Design of Multi-Source Fusion Wireless Acquisition System for Grid-Forming SVG Device Valve Hall

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Aiming at the requirement for status monitoring of the GFM-SVG valve hall, this paper designs a multi-source fusion wireless acquisition system. The constructed system possesses certain engineering application value. The following are some suggestions.

1. The designed system appears to be applicable to other types of converters. Please explain whether the system constructed in the paper is targeted specifically for GFM-SVG. If so, please elaborate on how the proposed system adapts to the application requirements of GFM-SVG. These requirements should differ from that of GFL-SVG.
2. The two contributions mentioned in the introduction are identical. Please verify whether it is a duplicate writing.
3. Although the sensing system constructed in the paper involves multi-type physical quantities, it is not clear how these multi-physical quantities are fused and how they support the operation of the GFM-SVG. Please supplement the relevant description.
4. The data fusion part of the paper lacks equation descriptions. Please supplement the core equations concerning the design of the fusion wireless acquisition system.

Author Response

1. The designed system appears to be applicable to other types of converters. Please explain whether the system constructed in the paper is targeted specifically for GFM-SVG. If so, please elaborate on how the proposed system adapts to the application requirements of GFM-SVG. These requirements should differ from that of GFL-SVG.

 

REPLY: Thank you for suggesting us these works. The proposed system is an exclusive monitoring solution for the GFM‑SVG valve hall. Detailed adaptation logic and the distinctions from the GFL‑SVG have been added in the newly revised Section 4.1 “GFM‑SVG‑Specific Adaptation Design.”

In particular, the differences are explained from three perspectives—operating mode, fault risk, and maintenance requirements. Moreover, additional descriptions of the hardware EMI‑resistance design and the algorithmic dynamic‑weight allocation mechanism customized for GFM‑SVG operation have been included. These revisions comprehensively address the reviewer’s request.

 

 

2. The two contributions mentioned in the introduction are identical. Please verify whether it is a duplicate writing.

 

REPLY: Thank you for suggesting us these works. After verification, the two contributions in the Introduction were indeed repeated. The second contribution has been revised and refined. The two contributions now focus respectively on:

• 1) Development of a heterogeneous wireless sensing network. A WSN integrating temperature, acoustic, and visual sensing nodes is designed. An EMI‑resilient communication scheme is proposed to ensure reliable collection and stable trans-mission of multi‑physics information within the valve hall.
• 2) Proposal of an improved Dempster–Shafer (D‑S) evidence‑theory–based fusion model. The proposed multi‑source information‑fusion fault‑diagnosis algorithm effectively resolves data uncertainty and evidence conflict across heterogeneous sensing modalities, achieving high‑accuracy and high‑reliability diagnosis of ab-normal sound, local overheating, coolant leakage, and compound faults.

 

This revision clearly distinguishes the technical innovations and eliminates redundant expressions. The specific changes can be found in the Introduction section under “The main contributions of this paper are as follows.”

 

3. Although the sensing system constructed in the paper involves multi-type physical quantities, it is not clear how these multi-physical quantities are fused and how they support the operation of the GFM-SVG. Please supplement the relevant description.

 

REPLY: Thank you for suggesting us these works. A detailed explanation has been added in the newly revised Section 6.2 “Construction of the Multi‑Source Information Fusion Algorithm.”

The section now specifies a three‑level fusion framework combining accurate visual recognition with infrared and acoustic cross‑verification. It elaborates the complete process of single‑sensor feature extraction, dynamic weight assignment, and evidence‑conflict handling, and further clarifies how multi‑physics information collaboratively supports fault diagnosis in GFM‑SVG systems. This addition fully improves the fusion mechanism and operational support logic.

 

4. The data fusion part of the paper lacks equation descriptions. Please supplement the core equations concerning the design of the fusion wireless acquisition system.

 

REPLY: Thank you for suggesting us these works. The core equations have been added in Sections 6.1 and 6.2, covering both the YOLO‑based model and the proposed fusion algorithm. These include:

the combined loss function (Equation 1);

the CIoU bounding‑box regression loss (Equation 2);

the visual‑evidence confidence formulation (Equation 3);

the infrared/acoustic‑evidence confidence calculation (Equation 4);

the conflict factor (Equation 5); and

the composite confidence synthesis (Equation 6).

Together, these additions provide a complete mathematical representation of the data‑fusion model and ensure algorithmic reproducibility.

Reviewer 2 Report

Comments and Suggestions for Authors

This paper describes the design and implementation of a multi-source fusion wireless data acquisition system for Grid-Forming Static Var Generator (GFM-SVG) valve halls; integrating temperature, acoustic, visual, and infrared sensors into a heterogeneous wireless sensor network (WSN) that can operate under strong electromagnetic interference (EMI) conditions. The results are demonstrated through experiments of high diagnostic accuracy (up to 98.5%) and low packet loss rates (less than 1.5%), showing practical robustness and effectiveness in enhancing the safety and stability of grid-forming SVG operations.

The flaws and limitations identified are:

1) The system's image acquisition resolution (320x240 pixels at 2 frames per second) balances real-time performance and bandwidth but may limit detail for some complex visual fault detection.

2) While the study comprehensively covers multi-source sensor fusion, the scale and scope are limited to valve hall environments of grid-forming SVG devices and may not generalize directly to other types of equipment or power systems without further adaptation.

3) The approach depends heavily on accurate sensor deployment and calibration, which can be challenging in real industrial environments with harsh conditions.

4) The fault diagnosis relies on thresholds and rules that may require tuning or adaptation for varying operational conditions or new types of faults not considered in the current study.

