A multi-modal sensor-based power equipment partial discharge monitoring and positioning system

Abstract

The application relates to the technical field of partial discharge of power equipment, and particularly discloses a power equipment partial discharge monitoring and positioning system based on multi-modal sensing, which comprises a multi-modal sensing unit configured to be arranged around power equipment to be monitored and used for synchronously collecting various physical signals generated by partial discharge, wherein the multi-modal sensing unit at least comprises an ultrahigh frequency sensing array used for receiving ultrahigh frequency electromagnetic wave signals generated by partial discharge, and an ultrasonic sensing array used for receiving ultrasonic signals generated by partial discharge; and a data acquisition and synchronization unit connected with the multi-modal sensing unit. The power equipment partial discharge monitoring and positioning system based on multi-modal sensing effectively solves the problems of insufficient multi-modal cooperation, low time synchronization precision and lack of closed-loop optimization mechanism in the existing partial discharge monitoring by constructing an integrated architecture of multi-modal sensing cooperation + high-precision time synchronization.

Description

Technical Field

[0001] This invention relates to the field of partial discharge technology for power equipment, specifically to a partial discharge monitoring and location system for power equipment based on multimodal sensing. Background Technology

[0002] Power equipment (such as gas-insulated switchgear, power transformers, and high-voltage switchgear) is a core infrastructure ensuring the safe and stable operation of the power grid, and its insulation performance directly determines the reliability of the grid operation. Partial discharge is a core cause of insulation degradation in power equipment: in areas of concentrated electric field inside the equipment (such as inside bushings, winding joints, and disconnector contacts), partial breakdown discharge can occur due to insulation defects (such as bubbles, floating potential, and surface contamination). This discharge process continuously erodes the insulating medium (such as SF6 gas decomposition, transformer oil aging, and epoxy resin cracking). If it is not monitored and located in time, it will eventually lead to major faults such as insulation breakdown and equipment explosion, causing large-scale power outages and huge economic losses. Therefore, accurate monitoring and location of partial discharge in power equipment is a core requirement for power industry operation and maintenance.

[0003] Currently, the mainstream partial discharge monitoring technologies in the power industry mainly revolve around two single-mode technologies: ultrasonic monitoring and ultra-high frequency monitoring. Some solutions attempt to combine temperature sensing, but due to limitations in technical architecture and algorithm design, they still suffer from the following key technical defects, making it difficult to meet actual operation and maintenance needs:

[0004] Existing monitoring schemes generally suffer from the problem of "single-modal dependence":

[0005] For example, pure ultrasonic monitoring solutions (such as online monitoring devices based on a single ultrasonic probe) can capture mechanical vibration signals generated by partial discharge, but they are highly susceptible to external environmental interference, such as noise from the switch room fan, vibration from the movement of maintenance personnel, and ultrasonic interference caused by electromagnetic radiation from motors. Furthermore, they cannot distinguish between "internal discharge of equipment" and "external environmental interference," resulting in a high false alarm rate.

[0006] For example, pure UHF monitoring solutions (such as built-in UHF sensor systems) can detect UHF electromagnetic waves (300MHz-3GHz) generated by discharge and have strong anti-electromagnetic interference capabilities, but they are not sensitive enough to "non-corona weak discharges" (such as floating potential discharges), and the signal is easily shielded by the metal can, and cannot cover the external area of ​​the equipment (such as the outgoing wires and the top of the sleeve).

[0007] A few attempts to combine ultrasound and UHF are merely "parallel data acquisition" without establishing a spatiotemporal correlation logic between the two. For example, they fail to determine whether the two signals are generated by the same discharge event, and still have the problem of "misjudging external ultrasound interference and distant UHF interference as joint discharge," thus failing to reduce the false alarm rate. Summary of the Invention

[0008] This invention aims to at least partially address one of the technical problems in related technologies. Therefore, the objective of this invention is to propose a partial discharge monitoring and location system for power equipment based on multimodal sensing, in order to ensure the safe and stable operation of power equipment.

