of Aegean Region, located in the west of our country. Greenhouse gothic roof, side walls hard plastic other places soft plastic covered, steel construction 50000 m2 closed area of soilless agriculture full automation partitionless block greenhouse. In this study, tomato development, air temperature between 2021-2023, wind speed, wind direction, sunbathing time, radiation value, radiated radiation value, relative humidity, cloudiness value, etc, data on air pressure and electricity consumption values were used. During this time, approximately 5808 data were obtained from each dataset. This data was analyzed in the Weka program, a software with General Public License (GPL) based on Java and open data, developed by Waikato University in New Zealand. In total, about 63888 data analyses were performed with Weka. Results: Using the Weka program, the algorithm, which is "Random Tree" (Random decision tree) under the classification algorithm, determined the factors that affect the electricity consumption of the greenhouse. As a result of this analysis, the factors affecting energy consumption were determined as the maturation-harvest period of the tomato, the external environment temperature, the process of rooting and flowering, the radiated radiation value and wind speed. Multilayer Perceptron algorithm using different hidden layers and different machine learning repetitions to achieve optimum precision artificial neural network (ANN), energy prediction needed for the next 5 seasons. According to these estimation results, it was found that the energy needed by the greenhouse with the solar energy system that produces 20 kW of energy per hour and has a storage system. Conclusion: When the previous studies were examined, it was emphasized that a large part of the energy consumption of the greenhouse is realized in heating systems. In addition to previous findings, this study found that the greenhouse is more effective on energy consumption than the heating system of the nutrient solution system. The cost of the solar energy system, which is calculated according to the estimation results of energy needs, was found to be 132870$. This facility produces 836046 kW/year of electricity per year, it was calculated that 700 days later the plant amortizes the installation costs when calculated according to the current electricity price. According to this model, it is seen that the electricity need can be met with a solar panel surface area up to 4.6% of the surface area of the existing greenhouse. Due to the fact that this region is a greenhouse region, the rate of 4.6 '% can be taken as a reference value in the panel surface area account to meet the energy needs of other greenhouses growing the same product in the region. In addition, this study revealed that the heating and feeding solution system should be evaluated together while energy efficiency work is carried out in fully automation greenhouses.