Scholarlink

<p>To address the urgent need for energy conservation and carbon reduction in the steel industry and achieve converter energy refined management, this paper takes converter process energy consumption as the research object and proposes a prediction model based on Support Vector Regression (SVR). Ten key influencing factors are determined through mechanistic analysis, and their correlation with energy consumption is verified using Grey Relational Analysis (GRA). Dimensionality reduction via Principal Component Analysis (PCA) is then employed to mitigate overfitting, and SVR hyperparameters are optimized using grid search and <b>K</b>-fold cross-validation. Experimental results demonstrate that the prediction accuracy (RMSE = 3.9854, MAE = 2.9524, MAPE = 7.33%) and computational efficiency of this model are superior to those of Backpropagation (BP) Neural Network and Random Forest (RF) models. The model can provide a reliable basis for energy consumption optimization in actual production, strongly supporting the green and low-carbon transformation of the steel industry. Finally, to enhance the model’s reliability, the results are analyzed using Shapley Additive exPlanations (SHAP).</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

Prediction of Converter Process Energy Consumption Based on Support Vector Regression

  • 期刊论文

  • 发表日期：

    2026年6月29日

查看全文
  • Feixiang Dai,
  • Xiaojing Yang,
  • Xiangjun Bao,
  • Lu Zhang,
  • Guang Chen

摘要

To address the urgent need for energy conservation and carbon reduction in the steel industry and achieve converter energy refined management, this paper takes converter process energy consumption as the research object and proposes a prediction model based on Support Vector Regression (SVR). Ten key influencing factors are determined through mechanistic analysis, and their correlation with energy consumption is verified using Grey Relational Analysis (GRA). Dimensionality reduction via Principal Component Analysis (PCA) is then employed to mitigate overfitting, and SVR hyperparameters are optimized using grid search and K-fold cross-validation. Experimental results demonstrate that the prediction accuracy (RMSE = 3.9854, MAE = 2.9524, MAPE = 7.33%) and computational efficiency of this model are superior to those of Backpropagation (BP) Neural Network and Random Forest (RF) models. The model can provide a reliable basis for energy consumption optimization in actual production, strongly supporting the green and low-carbon transformation of the steel industry. Finally, to enhance the model’s reliability, the results are analyzed using Shapley Additive exPlanations (SHAP).

Graphical Abstract