[0009] To achieve the above objectives, a first aspect of the present invention provides a power equipment partial discharge monitoring and location system based on multimodal sensing, comprising:

[0010] A multimodal sensing unit, configured to be deployed around the power equipment to be monitored, is used to simultaneously acquire multiple physical signals generated by partial discharge. The multimodal sensing unit includes at least:

[0011] Ultra-high frequency sensor array, used to receive ultra-high frequency electromagnetic wave signals generated by partial discharge;

[0012] An ultrasonic sensor array is used to receive ultrasonic signals generated by partial discharge.

[0013] The data acquisition and synchronization unit is connected to the multimodal sensing unit and is used to perform high-precision synchronous sampling and timestamp marking on the ultra-high frequency electromagnetic wave signal and the ultrasonic signal to generate synchronous multimodal data.

[0014] The central processing unit, connected to the data acquisition and synchronization unit, is configured as follows:

[0015] Receive the synchronous multimodal data;

[0016] A multimodal data fusion and discrimination algorithm is executed to identify real joint discharge events based on the spatiotemporal correlation between the ultra-high frequency electromagnetic wave signal and the ultrasonic signal;

[0017] Once the combined discharge event is identified, a joint localization algorithm is executed to calculate the discharge location coordinates of the combined discharge event in three-dimensional space.

[0018] To achieve the above objectives, a second aspect of the present invention provides a method for monitoring and locating partial discharge in power equipment based on multimodal sensing, comprising the following steps:

[0019] First, multimodal sensing units, including ultra-high frequency sensing arrays and ultrasonic sensing arrays, are deployed around the power equipment to be monitored to simultaneously acquire ultra-high frequency electromagnetic wave signals and ultrasonic signals generated by partial discharge.

[0020] The data acquisition and synchronization unit uses BeiDou / GPS second pulse signals to achieve nanosecond-level time synchronization of each sensor channel, and performs high-precision synchronous sampling and timestamp marking on the signals to generate synchronous multimodal data;

[0021] After receiving the data, the central processing unit executes a multimodal data fusion and identification algorithm. Using the UHF signal as the trigger reference, it searches for the existence of a valid ultrasonic signal within a preset time window and identifies the real joint discharge event through spatiotemporal correlation.

[0022] After successful identification, the joint localization algorithm is executed. First, the initial discharge area is calculated based on the time difference of arrival of the UHF signal. Then, the sound velocity is corrected by combining the temperature data provided by the thermal imaging unit. Finally, the time difference of arrival of the ultrasonic signal is used to perform a precise search in the initial area, and the three-dimensional coordinates of the discharge location are calculated.

[0023] Simultaneously, discharge feature vectors are extracted from multimodal signals and input into a pre-trained classification model to identify the discharge type;

[0024] The final result is displayed on the device's 3D model by a visualization unit, indicating the location, display type, and severity. When the threshold is exceeded, a graded alarm is triggered.

[0025] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory. When the computer program is executed by the processor, it implements the above-described method for monitoring and locating partial discharge of power equipment based on multimodal sensing.

[0026] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0027] This invention discloses a power equipment partial discharge monitoring and location system based on multimodal sensing. By constructing an integrated architecture of multimodal sensing collaboration and high-precision time synchronization, it effectively solves the problems of insufficient multimodal collaboration, low time synchronization accuracy, and lack of closed-loop optimization mechanism in existing partial discharge monitoring. Among them, the multimodal sensor array combined with the spatiotemporal correlation fusion discrimination algorithm can accurately distinguish between the actual internal discharge of the equipment and external environmental interference, avoiding false alarms and missed alarms.

[0028] The nanosecond-level synchronization achieved by UWB bilateral bidirectional ranging ensures the effective correlation of signals of different modes, laying the foundation for multi-source fusion. The adaptive dynamic compensation mechanism combines equipment structure and real-time environmental parameters to correct signal propagation characteristics, improves the accuracy of discharge positioning, and ultimately realizes accurate monitoring, comprehensive positioning, intelligent early warning and efficient operation and maintenance of partial discharge of power equipment, ensuring the safe and stable operation of the power grid. Attached Figure Description

[0029] The disclosure of this invention is illustrated with reference to the accompanying drawings. It should be understood that the drawings are for illustrative purposes only and are not intended to limit the scope of protection of this invention. In the drawings, the same reference numerals are used to refer to the same parts. Wherein:

[0030] Figure 1 This is a schematic diagram of the framework structure of a power equipment partial discharge monitoring and positioning system based on multimodal sensing provided by the present invention.

