Scholarlink

<p>The bath temperature of an aluminum reduction cell is a reflection of the energy balance and is influenced by various complex factors. Controlling this temperature within an optimal range is critical for aluminum production. In this work，the bath temperature of a 500-kA aluminum reduction cell was predicted based on its historical production data using five machine learning models, including support vector regression (SVR), kernel ridge regression (KRR), random forest (RF), gradient boosted decision tree (GBDT), and lightweight gradient boosting machine (LightGBM). The results show that the LightGBM model has the best performance on accuracy, followed by GBDT and RF models, and the accuracy of SVR and KRR models is relatively low. SHapley Additive exPlanations (SHAP) model was employed to interpret LightGBM. It was found that the importance of features in descending order is last bath temperature, fluoride salt feeding weight, normal break and feed cycles, set voltage, and metal tapping weight. LightGBM can be applied to predict the bath temperature of aluminum reduction cell with an average difference of 1.6&#xa0;℃ between the predicted temperature and the actual temperature.</p> Graphical Abstract <p></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

Bath Temperature Prediction of Aluminum Reduction Cell Based on Machine Learning Algorithm

  • 期刊论文

  • 发表日期：

    2025年3月14日

查看全文
  • Jingkai Lin,
  • Aimin Liu,
  • Zhaowen Wang,
  • Zhongning Shi,
  • Fengguo Liu

摘要

The bath temperature of an aluminum reduction cell is a reflection of the energy balance and is influenced by various complex factors. Controlling this temperature within an optimal range is critical for aluminum production. In this work，the bath temperature of a 500-kA aluminum reduction cell was predicted based on its historical production data using five machine learning models, including support vector regression (SVR), kernel ridge regression (KRR), random forest (RF), gradient boosted decision tree (GBDT), and lightweight gradient boosting machine (LightGBM). The results show that the LightGBM model has the best performance on accuracy, followed by GBDT and RF models, and the accuracy of SVR and KRR models is relatively low. SHapley Additive exPlanations (SHAP) model was employed to interpret LightGBM. It was found that the importance of features in descending order is last bath temperature, fluoride salt feeding weight, normal break and feed cycles, set voltage, and metal tapping weight. LightGBM can be applied to predict the bath temperature of aluminum reduction cell with an average difference of 1.6 ℃ between the predicted temperature and the actual temperature.

Graphical Abstract