Researchers in China have analyzed the effect of phase change material on BIPV and created an artificial neural network to predict its effect on system temperature. The proposed approach reportedly achieved superior predictive performance compared to previous methods.
A Chinese research group has investigated the effect of using phase change material (PCM) for cooling building-integrated photovoltaic panels (BIPV) and developed a method for predicting BIPV panel temperature using an artificial neural network (ANN) .
PCMs can do that absorb, store and release large amounts of latent heat within defined temperature ranges. They are often used at research level for the cooling of PV modules and the storage of heat.
“The use of ANNs in predicting the performance of PV-PCM systems is of great importance in promoting the efficiency and reliability of these integrated technologies,” the team said. “ANNs excel at capturing complex relationships within large data sets, allowing for more accurate modeling of the complex interplay between various factors that influence the performance of PV-PCM systems.”
In the first part of their research, the academics developed a numerical model by simulating a BIPV system based on a polycrystalline module with cells with a thickness of 0.4 mm, a glass layer of 3 mm thickness and an encapsulant of ethyl vinyl acetate (EVA). ). ) layer of 0.5 mm. The module contained another layer of EVA, Tedlar with a thickness of 0.33 mm used on the back surface, and a 0.5 mm aluminum chamber with PCM was placed for cooling. This panel had a nominal efficiency of 17.5% at a reference temperature of 25 °C. The irradiance and ambient temperature were simulated as that of Jiangsu, China, on October 15, 2022.
“The highest and lowest difference between the PV panel electrical efficiency of systems with and without PCM is equal to 2.17% and 0.1%, respectively, which occurs at 1:55 PM and 4:00 PM,” the scientists said. “The maximum and minimum difference between the efficiency of the two systems is 2.17% and 0.10% respectively, which occurs at 1:55 PM and 4:00 PM”
Following these results, the academics developed the ANN prediction method, using the deep learning algorithm group method of data processing (GMDH). “A GMDH-type multi-layer neural network was applied to estimate the daytime solar cell temperature of the PV-PCM system using the input variables of hourly total solar radiation and ambient temperature,” they explained.
For training, the researchers extracted 56 data points from their numerical model and then tested it on 24 prediction points. “The correlation of the coefficient (R2) is obtained as 0.97602,” they found. “Additionally, the root mean square error (RMSE) and mean square error (MSE) were calculated as 1.483 and 2.22, respectively.”
The proposed approach reportedly achieved superior predictive performance compared to previous methods.
However, the group also emphasized that “since the model developed in this study can only predict the performance of the system on a specific day, future research efforts will focus on developing a model that can predict the performance of the system throughout the year .”
The results are presented in “Predicting the temperature of a building-integrated photovoltaic panel equipped with phase change material using an artificial neural network”, published in Case studies in thermal engineering. Scientists from China’s Jiangxi University of Software Professional Technology, the Postdoctoral Research Workstation of Aheadsoft Software and the Huaiyin Institute of Technology conducted the research.
This content is copyrighted and may not be reused. If you would like to collaborate with us and reuse some of our content, please contact: editors@pv-magazine.com.