[0031] Figure 2 These are schematic diagrams (a) and (b) showing the alignment of UHF and AE signals under nanosecond-level synchronization in a power equipment partial discharge monitoring and positioning system based on multimodal sensing provided by the present invention.

[0032] Figure 3 This invention provides a power equipment partial discharge monitoring and location system based on multimodal sensing that exceeds... and Schematic diagrams (c) and (d) for determining the signal energy threshold;

[0033] Figure 4 This invention provides a schematic diagram (e) of the UHF coarse positioning area superimposed with AE for precise positioning in a power equipment partial discharge monitoring and positioning system based on multimodal sensing, and a schematic diagram (f) of the positioning error distribution.

[0034] Figure 5 This is a curve showing the change in sound velocity versus temperature in a power equipment partial discharge monitoring and location system based on multimodal sensing provided by the present invention.

[0035] Figure 6 This is a schematic diagram illustrating the change in positioning error with temperature in a power equipment partial discharge monitoring and positioning system based on multimodal sensing provided by the present invention.

[0036] Figure 7 This is a schematic diagram of a three-dimensional model for calibrating the discharge location in a power equipment partial discharge monitoring and positioning system based on multimodal sensing provided by the present invention.

[0037] Figure 8 This is a schematic diagram of the color mapping of early warning levels in a power equipment partial discharge monitoring and positioning system based on multimodal sensing provided by the present invention.

[0038] Figure 9 This is a flowchart illustrating a method for monitoring and locating partial discharge in power equipment based on multimodal sensing, provided by the present invention.

[0039] Figure 10 This is a schematic diagram of the structure of an electronic device provided by the present invention. Detailed Implementation

[0040] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0041] The following description, with reference to the accompanying drawings, illustrates an embodiment of the present invention: a power equipment partial discharge monitoring and location system based on multimodal sensing. Example 1

[0042] Figure 1 This is a schematic diagram of the framework structure of a power equipment partial discharge monitoring and positioning system based on multimodal sensing according to an embodiment of the present invention.

[0043] like Figure 1 As shown, in order to achieve accurate monitoring and location of partial discharge in large power equipment, the deployment and composition of the three core units of this system are as follows:

[0044] 1. A multimodal sensing unit, configured to be deployed around the power equipment to be monitored, for synchronously acquiring multiple physical signals generated by partial discharge. The multimodal sensing unit includes at least:

[0045] Ultra-high frequency sensor array, used to receive ultra-high frequency electromagnetic wave signals generated by partial discharge;

[0046] An ultrasonic sensor array is used to receive ultrasonic signals generated by partial discharge.

[0047] The ultra-high frequency (UHF) sensor array employs at least four built-in or external UHF sensors. For oil-immersed transformers, the sensors can be installed via pre-installed valves (such as drain valves) on the transformer tank wall, enabling their antennas to effectively receive UHF electromagnetic wave signals generated by partial discharge within the transformer. The frequency range is typically 300MHz-3GHz. These four sensors are spatially non-coplanar and non-collinear, forming a three-dimensional sensing array, which forms the geometric basis for achieving three-dimensional positioning.

[0048] Ultrasonic (AE) sensor array: This also employs at least four piezoelectric ultrasonic sensors, tightly attached to the outer surface of the transformer tank wall via magnetic adsorption or bonding. The sensor arrangement should fully consider the tank structure, evenly covering critical areas such as the vicinity of windings, bushings, and tap changers, while minimizing obstructions. These sensors are used to receive the signal of stress waves generated by partial discharge propagating through the tank structure and reaching the outer wall. These stress waves are ultrasonic signals, typically with a frequency range of 20kHz-300kHz.

[0049] 2. Data Acquisition and Synchronization Unit: Connected to the multimodal sensing unit, this unit performs high-precision synchronous sampling and timestamping of UHF electromagnetic wave signals and ultrasonic signals, generating synchronous multimodal data. This unit acts as a bridge between front-end sensing and back-end processing, consisting of multiple parallel acquisition channels and a unified synchronous clock source.

