A data-driven model for energy consumption analysis along with sustainable production: A case study in the steel industry. Issue 2 (15th June 2022)

• Record Type: Journal Article
• Authors: Chavosh Nejad, Mohammad; Hadavandi, Esmaeil; Nakhostin, Mohammad Masoud; Mehmanpazir, Farhad
• Is Part Of: Energy sources. Volume 44:Issue 2(2022)
• Page Start: 3360
• Page End: 3380
• Publication Date: 2022-06-15
• Subjects: Energy efficiency -- process optimization -- sustainable production -- data-mining -- scenario analysis; Natural resources -- Periodicals; Energy consumption -- Periodicals; Energy consumption -- Climatic factors -- Periodicals; Energy conversion -- Periodicals; Energy conversion -- Environment aspects -- Periodicals; Power (Mechanics) -- Periodicals; 333.7905
• DOI: 10.1080/15567036.2022.2064943
• Languages: English
• ISSNs: 1556-7036

Abstract

ABSTRACT: Sustainable production is of the most serious concerns that affect production systems. In a manufacturing company, efficient energy consumption, which leads to significant environmental benefits, is an important factor that indicates the performance of sustainable production. This paper proposes a three-stage data-driven model to analyze energy consumption in production systems. The first stage develops energy consumption predictors, the second stage extracts production scenarios, and the last stage predicts the energy consumption for each scenario. We implemented the model in a steel manufacturing plant for investigating the electricity consumption (EC) of Electric Arc Furnace (EAF). First, we developed four groups of predictors where Boosted Neural Network achieved the best result in predicting EAF's electricity consumption (RMSE = 587, R-Squared = 0.859, MAPE = 0.073). Second, we extracted eight distinct production scenarios based on different amounts of input materials through a descriptive data-mining algorithm, K-means. Third, the EC of production scenarios was predicted by the best predictor. Feature analysis showed that Direct Reduced Iron(DRI), ladle age, and scrap grade-3 have the most effect on predicting EC. Scenario analysis illustrated that scenarios with a higher share of DRI cause a higher amount of EC. Contrastingly, input materials with more share of high-grade scrap types lead to more efficient EC.