Next Article in Journal
Critical Review of Advances and Numerical Modeling in Absorbers and Desorbers of Absorption Chillers: CFD Applications, Constraints, and Future Prospects
Previous Article in Journal
Efficient Production of Fuel Ethanol via the Simultaneous Use of Distillery Stillage Biomass and Beet Molasses
 
 
Article
Peer-Review Record

Dynamic Spatial–Temporal Graph Neural Network for Cooling Capacity Prediction in HVDC Systems

Energies 2025, 18(2), 313; https://doi.org/10.3390/en18020313
by Hao Sun 1, Shaosen Li 1, Jianxiang Huang 1, Hao Li 1, Guanxin Jing 1, Ye Tao 1 and Xincui Tian 2,*
Reviewer 1:
Reviewer 2: Anonymous
Energies 2025, 18(2), 313; https://doi.org/10.3390/en18020313
Submission received: 13 December 2024 / Revised: 30 December 2024 / Accepted: 7 January 2025 / Published: 12 January 2025
(This article belongs to the Topic Advances in Power Science and Technology, 2nd Edition)

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript explores a hybrid framework integrating Graph Neural Networks (GNNs) with temporal dynamics to predict the cooling capacity of converter valves in HVDC systems. The approach attempts to address spatial and temporal dependencies simultaneously using graph-based structures and attention mechanisms. While the work focuses on a relevant problem in predictive maintenance, several key issues require attention.

- The manuscript does not clearly articulate its novelty compared to existing methods in spatial-temporal prediction. The proposed integration of GNNs and temporal dynamics appears incremental rather than groundbreaking. 

- The introduction and related work sections do not sufficiently differentiate the current work from prior studies, particularly those employing Graph Attention Networks (GATs) or Temporal Graph Neural Networks (TGCNs). 

- A significant portion of the references cited in the manuscript is relatively outdated, with many papers dating back several years. Given the rapid developments in graph neural networks and spatial-temporal modeling, the authors must incorporate more recent studies (post-2020) to reflect the state of the art and situate their work in the current research landscape. 

- Some figures, such as Figure 2, are not readable, which makes it challenging to interpret the presented results effectively. The resolution and labeling of figures should be improved to ensure clarity. 

- Additional visualizations (e.g., heatmaps of graph attention weights) could help clarify how relationships between variables influence predictions, enhancing the interpretability of the proposed framework. 

- The adjacency matrix is based on Pearson correlation, which may not effectively capture non-linear or dynamic relationships between variables. This limitation should be addressed, or alternative methods for dynamic graph learning should be explored. 

- The dataset used for validation is limited to a specific HVDC system. The authors should discuss the generalizability of their model to other systems or larger datasets. 

- Scalability remains a concern, as the computational complexity of the model increases with larger graphs and time-series windows. This aspect requires further analysis. 

 

To conclude, the manuscript has potential but requires significant revisions to clearly establish its novelty, improve figure clarity, update references, and address concerns regarding graph construction and scalability.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The authors propose a novel framework that integrates Graph Neural Networks (GNN) with Temporal Dynamics for cooling capacity prediction in HVDC systems.

 

Questions:

 

1. There are many arXiv preprint references. ArXiv preprint files have not gone through a peer-review process and have not yet been accepted by the editorial board. Consider presenting more previously published work.

 

2. It is unclear why the integrated use of Graph Convolutional Network, Graph Attention Networks and Long Short-Term Memory improves the performance of cooling capacity prediction.

 

3. It is not clear what the input variables of the proposed machine learning method are. Is there a selection or are there multiple types of input?

 

4. There is little information about the test system. How was this database generated? How big is the database? What is the cooling capacity distribution of the database?

 

5. The proposed method consists of many parameters that must be defined by users. What were the values ??of these parameters in the case studies carried out by the authors?

 

6. Figures 2(a), 2(b) and 2(c) are too small in the paper and they depict important results. Please enlarge them a bit.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have addressed all the issues indicated in the first-round review. Therefore I recommend accepting the papaer.

Reviewer 2 Report

Comments and Suggestions for Authors

The authors propose a novel framework that integrates Graph Neural Networks (GNN) with Temporal Dynamics for cooling capacity prediction in HVDC systems.

 

The article has been improved, the contribution is good and all questions have been effectively answered.

Back to TopTop