[0050] The importance of high-precision synchronization here is as follows: to ensure the accuracy of the subsequent TDOA (Time Difference of Arrival) positioning algorithm, time synchronization is the technical cornerstone of this system. In this embodiment, a BeiDou / GPS timing module is used, and its output pulse-per-second (PPS) signal is transmitted to each acquisition channel. The pulse leading edge accuracy of this PPS signal can reach the nanosecond level. Based on this, each channel can achieve strict time alignment and high-precision timestamp marking.

[0051] Synchronous Sampling: Under the triggering and calibration of the PPS signal, the analog-to-digital converters (ADCs) of all channels sample the UHF and AE signals in a fully synchronous manner. The sampling rate of the UHF signal is no less than 2GSa / s to ensure accurate reproduction of the pulse waveform; the sampling rate of the AE signal is no less than 5MSa / s.

[0052] like Figure 2 A schematic diagram illustrating the alignment of UHF and AE signals at nanosecond-level synchronization is shown in Figure (a). The solid blue line in Figure (a) represents the actual acquired UHF signal containing background noise. The dashed red line delineates the shape of the ideal discharge pulse submerged in noise, characterized by a steep rise time on the nanosecond level (~10 ns) and an extremely narrow pulse width. The dashed black line precisely marks the arrival time of the UHF pulse. =2.0μs. In Figure (b), the solid green line represents the acquired AE signal, which also contains significant noise, simulating mechanical noise sources such as oil pump vibration and fan rotation. The dashed magenta line represents the ideal AE signal, which is a decaying oscillation wave with a frequency of approximately 150kHz. The dashed black line again marks the UHF trigger time. The red dashed line indicates the arrival time of the AE signal. ≈62.0μs. The yellow highlighted area in the figure represents the preset time window. The window is... Starting from a certain point, its width (set to ~70μs in the figure) is (Maximum device size / velocity of sound) ensures that the AE signal from the furthest sensor can be captured.

[0053] 3. Central Processing Unit (CPU): Connected to the data acquisition and synchronization unit, the CPU typically consists of a high-performance embedded industrial computer or server. It has a built-in high-speed data bus and a powerful processor (CPU / GPU / FPGA) capable of real-time analysis and computation of massive amounts of synchronous multimodal data transmitted from the data acquisition unit.

[0054] The central processing unit is configured as follows:

[0055] Receive synchronous multimodal data;

[0056] A multimodal data fusion and discrimination algorithm is executed to identify real combined discharge events based on the spatiotemporal correlation of ultra-high frequency electromagnetic wave signals and ultrasonic signals;

[0057] Once a combined discharge event is identified, a combined localization algorithm is executed to calculate the discharge location coordinates of the combined discharge event in three-dimensional space.

[0058] This system's workflow aims to address two major technical pain points commonly found in traditional single-mode monitoring methods: noise interference and low positioning accuracy. Its core principle leverages the physical correlation between electromagnetic and acoustic signals generated simultaneously by the phenomenon of partial discharge. Through a two-step strategy of identification followed by positioning, it accurately locates the actual discharge signal from complex background noise. Specific steps include:

[0059] The first step is to use multimodal data fusion and identification based on spatiotemporal correlation to explore whether the signal just now was a real partial discharge.

[0060] The physical basis of this step is that when partial discharge occurs, the generated UHF electromagnetic waves travel at near the speed of light. ) propagation, while the propagation speed of ultrasound in insulating oil ( The signal transmission is much slower. Therefore, for the same discharge event, a UHF sensor will receive the signal almost instantaneously, while an AE sensor will have a significant delay proportional to the propagation distance. Random electromagnetic interference (such as cell phone signals) or mechanical vibration noise (such as accessory vibration) does not possess this strict "electricity first, sound later" characteristic with a reasonable time delay.

[0061] The algorithm implementation process for this step is as follows:

[0062] UHF signal initial triggering: The central processing unit monitors the data stream of all UHF channels in real time. When the signal amplitude of any UHF channel exceeds the preset noise baseline threshold, the system records the arrival time of the pulse. This is then marked as a candidate discharge event. The high sensitivity of UHF signals ensures that even weak, real discharges are not missed.

