Scholarlink

<p>It is essential to accurately predict thermoelectric properties for the rational design of high-efficiency energy conversion materials. In this study, a data-driven model based on extremely learned Deep Neural Networks (DeepELM-DNNs) was developed to forecast the power factor which is a key thermoelectric performance indicator for two polycrystalline systems: (GeTe)₁₀Sb₂Te₃ and (GeTe)₂₄Sb₂Te₃. Three DeepELM-DNN models were implemented and tested using experimentally collected features, namely seebeck coefficient, electrical resistivity, and temperature datasets. The models were assessed by four metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Coefficient of Determination (R²), and Mean Absolute Percentage Error (MAPE). The best model, DeepELM-DNN-3, achieved good performance for (GeTe)₁₀Sb₂Te₃ with MAE = 0.0932, RMSE = 0.1112, R² = 0.9837, and MAPE = 4.3261%, and for (GeTe)₂₄Sb₂Te₃ with MAE = 0.1032, RMSE = 0.1387, R² = 0.9834, and MAPE = 4.9748%. These findings illustrate the robustness of the model to describe zT magnitudes with a manner corresponding to their nonlinearity in temperature and composition for complex telluride materials. The results also reveal that the DeepELM-DNN is highly accurate and efficient in modeling thermoelectric materials, and has the potential to significantly accelerate AI-driven discovery of high-performance heat-to-electricity conversion materials.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Data-driven prediction of thermal and thermoelectric performance in GeTe-Sb2Te3 systems: extreme learning of deep neural networks

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年12月10日

查看全文
  • Kadhum Hassan Ali,
  • Luttfi A. Al-Haddad

摘要

It is essential to accurately predict thermoelectric properties for the rational design of high-efficiency energy conversion materials. In this study, a data-driven model based on extremely learned Deep Neural Networks (DeepELM-DNNs) was developed to forecast the power factor which is a key thermoelectric performance indicator for two polycrystalline systems: (GeTe)₁₀Sb₂Te₃ and (GeTe)₂₄Sb₂Te₃. Three DeepELM-DNN models were implemented and tested using experimentally collected features, namely seebeck coefficient, electrical resistivity, and temperature datasets. The models were assessed by four metrics: Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Coefficient of Determination (R²), and Mean Absolute Percentage Error (MAPE). The best model, DeepELM-DNN-3, achieved good performance for (GeTe)₁₀Sb₂Te₃ with MAE = 0.0932, RMSE = 0.1112, R² = 0.9837, and MAPE = 4.3261%, and for (GeTe)₂₄Sb₂Te₃ with MAE = 0.1032, RMSE = 0.1387, R² = 0.9834, and MAPE = 4.9748%. These findings illustrate the robustness of the model to describe zT magnitudes with a manner corresponding to their nonlinearity in temperature and composition for complex telluride materials. The results also reveal that the DeepELM-DNN is highly accurate and efficient in modeling thermoelectric materials, and has the potential to significantly accelerate AI-driven discovery of high-performance heat-to-electricity conversion materials.