【摘要】 为优化连铸工艺，实现对二冷区各区段水量的精准预测，建立了温度预测模型和二冷配水预测模型。充分利用生产过程中积累的数据，挖掘生产数据间的内在规律和关联性。首先，温度预测模型将影响温度的特征参数预处理后，按照时间序列的顺序输入到Transformer模型中，模型通过对这些特征参数进行编码和自注意力机制的学习，捕捉序列中的时间依赖关系和特征间的交互作用，准确预测二冷区的温度分布。其次，由于温度是影响二冷配水准确性的关键因素，由温度预测模型预测得到的温度，再结合影响二冷水分配的其他影响参数，作为特征序列输入到二冷配水预测模型，该模型通过随机森林的集成学习，利用特征参数的内在规律和关联性，预测二冷区各段的水量分配。研究结果表明，所提出的动态二冷配水模型能有效提升铸坯质量和生产效率，并且该预测模型不受固定参数和专家经验的限制，能够自动适应生产过程中的变化，为连铸过程中动态二冷配水提供参考，对提高连铸生产线的智能化水平具有重要意义。更多还原

【Abstract】 To optimize the continuous casting process and achieve accurate prediction of water volume in each section of the secondary cooling zone, temperature prediction models and secondary cooling water distribution prediction models were established. Fully utilize the accumulated data in the production process, and explore the inherent laws and correlations between production data. Firstly, the temperature prediction model inputs the feature parameters affecting the temperature into the Transformer model in the order of time series by preprocessing them, and the model accurately predicts the temperature distribution in the second cooling zone by capturing the time-dependent relationship in the sequence and the interaction between the features through encoding these feature parameters and learning by the self-attention mechanism. Secondly, since temperature is a key factor affecting the accuracy of water distribution in the second cooling zone, the temperature predicted by the above temperature prediction model, combined with other influential parameters affecting the distribution of water in the second cooling zone, is inputted as a sequence of features into the second cooling water distribution prediction model, which predicts the distribution of water in each section of the second cooling zone by utilizing the intrinsic regularity and correlation of the feature parameters through the integration of the Random Forest learning. The results show that the dynamic two-cooling water allocation model proposed in this paper can effectively improve the quality of billet casting and production efficiency, and the prediction model is not limited by fixed parameters and expert experience, and it can automatically adapt to the changes in the production process, which provides a reference for the dynamic two-cooling water allocation in the continuous casting process, and it is of great significance to improve the intelligent level of continuous casting production line.更多还原