[0063] Establish a dynamic time window: Starting from this point, the system will open a preset time window. The width of the window It is based on the maximum geometric dimensions of the device under test. and speed of sound The formula, derived from scientific calculations, can be expressed as: It represents the longest time required for an ultrasonic wave to travel from any point inside the device to the farthest sensor.

[0064] AE signal association confirmation: The system then checks within this time window whether there is a valid ultrasonic signal in all AE channels, that is, a signal whose energy exceeds its own noise threshold.

[0065] Then, the incident is investigated and judged.

[0066] Real-life event: If in If the system successfully detects valid signals from multiple (e.g., at least three) AE sensors, the central processing unit determines that this "candidate discharge event" is a "real joint discharge event," which means that the electromagnetic and acoustic phenomena have been cross-validated.

[0067] Noise removal: If in If no valid AE signal is detected, the initial UHF pulse is determined to be external electromagnetic interference and removed from the dataset; conversely, if there is only an AE signal without UHF triggering, it is determined to be mechanical vibration noise.

[0068] The second step is joint positioning based on multimodal data collaboration, which aims to solve the positioning accuracy problem.

[0069] The algorithm implementation process for this step is as follows:

[0070] UHF coarse localization: The system first uses the arrival timestamps of a set of UHF signals associated with the event. By calculating the time difference of arrival (TDOA) between different sensors and solving a set of hyperbolic equations, an initial estimated area of ​​the discharge location can be obtained. Because UHF signals have complex propagation paths inside the equipment, including reflection and refraction, their positioning results may contain some errors, but are sufficient to define a reliable search range.

[0071] Precise AE localization: Next, the system uses the arrival timestamps of a set of AE signals associated with the event. The TDOA algorithm is also used, but the key difference is that the search algorithm for determining the sound source location is constrained to the initial estimation region obtained in the first step. Internally. This electromagnetically guided acoustic strategy has two major advantages:

[0072] First, it eliminates multiple solutions: When using only AE signals for global positioning, multiple spurious solutions often arise due to sensor layout and time measurement errors. By limiting the search range, these spurious solutions can be effectively eliminated, and the unique true solution can be found.

[0073] Secondly, it improves computational efficiency: it eliminates the need to search within the entire three-dimensional space of the device, greatly accelerating the convergence speed of the positioning algorithm.

[0074] Ultimately, the result of AE's precise positioning, namely the precise coordinates (x, y, z) of the combined discharge event in the device's three-dimensional model, is calculated and output.

[0075] like Figure 4 The diagram (e) shows the UHF coarse positioning area overlaid with the AE precise positioning result, and the positioning error distribution diagram (f). In the left diagram (e), the blue circular area represents the coarse positioning result based on the UHF signal, with a radius of approximately 1.2m. This reflects that the UHF method can pinpoint the approximate area of ​​partial discharge, but its accuracy is limited. The red scatter dots represent the precise positioning result calculated by the AE algorithm, distributed near the actual discharge source, indicating that the AE positioning achieves secondary optimization within the UHF region. The yellow pentagram marks the actual discharge source coordinates (2.5m, 3.5m), clearly showing that the AE positioning result is significantly closer to the actual location.

[0076] Figure (f) on the right shows a histogram of the error distribution between the AE positioning results and the actual location. Most positioning errors are concentrated in the range of 0.1–0.3m, with an average error of approximately 0.18m and a maximum error of no more than 0.5m. This figure intuitively demonstrates that the joint algorithm of UHF coarse positioning + AE precise positioning can balance range locking and accuracy optimization.

