Predicting the Electrical Energy Consumption of Electric Arc Furnaces Using Statistical Modeling
Round 1
Reviewer 1 Report
The authors have submitted a very interesting review about the state of the art of statistical modelling of electrical energy consumption in Electric Arc Furnaces (EAF) for steel production.
They have reviewed the existing literature regarding this subject by comparing the methodology and the accuracy of existing published models. Furthermore, the usability of the model is also taken into account. This is an important, often forgotten, factor.
The subject is very convenient and the research has been addressed in a systematic and clear way, only few suggestions are proposed to the authors. The manuscript is well written, although some minor mistakes have been found as indicated below.
I personally welcome very much a review indicating that several important aspects are too often missed in published statistical models:
Sufficient validation with future heats in order to get long-term stability of models Systematic and judicious selection of variables Adequate metrics of model performance The often poor performance of “global” process models when different furnaces may differ significantly. Inclusion of not previously known variables as inputs for the model.Authors detected that all post-2010 papers lack important furnace input variables (lines 564-567) and express that “This could indicate some recent “decoupling” between the fields of statistical modelling and process metallurgy”. The reviewer totally agrees with this impression and suggest to expand this contribution by highlighting the fact that, as found, previous process knowledge is a must in order to obtain meaningful process models. Processes and models cannot be independently developed.
Other comments:
I would suggest to check Worldsteel statistical yearbook for a more updated and accurate value of EAF steel share (2018: 25.5%, 2017: 25.2%, etc.). Annual reports from 1978 to 2018 are ready available at https://www.worldsteel.org/steel-by-topic/statistics/steel-statistical-yearbook.html. Current wording may give the impression that EAF share of steel production will increase by 11% annually, which seems not to be the actual trend. I would suggest reword this sentence in order to clarify that the expected increase refers to the total EAF steel world production. (also in abstract, line 11), it is said that “some input variables” that heavily influence the EE consumption are rarely used in the models”. Would it be interesting to give one or two examples taken from the most often or most evident missing variables at this point? Please, check “x” in Equation 4. Please, check line indentation according to MDPI template. Should “unique delay types” be statically independent? If so, is it the case in real plant practice? Please, consider to complete symbol explanation with shape, scale distribution parameters. “with a standard deviation of only 5 kWh/t” gives the impression to be referring to the standard deviation of EE. Please, check if “standard deviation of error” or “standard deviation of model error” is better. “Using the models together predicts the EE demand per ton produced steel”. Please, consider rewording this sentence. Please check if “preheating energy” is better than “Preheater energy”. Please, consider to reword “The model was compared with models created MLR, SVM, and DT” (based on? using?). Please, check if the coefficient value is a good indicator of variable’s importance? Were all the input variables normalized to the same scale? “The net contribution from oxidation contributes to between 20-50%” Please, improve wording and consider to make a reference to table 2 here. “Of course, this is only sensible if the scrap composition [for each type] is expected to stay the same”. This seems to be referred to the scrap composition of each scrap type, since if the total scrap composition remains constant it is no longer relevant for EE prediction. “The charge mix charged”. Please, improve wording here. “One ton of molten steel contains approximately 390 kWh than solid scrap”. Please, consider that hot metal has possibly much higher Si and C content than scrap which should also be quantified in order to illustrate the effect of hot metal addition. “However, hot metal… is not mentioned in this [current?] study” [26,44-46]. Please, check whether this reference should be placed here or not.356-364. Please, explain how scrap preheating is achieved and why this stage must be separately considered.
“to account for affects that…”. Please, check if “effects” should be used instead. “furnace pressures”. Please, indicate the fluid(s).258 and 580. The point regarding power-on time is whether it is accurately known in advance or not. If it is known, why not use it? I would suggest to clarify this point.
Formatting and English:
Please, check reference formatting along the text (as the space between text and reference in line 20: “2023[1]”) and please, check MDPI referencing style (as in line 211 “the authors in [47]” or in line 121: “in reference [33]”). Check line 173.
Please check subject-verb agreement along the text:
“non-linear models was..” “all of which was taken” “the majority of the terms … is very similar” “The model were one of the few found in literature which were tested on external data”. “the models by Sandberg calculates the yield” “They usually consists of…” “some frameworks only works…” “Only two studies reports…” “are data points that satisfies…” “Non-regular operations includes…” “Transparent models are models that reveals…”Author Response
The answers are in the submitted PDF.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This is a very well written review paper on EE statistical models for EAF systems. The paper covers a good range of literature and compares various literature in a critical and well organised manner. There are a few minor suggestions I have for the authors:
Define all parameters/symbols upon first appearance in equations in the text. Some are shown but not all of them. Table 1 is very useful, but I wonder if you can include citations? Are any of the comparison tables able to be presented in a graphical format to compare a certain numerical metric? That might be more visual to compare and contrast a couple of key numerical metrics. Are 15 studies an exhaustive review? It might be useful to set the scope of the literature survey and discuss any uncertainties of other studies in the wider literature? Any recommendations following the conclusions?
Author Response
The answers are in the PDF
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
The reviewer would like to thank the carefully revision performed and the detailed answers provided by the authors.
