Machine learning-based modeling of thermoelectric materials and air-cooling system developed for a humid environment

• Published in: Materials express 2021-02, Vol.11 (2), p.153-165
• Main Authors: Ameenuddin Irfan, Sayed, Irshad, Kashif, Algahtani, Ali, Azeem, Babar, Tirth, Vineet, Algarni, Salem, Islam, Saiful, Abdelmohimen, Mostafa A. H.
• Format: Article
• Language: English
• Subjects: Air Cooling; Cooling systems; Empirical analysis; Empirical Modeling; Gaussian Process Regression; Machine Learning; Mathematical Model; Polynomials; Relative humidity; Renewable energy sources; Room temperature; Thermoelectric cooling; Thermoelectric materials; Thermoelectric Module
• ISSN: 2158-5849; 2158-5857
• DOI: 10.1166/mex.2021.1909

The thermoelectric air-cooling system (TE-ACS) has witnessed exponential progress due to its ability in tackling the environmental pollution issues by utilizing renewable energy sources. In Present research focused on, different machine learning models developed to predict TE-ACS parameters and comparisons are made with the empirical models. The empirical models, based on actual experimental data for the same input variables, are developed by using the polynomial fitting method. The machine learning models, especially the nonlinear regressions using the Gaussian exponential method, have shown less error in comparison with experimental results. The highest prediction accuracy of the machine learning model for hot side temperature of thermoelectric material is achieved with R2 = 0.87 and RMSE = 0.52. For the cold side temperature, R2 = 0.92 and RMSE = 0.44. The machine learning prediction for the inside-room temperature results in R2 = 0.86 and RMSE = 1.18. The model for relative humidity inside the room produced R2 = 0.87 with an RMSE value of 0.89. These models may be utilized to evaluate the TE-ACS performance for the larger input values that are difficult or, at times, impossible to perform in the actual experimental setup. Therefore, these machine learning models gives a strong basis for the design and analysis of TE-ACS methods.