[0077] Compared to existing single-mode monitoring technologies, this system utilizes the inherent correlation between different physical signals to establish a powerful logical filter. Any single-source noise that cannot simultaneously cause electromagnetic and acoustic phenomena (whether external electromagnetic interference or internal mechanical vibration) is effectively eliminated, thus fundamentally solving the problems of false alarms and missed alarms. Every discharge event output by the system has undergone cross-validation, resulting in highly reliable results. Furthermore, the joint positioning algorithm creatively adopts a two-step strategy of UHF coarse positioning + AE fine positioning. It is not a simple weighted average of the two positioning results, but rather utilizes the global arrival characteristics of UHF signals to provide prior information for AE positioning, and then uses the advantage of the relatively clear propagation path of AE signals for accurate solution. This collaborative mechanism overcomes the inherent defects of single-mode positioning, improving positioning accuracy from the meter level to the centimeter level, and achieving precise source tracing of fault points.

[0078] In addition to answering questions about the presence and location of discharge, this system's synchronously acquired multimodal data provides rich information for subsequent fault diagnosis. Because it simultaneously captures the electromagnetic and acoustic waveforms of the discharge, future analysis of the intrinsic relationships between these waveform characteristics—for example, the steepness of the UHF pulse and the ratio of AE signal energy—can lead to a more accurate determination of the type (corona, surface, air gap) and severity of partial discharges. This is something that is difficult to achieve with any single information source.

[0079] For example, the above-mentioned multimodal sensing unit further includes:

[0080] One or more long-wave infrared thermal imaging sensor units were added. These units were installed in locations that overlook key areas of the power equipment being monitored (such as large oil-immersed transformers), ensuring that their field of view completely covers the main heat dissipation surfaces of the transformer enclosure.

[0081] The thermal imaging sensor periodically (e.g., once per minute) captures infrared radiation from the device surface and converts it into a two-dimensional temperature distribution map. This image, presented as a digital matrix, is transmitted to the central processing unit in real time. The central processing unit then registers this real-time temperature image with a pre-established three-dimensional digital model of the power equipment, thereby obtaining real-time temperature data for any point on the equipment's surface.

[0082] In addition, the central processing unit is further configured to compensate for the propagation speed of the ultrasonic signal using temperature distribution data. This is because the accuracy of the ultrasonic positioning algorithm fundamentally depends on the accuracy of the formula distance = speed × time.

[0083] In traditional acoustic positioning techniques, the propagation speed of ultrasound in insulating media (such as transformer oil) is typically assumed to be a constant. However, this is an idealized assumption that deviates significantly from reality. Physics shows that the speed of sound in a liquid medium is closely related to the temperature and density of the medium. For large power transformers, the internal oil temperature distribution is extremely uneven during operation, with the top and bottom oil temperatures differing by tens of degrees Celsius. Ignoring this temperature gradient and using a fixed average sound speed for calculation inevitably leads to huge positioning errors, ranging from tens of centimeters or even larger, rendering the positioning results meaningless in practical terms. This has long been a major technical bottleneck hindering high-precision acoustic positioning in this field.

[0084] How to use temperature data for compensation? This invention proposes an innovative compensation process for electromagnetic pre-positioning guidance and thermal imaging data correction, the working principle of which is as follows:

[0085] Step 1: Initial Location and Delineation of the Area. After the system confirms a real joint discharge event through spatiotemporal correlation, it first uses the time difference of arrival (TDOA) of the UHF sensor array to calculate an initial estimated area of ​​the discharge. ;

[0086] Step 2: Extract local equivalent temperature. The central processing unit extracts the local equivalent temperature based on the initial estimated region. Within the device's 3D model, the real-time surface temperature value corresponding to that area is retrieved and extracted. Using a heat conduction model, the local equivalent temperature within that area can be estimated. ;

[0087] Step 3: Calculate the corrected speed of sound. The central processing unit calls a preset physical model or looks up a table, based on the local equivalent temperature. Calculate the corrected local sound velocity in this region. This model is typically a temperature-dependent function, and the formula can be expressed as a linear approximation:

[0088]

[0089] in, , and All of these are known constants related to the properties of the insulating medium;

[0090] Step 4: Perform high-precision acoustic positioning. Finally, the system will use this precisely compensated local sound velocity... Substituting these values ​​into the ultrasonic TDOA positioning algorithm, the final discharge location coordinates are calculated.

