Title: Prediction of electrical energy consumption of CONARC^® furnace using machine learning techniques

Authors: Parul Kota; Albin Rozario Samiraj; Sujoy S. Hazra

Addresses: American Express, American Express India Campus, Harizan Colony, DLF Phase 5, Sector 43, Gurugram, Haryana, 122003, India ' JSW Steel Limited Dolvi Works, Geetapuram, Dolvi, Taluka Pen, District Raigad, Maharashtra, 402107, India ' JSW Steel Limited Dolvi Works, Geetapuram, Dolvi, Taluka Pen, District Raigad, Maharashtra, 402107, India

Abstract: Due to the complex nature and vast number of variables involved in the CONARC^® steelmaking process, predicting the electrical energy (EE) consumption is a non-trivial task. Although many machine learning (ML) techniques have been utilised to predict EE consumption in electric arc furnace (EAF), not much attention has been focused on predicting EE of CONARC^® furnace that uses a variety of raw material in different proportions. In the present work, EE is predicted using various ML algorithms after identifying the relevant variables through a physiochemical model. Random forest (RF) resulted in the best accuracy for the prediction with mean squared error (MSE) of 0.39 having 100 numbers of trees. RF and support vector machine (SVM) were further tuned using boosting method XGBoost (MSE = 0.006) and grid search (0.034) respectively to improve the prediction accuracy.

Keywords: CONARC^®; steel making; electrical energy; statistical modelling; machine learning.

DOI: 10.1504/IJSPM.2022.131563

International Journal of Simulation and Process Modelling, 2022 Vol.19 No.3/4, pp.193 - 203

Received: 05 Aug 2022
Received in revised form: 23 Mar 2023
Accepted: 23 Mar 2023
Published online: 19 Jun 2023 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article