Holographic Sound Field Inspection Instrument — Multi-Scenario Power Industry Application Solution

A Panoramic Acoustic Monitoring Solution Based on a 64-Element Omnidirectional Array and AI Panoramic Sound Field Reconstruction

I. Industry Pain Points: Common Limitations of Acoustic Monitoring for Power Equipment

Partial discharge (PD) is one of the most important early indicators of insulation failure in power equipment. Detection methods commonly used in the industry today include the ultra-high frequency (UHF) method, the acoustic emission (AE) method, the high-frequency current transformer (HFCT) method, and the transient earth voltage (TEV) method, whose operating bands lie respectively in the ultra-high frequency (approximately 300 MHz–3 GHz), ultrasonic (approximately 20 kHz–300 kHz), high-frequency current (approximately 0.1 MHz–50 MHz), and transient earth voltage (approximately 3 MHz–100 MHz) ranges. These bands essentially do not overlap the 20 Hz–20 kHz audible-sound monitoring bandwidth of the array used in this solution, and each method has its own technical strengths in specific scenarios. With respect to the technical direction of acoustic monitoring (the audible-sound band), the industry generally faces the following common limitations:

• Limited field of view, with a tendency toward blind spots. Handheld devices and fixed sensors mostly detect within a limited angular range. In large, multi-equipment scenarios such as substations and converter stations, a single device has difficulty covering acoustic events in all directions, and there may be a risk of missed detection.

• Weak signals are difficult to capture in interference-heavy environments. Strong electromagnetic interference, overlapping noise from multiple devices, and spatial reverberation are present on site. Conventional beamforming algorithms tend to lose localization accuracy at low signal-to-noise ratios, and extracting the weak audible-sound component of partial discharge from background noise poses a certain degree of difficulty.

• High reliance on operator experience. Detection results depend largely on the operator’s judgment; the establishment of standardized abnormal-sound identification models is still being refined, and there is room to raise the degree of closed-loop intelligence across “monitoring → identification → early warning.”

• Intermittent inspection leaves “blind windows.” Handheld-device inspection as the primary mode leaves intervals between inspections. A fault may occur and evolve between two inspections, which is at odds with the preventive-maintenance objective.

Note: The common industry limitations described in this section are summarized from publicly available technical literature and industry practice. They are not directed at the products or solutions of any particular manufacturer and do not constitute an evaluation or disparagement of the goods or services of others.

II. Solution Overview: Near-Omnidirectional Panoramic Acoustic Monitoring from a Single Device

Built around a 64-element omnidirectional, high-SNR microphone array, the Holographic Sound Field Inspection Instrument can achieve near-omnidirectional panoramic sound field coverage of the entire spherical space (360° horizontal, 180° vertical) from a single unit, and uses AI algorithms to complete full-chain intelligent monitoring—“panoramic acquisition → source separation → anomaly identification → risk warning” (each stage of this capability is a design target). Relative to conventional technical approaches, its design differences are as follows:

Table 1. Comparison of Design Differences in Technical Approach

The core technologies of this solution are currently undergoing adaptation research for typical scenarios such as substations, converter stations, transmission lines, and distribution station rooms.

Note: This table is a comparative statement at the technical-approach level, compiled from publicly available information and not directed at the products of any particular manufacturer. The actual application performance in each scenario must be determined after on-site adaptation and verification.

III. Core Technologies

3.1 64-Element Omnidirectional Array: Near-Omnidirectional Panoramic Coverage

A spherical microphone array design is adopted, dispensing with any mechanical pan-tilt rotation, to achieve near-omnidirectional coverage of the entire spherical space. A single deployment panoramically maps all acoustic hotspots, offering an improved approach to the “limited field of view, tendency toward blind spots” problem of conventional acoustic imaging cameras (design target).

3.2 AI Panoramic Sound Field Reconstruction

Based on a hierarchical multi-task deep learning model, spherical harmonic domain features of different orders are assigned to different frequency layers to achieve high-fidelity reconstruction of broadband panoramic sound fields (design target). A single device can localize medium- and long-range sound sources (design target), with localization accuracy improving as frequency rises.

3.3 AI Multi-Source Blind Separation

For complex scenarios in which multiple devices in a substation emit sound simultaneously, a deep neural network source separation algorithm achieves multi-source separation from mixed acoustic signals without prior information (design target), with a design separation signal-to-interference ratio of ≥ 12 dB (see Note 3 for test conditions), supporting the separation and tracking of abnormal sounds from different devices.

3.4 Adaptation to Strong Electromagnetic Environments in Power Grids

To address the specific demands of substations—strong electromagnetic interference, wide-temperature-range operation, and outdoor deployment—electromagnetic shielding and filter circuit design are used to suppress electromagnetic coupling interference and reduce the effect of power-frequency and high-frequency electromagnetic interference on the acoustic acquisition chain. It should be noted that the 50 Hz power frequency in substations, the 100/200 Hz vibration of transformer bodies, and cooling fan noise all fall within the 20 Hz–20 kHz operating band of this solution. These constitute acoustic interference and are handled by the AI multi-source blind separation of Section 3.3 at the signal level (design SIR ≥ 12 dB), rather than suppressed by filter circuits. In addition, wide-temperature-range adaptive design ensures engineering reliability of the system between −20 °C and 50 °C (design target; the specific temperature range is subject to measured calibration).

3.5 Acoustic-Optical Multimodal Fusion

Sound field imaging is overlaid on panoramic optical images to present a visualized acoustic monitoring view. Fusion with multimodal data such as infrared thermal imaging is also supported, underpinning the construction of power grid digital twins.