[0091] The core contribution of the above steps lies in solving the critical problem of sound speed uncertainty at the physical level. By introducing thermal imaging as a new information dimension, the positioning model is transformed from a static, idealized model into a dynamic, accurate model capable of adapting to the device's operating status in real time. This allows the positioning error to be stably improved from the traditional decimeter level to the centimeter level, achieving a qualitative leap in positioning accuracy.

[0092] like Figure 5 The sound speed vs. temperature curve is displayed, showing the fit between the theoretical formula and the measured data. Figure 5The solid blue line represents the theoretical formula for the speed of sound as a function of temperature (approximately 331.4 + 0.6 Tm / s), the red dots represent measured sound speed data, and the dashed green line represents the linear fitting result based on the measured data. The curves show that the measured data and the theoretical formula generally follow the same trend, but a slight deviation occurs above 40°C, with an average deviation of approximately ±2 m / s. This result indicates that the sound speed-temperature fitting method corrected based on measured data can be optimized on the basis of the theoretical model, making the sound speed estimate closer to actual operating conditions.

[0093] In some embodiments of the present invention, spatiotemporal correlation is defined as follows:

[0094] Within the preset time window Inside, both the ultra-high frequency sensor array and the ultrasonic sensor array detected energy levels exceeding their respective preset thresholds. and The signal;

[0095] Among them, the preset time window The width of this window is not an arbitrary value, but a scientifically determined value based on physical principles. This ensures that the window size is large enough to cover the longest possible time it takes for an ultrasonic wave generated from any point within the device to travel to any sensor.

[0096] Energy threshold ( and These two thresholds are set above the background noise level of their respective channels to initially determine the validity of the signal.

[0097] High-voltage equipment operating environments are extremely complex, involving electromagnetic and mechanical noise sources such as radio broadcasts, mobile phone communications, and switch operations, which generate significant pulse interference that can be easily misinterpreted as discharge signals by sensitive UHF sensors. Mechanical noise sources, such as oil pump vibrations, fan rotations, and loose accessories, produce stress waves that can also be misinterpreted by AE sensors. Relying solely on a single sensor signal to diagnose partial discharge would result in a very high false alarm rate, rendering the monitoring system ineffective. Therefore, accurately extracting the true discharge signal from a massive amount of noise signals is the primary challenge that all monitoring systems must address.

[0098] By defining the spatiotemporal correlation, this system can perform event identification. The identification algorithm in this embodiment cleverly utilizes the physical principle that partial discharge inevitably generates both electromagnetic waves (near-light speed) and sound waves (slow speed), designing a logical criterion of "electromagnetic triggering, acoustic confirmation":

[0099] Step 1 (Trigger): The algorithm uses the UHF signal as a "sentinel." When any UHF sensor detects an energy exceeding... When a pulse signal arrives, its arrival time is recorded. Based on this, immediately initiate a time period of... Confirmation window;

[0100] Step Two (Confirmation): The algorithm is displayed in this confirmation window. Inside, check the data from all ultrasonic sensors;

[0101] Step 3 (Decision): If, within this window, at least one (or a predetermined number) ultrasonic sensors detect an energy exceeding [a certain threshold], [the following is a separate, unrelated statement:] If the system detects two signals, it determines that they originate from the same physical event and have a spatiotemporal correlation, thus confirming them as a real joint discharge event.

[0102] If no valid ultrasonic signal is detected when the confirmation window is closed, the system determines that the initial UHF signal is random electromagnetic interference and discards it.

[0103] In summary, this definition establishes a very strong logical constraint: a random external electromagnetic interference cannot spontaneously generate a matching ultrasonic signal within the device within a suitable time delay; conversely, a random mechanical vibration cannot spontaneously generate a precise and leading ultra-high frequency electromagnetic pulse. Therefore, this spatiotemporal correlation criterion can fundamentally filter out the vast majority of irrelevant noise, reducing the false alarm rate of the system to an extremely low level and ensuring the high reliability of the monitoring results. Furthermore, by eliminating a large amount of invalid noise data at the very beginning of data processing, it greatly reduces the computational burden on subsequent positioning and analysis algorithms, ensuring the real-time response capability of the entire system.

