| About this Abstract |
| Meeting |
2021 TMS Annual Meeting & Exhibition
|
| Symposium
|
Energy Technologies and CO2 Management
|
| Presentation Title |
Prediction Model of Converter Oxygen Consumption Based on Recursive Classification and Feature Selection |
| Author(s) |
Liu Zhang, Zhong Zheng, Kaitian Zhang, Xinyue Shen , Yongzhou Wang |
| On-Site Speaker (Planned) |
Liu Zhang |
| Abstract Scope |
To improve the prediction accuracy of oxygen consumption in the steelmaking converter, a prediction model of converter oxygen consumption based on long short-term memory (LSTM) network ensemble learning was proposed. The converter production data were clustered to construct multiple training subsets. And the prediction model of converter oxygen consumption was constructed by using the LSTM network for each subset. For the test sample, the weight of the prediction results can be determined by the mode matching degree between the test sample and different types of multiple models, and the oxygen consumption of the sample was estimated by the weighted sum from the prediction of the multiple models. The converter production data from a steel enterprise were used for testing. The results showed that the prediction accuracy of the ensemble learning model was higher than that of the single prediction model. |
| Proceedings Inclusion? |
Planned: |