Using Statistical Modeling to Predict the Electrical Energy Consumption of an Electric Arc Furnace Producing Stainless Steel
Round 1
Reviewer 1 Report
Thank you for an interesting and very thorough paper, which presents the used metods with enough information and details to repeat the study with another set of data. And I also see here a more wider possibility to use the presented method with data from completely different processes. In my opinion the paper can be accepted after a minor spell check.
Author Response
See attached file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
A graph comparing the predicted (y-axis) to the actual (x-axis) electrical energy consumption should be provided, to better visualise the prediction accuracy.
Accuaries and error standard deviations could be provided as kWh/t in addition to kWh/heat
Author Response
See attached file.
Author Response File: Author Response.pdf
Reviewer 3 Report
The techniques used for the statistical modelling of the electric energy consumption of an EAF are well chosen. The statistical modelling has been conducted and evaluated well. The modelling delivered valid results for the prediction of the EE consumption of a test set of heats and the prediction behaviour was investigated by a number of statistical tools.
The references are up-to-date and appropriate. There are very current references, even from 2019, but also references to older fundamental publications, giving a good picture of publications on the paper topic.
However, the paper could be improved by including the very important information, what packages or libraries (and which version) were used for the statistical model and tests (NumPy, SciPy, Pandas, Statsmodels, R or others?). Since Python is only a very generic programming language, the information given in A1 is on one hand not sufficient regarding software and on the other hand irrelevant (hardware specs) to the work conducted.
Another point that has to be improved is in my opinion the phrasing regarding the chemical variables. Section 3.2.1 currently reads as if e.g. metallic Cr is the percentage of charged metallic material, while e.g. Cr2O3 is charged in the given percentage as oxide. This of course would be completely illogical considering the EAF process. No EAF steel plant knows the exact composition of charged metals and oxides. Metalls especially like Cr are partly oxidised by oxygen and end up in the slag as oxides. A big part of the oxides in the slag was charged as metall and not as oxide as the phrasing currently implies. Only later on in the paper it becomes clear that the variables are given by slag and steel analysis presumably at tapping. So these variables are only known after the end of the process.
In section 2.3 it is stated that linear statistical models are sub-optimal tools for the prediction of EE consumption. Never the less, at least [35] and [36] report linear statistical models with R² of more than 0.9, which would be significantly better than the reported results of this paper. Since the paper is already quite long this should at least be mentioned in the discussion and a comparison with an MLR based model could be put in the outlook.
In the paper the phrase future heats is used when meaning the test dataset. The use of this phrase should be reconsidered or it should be discussed how a model can predict future heats when variables like metallic and oxide elements but also delays and therefore TTT are not available ex ante but only ex post. How would these variables be estimated for heats that have not been melted yet?
Finally, there is one linguistic error that is occuring regularly throughout the whole paper that should be corrected urgently: plural followed by 3rd person singular form e.g. line 98 "Statistical models, [...], acts on [...]". This kind of error can also be found on line 101, 114, 146, 154, 230, 255, 267, 273, 274, 286, 302, 308, 311, 319 and many more.
Author Response
See attached file.
Author Response File: Author Response.pdf