IV. Multi-Scenario Power Industry Applications

Table 2. Multi-Scenario Adaptation Notes for the Power Industry (Design Targets)

Note: This table is a design statement of technical capability. It does not constitute a commitment regarding detection conclusions or engineering outcomes in any specific scenario; the actual applicability in each scenario must be determined after on-site adaptation and verification.

V. Typical Inspection Objects and Fault Types

Table 3. Typical Inspection Objects and Acoustic Anomalies (Inspection reference value is a design target)

By building an abnormal acoustic signature library for power grid equipment that covers multiple fault types, operating conditions, and SNR levels, the system supports vertical transfer and cross-scenario generalization of the AI diagnostic model (design target).

VI. Technical Parameters and Performance Specifications

Note 1. Sound field reconstruction, localization, and separation are design specifications of the system hardware and algorithms and do not constitute guaranteed on-site values. Among these, the sound field reconstruction angular resolution, SNR, and sensitivity all vary with frequency and test distance—constrained by the diffraction limit of the array aperture, resolution degrades as frequency decreases. The values in the table above are design target values under specified frequency bands and test conditions and must be calibrated against actual test conditions before use.

Note 2. Regarding the applicable scope of partial discharge detection—the main energy of partial discharge acoustic emission is concentrated in the ultrasonic band (approximately 20 kHz–300 kHz), whereas the upper bandwidth limit of the array in this solution is 20 kHz. Accordingly, this solution covers the audible-sound and low-ultrasonic boundary acoustic components of PD and can provide auxiliary criteria at the acoustic monitoring level. It is recommended for use in conjunction with the AE ultrasonic method, the UHF method, the TEV method, and others, and does not replace ultrasonic partial discharge detection. The actual applicability for each equipment type must be verified through on-site comparative testing.

Note 3. The actual values of the separation SIR, array SNR, sensitivity, angular resolution, and other specifications vary with frequency, distance, sound source type, and on-site SNR conditions. The values in the table above are design target values under specific test conditions. It is recommended that they be cited only after a qualified testing institution has issued a test report, with the report number and issuing institution indicated.

Note 4. Regarding the cost estimate—the statements “reduce the number of sensors deployed” and “lower O&M costs” are based on an internal estimation model for typical scenarios (basis of estimate: [To be filled: estimation method, assumptions, report number]). They are internal estimation results; actual outcomes vary with station scale, monitoring scope, and acceptance requirements, and do not constitute a commitment.

VII. Application Value

7.1 Economic Value: Helping Reduce Total Cost of Ownership for Station-Wide Monitoring

A single Holographic Sound Field Inspection Instrument can achieve 360° horizontal coverage, so that a substation’s key monitoring area typically requires only 2 units (mutually redundant) to achieve panoramic monitoring; the specific number may be determined flexibly according to station scale, monitoring scope, and the division of key areas. Compared with fixed multi-point partial discharge monitoring systems, which typically require 6–8 nodes, internal estimates indicate room for a reduction in long-term O&M costs (see Note 4 for the basis of the estimate; actual results vary with station and do not constitute a commitment).

7.2 Early-Warning Value: From “Post-Failure Repair” to “Advance Warning”

Panoramic real-time online monitoring can significantly shorten inspection intervals and reduce fault blind windows. Through AI anomaly identification, it is designed to support early warning of faults (design target: 2–4 weeks in advance; not a guaranteed on-site value), helping to reduce the economic losses caused by unplanned downtime.

7.3 Efficiency Value: From “Operator Experience” to “Intelligent Closed Loop”

Establishing a full-chain intelligent monitoring paradigm—“panoramic acquisition → blind source separation → anomaly identification → risk warning”—upgrades inspection, which has relied heavily on operator experience, to standardized intelligent monitoring, improving O&M efficiency and consistency (design target).

VIII. Summary

Taking “omnidirectional array + AI panoramic sound field reconstruction + multi-source blind separation + intelligent anomaly identification” as its core technical approach, and addressing the application needs of panoramic acoustic monitoring for power equipment, the Holographic Sound Field Inspection Instrument provides a panoramic acoustic monitoring solution with single-unit near-omnidirectional coverage, strong-interference decoupling, and intelligent early warning for substations, converter stations, transmission lines, distribution station rooms, and other scenarios (design target).

The core technologies of this solution are currently undergoing adaptation research for typical scenarios such as substations, converter stations, transmission lines, and distribution station rooms—evolving from “limited field of view” toward “near-omnidirectional coverage,” from “operator experience” toward an “intelligent closed loop,” and from “intermittent inspection” toward “real-time online” monitoring. It provides a technical path reference for intelligent condition monitoring and digital O&M of power grid equipment and has application prospects in the construction of the new-type power system.

Hear the subtle, perceive the profound.

Compliance Statement

1. This material is product technical promotional material (advertisement). The technical specifications, performance data, and benefit estimates it contains are design specifications or internal estimation results and do not constitute guaranteed on-site values, product commitments, or contractual terms.

2. The technical-approach comparisons in this material are compiled from publicly available information. They are not directed at the products of any particular manufacturer and do not constitute an evaluation of the goods or services of others.

3. The descriptions of detection methods in this material are intended solely for technical exchange. Actual testing shall follow the applicable national and industry standards and shall be carried out by qualified institutions.

4. For the current status of the standards cited in this material, the officially issued text shall prevail. This solution makes no claim of conformity to any standard.

5. The actual functions and performance of the product shall be governed by the technical agreement, inspection report, and contract of the final delivered product.

© Beijing GroupCore Acoustic Technology Co., Ltd. This document is an internally revised version. Do not reproduce without permission.