电解铝分子比预测数据集
收藏资源简介:
(1)从企业电解槽控系统实时运行数据(温度、电压、铝水平、电解质水平等),从生产线化验记录分子比、氧化铝浓度等化验结果(定期),从物料系统采集氟化盐、氧化铝投料记录。 数据加工: (1)清洗异常值、空值、填补缺失数据(向上填充和向下填充),统一时间戳对齐。 (2)数据处理训练参数,加上效应系数、剔除电流、总功耗、Fe、Si含量。 (3)构建时间序列特征,用于机器学习模型训练。 预测算法: 采用华为盘古大模型算法建立分子比动态预测模型。
(1) Data sources: - Real-time operational data from the enterprise electrolytic cell control system, including temperature, voltage, aluminum level, electrolyte level and other related metrics; - Periodic laboratory test results from the production line, covering molecular ratio, alumina concentration and other test items; - Feeding records of fluoride salt and alumina collected from the material management system. Data processing: 1. Conduct data cleaning: remove outliers and null values, fill missing data via forward filling and backward filling, and unify and align timestamps; 2. Process training parameters: add the effect coefficient, and eliminate features including current, total power consumption, Fe and Si content; 3. Construct time-series features for machine learning model training. Prediction algorithm: A dynamic molecular ratio prediction model is established using Huawei Pangu Large Language Model (LLM) algorithm.




