Determination of electrical and thermal conductivities of n- and p-type thermoelectric materials by prediction iteration machine learning method

Hasan Tiryaki, Aminu Yusuf and Sedat Ballikaya

Energy, 2024, vol. 292, issue C

Abstract: Synthesising a novel high-performance thermoelectric (TE) material is time-consuming because different compositions of the chemical elements are usually varied using a trial-and-error approach. Moreover, the characterisation of TE materials requires both complex and expensive equipment; these measuring devices often fail during operation. Machine learning (ML) models can be used to accurately predict the properties of a novel composition, saving time as well as the cost of the material and equipment. In this study, two different prediction scenarios have been demonstrated, one for n-type with the general formula BixBayBzYbtTe3, and another for p-type with the general formula SbxBiyBazBtYbwTe3. From the experimental data of the above-mentioned n- and p-type compounds, transport properties of n-type Bi2-xTe3 and p-type Sb1.5Bi0.5-xTe3, where x ranges from 0 to 0.5, involving content variations of Ba, B, and Yb, are predicted. Case 1 deals with the prediction of resistivity and Seebeck values, while case 2 predicts the heat capacity (Cp) and thermal diffusivity values of the n- and p-type TE materials. Herein, different compositions of n-type BixBayBzYbtTe3 and p-type SbxBiyBazBtYbwTe3 are synthesised, and the experimental data are fed to 26 ML models. After training all the ML models, an Artificial Neural Networks (ANN) ML model with the highest R2 values of 0.9943 and 0.9995 in cases 1 and 2, respectively, is found to outperform the other models. The prediction iteration method is applied to the ANN to predict the transport properties of the p-type Sb1.5Bi0.2Ba0.3Te3 and n-type Bi1.9Ba0.1Te3. The accuracy of the prediction iteration method increases with the number of iterations. At the end of the 100th iteration, the prediction error of the ANN model in case 1 is as low as 7%, while it is 9% in case 2.

Keywords: ANN; Energy harvest; Machine learning; Prediction; Material synthesis; Thermoelectric materials (search for similar items in EconPapers)
Date: 2024
References: View references in EconPapers View complete reference list from CitEc
Citations:

Downloads: (external link)
http://www.sciencedirect.com/science/article/pii/S0360544224003694
Full text for ScienceDirect subscribers only

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:eee:energy:v:292:y:2024:i:c:s0360544224003694

DOI: 10.1016/j.energy.2024.130597

Access Statistics for this article

Energy is currently edited by Henrik Lund and Mark J. Kaiser

More articles in Energy from Elsevier
Bibliographic data for series maintained by Catherine Liu ().