An Open Industrial Energy Dataset with Asset-Level Measurements and High-Coverage 15-Minute Aggregates from a Manufacturing Facility

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript presents asset-level electrical energy measurements collected from a medium-scale manufacturing facility over an approximately one-year observation period in 2025.

The data is original and well defined. The authors have addressed all relevant aspects of data collection. The data is valuable insights for research in industrial energy analytics, load profiling, forecasting, anomaly detection, and data-quality assessment. The dataset is technically sound, well-documented, and ready for publication.

Author Response

The authors thank the reviewer for their positive evaluation of the manuscript. No specific revisions were requested.

Reviewer 2 Report

Comments and Suggestions for Authors

This data descriptor presents a valuable and well-documented contribution to the field of industrial energy informatics. The authors provide a comprehensive, real-world dataset of asset-level electrical measurements from a medium-scale manufacturing facility, addressing a notable gap in publicly available resources. The key strengths of the work are its transparency and practical representativeness. The manuscript meticulously details the "warts-and-all" nature of retrofit industrial monitoring—including heterogeneous meter fleets, mixed communication pathways, irregular sampling, staged commissioning, and documented instrumentation faults—rather than presenting a sanitized, idealized dataset. This makes it an excellent benchmark for developing robust analytical methods. The description of the medallion (Bronze-Silver-Gold) data processing architecture is clear and promotes reproducibility. The dataset itself, with 43 assets, ~1 million 15-minute intervals, and 2.96 GWh of energy, is substantial and accompanied by explicit data quality metrics (coverage, reliability flags), significantly enhancing its utility for research in load profiling, anomaly detection, and forecasting under imperfect data conditions.

1. The in-text references to figures, particularly Figure 6 and Figure 7, are somewhat ambiguous. The text mentions "Fig.6" but the caption in the provided document only shows a single plot without subfigure labels (a, b, etc.). For Figure 7, the text states it illustrates the distribution of data reliability, but does not explain what the two main panels (a and b, shown in the document) represent specifically (e.g., is (a) Data Coverage and (b) Reliable Coverage?). To prevent reader confusion, please ensure all figure citations precisely match the captions and provide a brief description in the text of what each subfigure depicts.
2. Section 2.7 describes a topology verification analysis, noting that many parent-child relationships have high violation rates of energy containment, which is correctly attributed to partial sub-metering. This is a critical and honest finding. To strengthen this section, consider adding a brief summary statement on the practical implication of this result. For example, explicitly state that therefore, the dataset should not be used as a benchmark for testing energy balance or disaggregation algorithms that assume a perfectly measured, closed hierarchy, but rather for studies where such real-world incompleteness is a key factor.
3. The solar balance model analysis in Section 2.7 is insightful. The description states the solar distribution meter signal is approximated as "downstream heat pump demand offset by photovoltaic generation." To be physically precise, consider rephrasing to: "the solar distribution meter reading is consistent with representing the net flowat that boundary, calculated as downstream heat pump demand minus​ photovoltaic generation" or similar, to accurately reflect the sign convention (generation is negative demand).

Author Response

Comment 1: 'The in-text references to figures, particularly Figure 6 and Figure 7, are somewhat ambiguous. The text mentions "Fig.6" but the caption in the provided document only shows a single plot without subfigure labels (a, b, etc.). For Figure 7, the text states it illustrates the distribution of data reliability, but does not explain what the two main panels (a and b, shown in the document) represent specifically (e.g., is (a) Data Coverage and (b) Reliable Coverage?). To prevent reader confusion, please ensure all figure citations precisely match the captions and provide a brief description in the text of what each subfigure depicts.'.

Response 1: the authors thank the reviewer for this observation. Figures 6 and 7 have been revised to improve clarity and consistency between in-text references and figure captions. For Figure 6, the caption has been updated to explicitly describe the multi-panel structure, clarifying that each horizontally stacked panel corresponds to a distinct asset. The in-text discussion has also been revised to focus on the interpretation of characteristic load behaviours across asset types. For Figure 7, both the caption and in-text description have been expanded to explicitly define the structure of the heatmap, specifying that rows represent individual assets and columns represent calendar dates, with colour intensity indicating the fraction of reliable 15-minute windows. This eliminates ambiguity regarding the visual structure and ensures consistency between the figure and its description.

 

Comment 2: "Section 2.7 describes a topology verification analysis, noting that many parent-child relationships have high violation rates of energy containment, which is correctly attributed to partial sub-metering. This is a critical and honest finding. To strengthen this section, consider adding a brief summary statement on the practical implication of this result. For example, explicitly state that therefore, the dataset should not be used as a benchmark for testing energy balance or disaggregation algorithms that assume a perfectly measured, closed hierarchy, but rather for studies where such real-world incompleteness is a key factor.".

Response 2: the authors thank the reviewer for highlighting the importance of clarifying this point. A summary statement has been added to Section 2.7 to explicitly describe the practical implications of partial submetering and incomplete hierarchical observability. The revised text now clarifies that the dataset should not be treated as a closed or fully balanced energy system and is therefore not suitable as a benchmark for methods assuming strict energy conservation across fully observed hierarchical boundaries. Instead, the dataset is now explicitly positioned as appropriate for evaluating analytical methods under realistic industrial conditions, where partial observability and instrumentation limitations are inherent.

 

Comment 3: "The solar balance model analysis in Section 2.7 is insightful. The description states the solar distribution meter signal is approximated as "downstream heat pump demand offset by photovoltaic generation." To be physically precise, consider rephrasing to: "the solar distribution meter reading is consistent with representing the net flowat that boundary, calculated as downstream heat pump demand minus​ photovoltaic generation" or similar, to accurately reflect the sign convention (generation is negative demand).".

Response 3: the authors thank the reviewer for this suggestion. The description of the solar distribution meter has been revised to reflect a physically precise formulation. The updated text now states that the solar distribution meter reading is consistent with representing the net power flow at the boundary, approximated as downstream heat pump demand minus photovoltaic generation, with generation treated as negative demand under the adopted sign convention. This revision aims to improve clarity and ensure correct physical interpretation.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have addressed all my concerns, the paper can be accepted for publication.