Artificial neural network modeling for evaluating the power consumption of silicon production in submerged arc furnaces. (5th February 2017)

• Record Type: Journal Article
• Authors: Chen, Zhengjie; Ma, Wenhui; Wei, Kuixian; Wu, Jijun; Li, Shaoyuan; Xie, Keqiang; Lv, Guoqiang
• Is Part Of: Applied thermal engineering. Volume 112(2017:Feb.)
• Page Start: 226
• Page End: 236
• Publication Date: 2017-02-05
• Subjects: Power consumption -- Submerged arc furnace -- Silicon -- Pearson correlation coefficient -- Contour diagrams -- Artificial neural networks; Heat engineering -- Periodicals; Heating -- Equipment and supplies -- Periodicals; Periodicals; 621.40205
• DOI: 10.1016/j.applthermaleng.2016.10.087
• Languages: English
• ISSNs: 1359-4311

Abstract

Highlights: The effect of metal oxide on power consumption was studied. Pearson correlation coefficient between oxide contents and power consumption was investigated. The interactive effects among the main oxide matter and remaining trace oxide matter was studied by the contour diagrams. Artificial neural network modeling was used for evaluating the power consumption of silicon production. Abstract: The Pearson correlation coefficient between different quantities of metal oxides and specific power consumption was used here to determine the effect of metal oxide content on the power consumed by an industrial silicon production process. The results showed that the effect of oxide content on power consumption falls into the order CaO > Fe2 O3 > Al2 O3 . The interactive effects among the main oxide matter (CaO, Fe2 O3, and Al2 O3 ) and remaining trace oxide matter (MgO, K2 O, TiO2, Cr2 O3, and NiO) of raw materials were also analyzed via contour diagrams; the results showed that the dominant metal oxides play a much more important role in power consumption than any of the trace oxides. The oxide content of the charge raw material critically affects the specific power consumption and electrical energy costs of the submerged arc furnace (SAF), and can be reduced by appropriately taking them into account. ANN (artificial neural network) modeling was used to evaluate the power consumption of silicon production in a typical SAF. The value R 2 = 0.80 of the neural network indicates that 80% of the variation in specific power consumption can be accounted for via the proposed model.