5) The paper focuses on experimental validation but does not detail long-term field testing or integration challenges with existing industrial control systems.

6) The paper language requires complete overhaul by a technical English expert as the present version is hardly comprehensible.

Comments on the Quality of English Language

The manuscript requires substantial language editing to improve grammar, sentence structure, and clarity. Several sections are difficult to comprehend due to poor English, which impacts the effective communication of the scientific content.

Author Response

1. The system's image acquisition resolution (320x240 pixels at 2 frames per second) balances real-time performance and bandwidth but may limit detail for some complex visual fault detection.

REPLY: Thank you for suggesting us these works. Thank you for pointing out this issue. Considering the need for complex visual‑fault detection, a supplementary note has been added in Section 5.1, specifying that the system supports dynamic resolution adjustment from 320 × 240 to 1600 × 1200 pixels. In practice, the resolution can be flexibly switched according to the monitoring requirements: high‑resolution mode is enabled when fine‑grained defect identification is required, ensuring a balance between real‑time response and detection accuracy. The corresponding revision has been incorporated into the camera‑data acquisition section.

 

2. While the study comprehensively covers multi-source sensor fusion, the scale and scope are limited to valve hall environments of grid-forming SVG devices and may not generalize directly to other types of equipment or power systems without further adaptation.

REPLY: Thank you for suggesting us these works. This limitation has been explicitly addressed in the Conclusion. A new paragraph clarifies the current shortcomings of system scalability and describes the adaptation process required for extending the system to other power‑equipment types such as transformers, switchgear, and photovoltaic inverters. In addition, Section 8 (Future Work) further elaborates on the proposed expansion and standardization directions to respond to the reviewer’s concern regarding research‑scope limitations.

 

3. The approach depends heavily on accurate sensor deployment and calibration, which can be challenging in real industrial environments with harsh conditions.

REPLY: Thank you for suggesting us these works. To address the difficulties of accurate sensor deployment and calibration under harsh industrial conditions, targeted solutions have been added in Section 4.2 “Sensor Deployment and Multi‑Source Data Calibration”, forming a complete support framework in conjunction with the hardware‑adaptation design in Section 4.1.

Simplified deployment strategy: The system adopts a “3D spatial pre‑setting + CAD‑assisted modeling” approach, where the three‑dimensional coordinates of sensors are predefined in the valve‑hall CAD model. This method ensures precise alignment between monitoring points and target devices, minimizes on‑site adjustments, and effectively reduces deployment complexity while maintaining positioning accuracy.

Optimized calibration procedure: Relying on industrial‑grade high‑precision sensors described in Section 4.1, the calibration process is simplified and can be completed without additional complex tooling. A multi‑sensor spatiotemporal‑calibration mechanism is also integrated: each STM32 microcontroller automatically time‑stamps every frame at millisecond precision. The JSON‑formatted data include a unique device ID and 3D coordinates; the host computer subsequently performs sequential sorting and spatial matching, achieving inter‑sensor synchronization with a time deviation ≤ 100 ms. This process both verifies and ensures the calibration accuracy, minimizing repeated on‑site adjustments.

This three‑tier solution—deployment simplification + hardware reliability + calibration verification—effectively resolves the practical challenges of sensor positioning and calibration in harsh industrial settings. All details are described in Sections 4.1 and 4.2 to confirm the system’s engineering feasibility and reliability.

 

4. The fault diagnosis relies on thresholds and rules that may require tuning or adaptation for varying operational conditions or new types of faults not considered in the current study.

REPLY: . Thank you for suggesting us these works. An additional dynamic adjustment mechanism has been introduced in Section 4.1 (Algorithm‑Adaptation Design). The system now dynamically modifies warning thresholds according to GFM‑SVG operating load and introduces a transient‑compensation coefficient to handle disturbances caused by operating‑condition fluctuations, improving robustness under varying scenarios. Furthermore, the need for adaptation to new fault types is clearly mentioned in the Conclusion, and future work proposes employing transfer‑learning and few‑shot learning techniques to enhance recognition of emerging faults not yet covered by the knowledge base.

 

5. The paper focuses on experimental validation but does not detail long-term field testing or integration challenges with existing industrial control systems.

REPLY: Thank you for suggesting us these works. Details regarding the system’s long‑term field testing in a 500 kV substation valve hall have been added in Section 7.2 “Performance Testing and Analysis of the Multi‑Source Information Fusion System.” In addition, the Conclusion now includes a discussion of potential integration challenges with existing industrial control systems. Section 4.2 also elaborates on the system’s spatiotemporal synchronization technology to ensure fusion precision for continuous long‑term operation.

 

6. The paper language requires complete overhaul by a technical English expert as the present version is hardly comprehensible.

REPLY: Thank you for suggesting us these works. A comprehensive language revision has been carried out by a professional scientific‑English editor. The revision corrected grammar, sentence structure, and technical terminology for consistency and clarity. Simultaneously, the Chinese version was refined for fluency and logical coherence to avoid ambiguity and enhance overall readability of the manuscript.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

no further comment

Author Response

Thank you for your recognition of our work!

Reviewer 2 Report

Comments and Suggestions for Authors

The paper has been improved after the first round of revision. However, some additional tasks are still needed.

Fig. 3 blocks inside the layers are hardly visible. Please change the colour configuration for a clear visibility.

Line 140-149: Please number or use bullets for these four ''perception-transmission-processing-application'' layers for clarity.

 

Author Response

Thank you for your recognition of our work. We have revised the paper according to your suggestions. Thank you!