Duan Yi, Chen Guang, Bao Xiangjun, Xu Jing, Zhang Lu, Yang Xiaojing
School of Energy and Environment, Anhui University of Technology, Ma'anshan, 243002, Anhui Province, People's Republic of China.
Sci Rep. 2025 Apr 1;15(1):11115. doi: 10.1038/s41598-025-95134-3.
Energy consumption serves as a critical indicator of energy utilization efficiency and environmental sustainability in the steel production process. Accurately predicting the Heat energy consumption per ton (HEC, GJ/t) of steel billet in Steel Rolling Reheating Furnace (SRRF) presents a formidable challenge owing to the complex interplay of factors such as production scheduling, raw material characteristics, process parameters, and equipment condition. This study proposes a novel approach to predict HEC (GJ/t) by utilizing actual production data from SRRF. A genetic algorithm (GA) optimized back-propagation neural network (BPNN) is developed and its performance is compared to that of a standard BP model. Experimental results reveal that the optimized GA-BP model, with a neural network structure of 17-10-1, achieves a prediction accuracy of 94.7% surpassing the 90.24% accuracy of the standard BP model. The proposed GA-BP model demonstrates superior predictive capabilities and robustness, offering valuable insights for optimizing process parameters and improving energy efficiency in SRRF operations.
能源消耗是钢铁生产过程中能源利用效率和环境可持续性的关键指标。由于生产调度、原材料特性、工艺参数和设备状况等因素的复杂相互作用，准确预测轧钢加热炉（SRRF）中每吨钢坯的热能消耗（HEC，GJ/t）面临着巨大挑战。本研究提出了一种利用SRRF实际生产数据预测HEC（GJ/t）的新方法。开发了一种遗传算法（GA）优化的反向传播神经网络（BPNN），并将其性能与标准BP模型进行了比较。实验结果表明，具有17-10-1神经网络结构的优化GA-BP模型实现了94.7%的预测准确率，超过了标准BP模型90.24%的准确率。所提出的GA-BP模型具有卓越的预测能力和鲁棒性，为优化工艺参数和提高SRRF操作中的能源效率提供了有价值的见解。
