Prediction of semi-coke yield using machine learning methods based on the analysis of raw material and technological parameters
Keywords:
semi-coke, pyrolysis, machine learning, prediction, regression, feature importance, R2, coalAbstract
This paper presents the development and validation of a highly accurate predictive model based on machine learning methods (Gradient Boosting algorithm) for determining semi-coke yield during low-temperature carbonization (pyrolysis). The study’s relevance is driven by the necessity to optimize coal chemistry processes, particularly when processing variable raw materials characteristic of Central Asian deposits (e.g., coals from the Kyrgyz Republic).
A complex set of ten input features was used, including raw material characteristics (coal type, moisture, ash, m1, m2) and technological parameters (temperature, speed, gas flow, tar yield). The model demonstrated high predictive capability on the test sample, achieving a coefficient of determination (R2) of 0.752 and a Mean Absolute Error (MAE) of 4.055%.
A key result is the feature importance analysis, which unequivocally revealed the dominant role of internal raw material characteristics over technological regimes. The parameters m1, type of coal, and m2 exerted the greatest influence on semi-coke yield, with their cumulative contribution exceeding 60%. It is established that process optimization should primarily focus on raw material control and preparation.
The developed model can be integrated into automated production control systems for operational forecasting and strategic planning, thereby enhancing the economic efficiency and stability of the pyrolysis process.
References
Петров, А. Н. Технология переработки твердого топлива. – М.: Химия, 2022. – 450 с.
Джаманбаев, А. Д., Ибраимова, А. М. Технологические особенности коксования углей месторождений Кыргызской Республики // Известия НАН КР. – 2021. – № 4. – С. 101–112.
Kozlov, V. N. Advanced kinetic modeling of coal pyrolysis: limitations and future directions //Fuel Processing Technology. – 2023. – Vol. 248. – P. 107990.
Смирнова, И. А. Прогнозирование теплоты сгорания угля с использованием искусственных нейронных сетей // Известия высших учебных заведений. Горный журнал. – 2024. – № 1. – С. 55–63.
Lee, J. H., Kim, M. S. Predicting coal tar yield in fast pyrolysis using ensemble machine learning models // Energy. – 2021. – Vol. 220. – P. 119642.
Абдыкадыров, Р. С. Анализ элементного состава бурых углей Кара-Кече // Вестник Кыргызско-Российского Славянского университета. – 2020. – Т. 20, № 11. – С. 120–126.
ГОСТ Р 7.0.7–2021. Система стандартов по информации, библиотечному и издательскому делу. Статьи в журналах и сборниках. Издательское оформление.
