| Predicting energy parameters based on artificial neural network methods in a photovoltaic-thermal solar system |
| Paper ID : 1238-NICAME1402 |
| Authors: |
|
Sabour Shoja Pour *1, Ali Motevali2, Seyed Hashem Samadi3, Azadeh Ranjbar Nedamani4, Pourya Biparva5 1Department of Biosystem Mechanical Engineering , Faculty of Agricultural Engineering, University of Sari Agricultural Sciences and Natural Resources, Iran 2Department of Biosystem Mechanics Engineering, University of Sari Agricultural Sciences and Natural Resources, Iran 3Department of Biosystems Mechanical Engineering, University of Tarbiat Modares, Iran 4Department of Biosystems Mechanical Engineering, University of Sari Agricultural Sciences and Natural Resources 5Department of Basic Sciences, University of Sari Agricultural Sciences and Natural Resources, Iran |
| Abstract: |
| The extensive use of energy in agricultural, industrial, and domestic activities, especially in developing countries, has led to a significant increase in energy demand. Meeting this energy demand with fossil fuels can result in greenhouse gas emissions and fuel price hikes. Therefore, more effective utilization of renewable energy sources can be instrumental in addressing these challenges. In this study, output power data recorded in a greenhouse were modeled using artificial neural networks. The modeling was performed using machine learning methods, including time-delay neural network, multi-layer perceptron neural network, and non-linear autoregressive neural network with logarithmic activation function, 60 neurons, and 100 iterations. The proposed models' performance and results were evaluated, showing that the predictive models for the system's power parameter outside the greenhouse performed better than in the greenhouse environment. Additionally, the results indicated that the multi-layer perceptron neural network modeling method with a close-to-zero overall coefficient of determination (R2 = 0.2) was not suitable for the studied parameter, while time-delay and non-linear autoregressive methods achieved desirable performance in photovoltaic-thermal system modeling (R2 = 0.9). These methods were recommended for use in this modeling approach. |
| Keywords: |
| Artificial Intelligence, Energy, Greenhouse, Renewable, Neural Network, Prediction. |
| Status : Paper Accepted (Poster Presentation) |