[0104] like Figure 3 Showing more and A schematic diagram illustrating the signal energy threshold determination. In Figures (c) and (d), the blue curve (UHF signal energy) and the green curve (AE signal energy) represent the partial discharge signal energies detected by the UHF and AE sensors, respectively. The UHF signal energy ranges from approximately 20dB to 100dB, with background noise maintained at around 30dB, while partial discharge events generate peak signals, reaching up to 65-89dB. The AE signal energy range is slightly lower, reflecting the attenuation characteristics during sound wave propagation. The red dashed line in the figure (… =50dB, =45dB) represents the set signal energy threshold. These thresholds effectively distinguish normal background noise from actual discharge events, avoiding false alarms common in traditional methods.

[0105] Not only that, Figure 3The solid red dots mark signal points exceeding the threshold, corresponding to partial discharge events occurring at different time points (e.g., 1.5 seconds, 3.2 seconds, etc.). Through dual-modal signal processing, the graph clearly shows the signal changes over time and whether they exceed the set threshold. Multimodal analysis of UHF and AE signals effectively improves the accuracy and reliability of event identification, significantly reducing the false alarm rate.

[0106] In some embodiments of the present invention, in order to achieve zero false alarm event identification from a strong noise background, the central processing unit of the system executes a multimodal data fusion identification algorithm, which utilizes cross-validation of information from multiple physical dimensions to ensure that only genuine partial discharge events can pass through, thereby fundamentally solving the false alarm problem. The specific steps include:

[0107] Step 1: Perform preliminary detection on the ultra-high frequency electromagnetic wave signal to identify candidate discharge pulses;

[0108] Technical details: 1. Signal preprocessing: The raw UHF signal received from the data acquisition and synchronization unit is not used directly. Instead, it is first filtered digitally (e.g., bandpass filtering) to remove noise outside the frequency band, and then denoised using techniques such as wavelet transform to improve the signal-to-noise ratio (SNR).

[0109] 2. Dynamic Threshold Setting: Instead of using a fixed amplitude threshold, the system implements an adaptive threshold algorithm. The central processing unit periodically analyzes the background noise level during periods without signal and dynamically sets a trigger threshold that is higher than the noise baseline (e.g., higher than 5 times the root mean square value).

[0110] 3. Candidate Event Generation: When the amplitude of the preprocessed UHF signal pulse continuously exceeds this dynamic threshold, the system identifies the pulse's starting point as a candidate discharge pulse. Simultaneously, the system accurately records the timestamp of the pulse's leading edge reaching the sensor. And extract its waveform data.

[0111] Step 2: Based on the arrival time of each candidate discharge pulse, within a preset time window... Within the range, it searches for the existence of valid ultrasonic signals acquired by the ultrasonic sensor array;

[0112] Technical details: 1. Scientific setting of time window: preset time window The setting of this window is crucial to the algorithm; it is based on explicit physical principles, not empirical guesses. This window represents the maximum theoretical time required for an ultrasonic wave generated from any point within the device to travel to the farthest sensor, ensuring that we do not miss any potentially relevant acoustic signals.

[0113] 2. Definition of a valid ultrasonic signal: Within this time window, the system will synchronously retrieve data from all ultrasonic sensors. A valid ultrasonic signal must meet two conditions: (a) its signal energy exceeds the background noise threshold of its channel; (b) its occurrence is not isolated, and it usually needs to be detected on at least N (e.g., N≥3) ultrasonic sensors to exclude local mechanical interference from individual sensors.

[0114] Step 3: If it exists, then the corresponding UHF signal and ultrasonic signal are identified as a combined discharge event;

[0115] Technical details: 1. Correlation confirmation: If the system successfully retrieves a valid ultrasonic signal that meets the conditions within the time window of step two, the authenticity of the candidate discharge pulse is cross-validated in the acoustic dimension.

[0116] 2. Data Encapsulation and Transmission: At this point, the central processing unit will officially confirm this event as a joint discharge event. It will package the UHF signal waveform associated with the event, all relevant AE signal waveforms, and their respective precise timestamps into a complete dataset and pass it to the subsequent joint localization algorithm module for processing.

[0117] 3. Noise Removal: If the time window If no valid ultrasonic signal is detected after the system is turned off, the system determines that the initial UHF pulse is random electromagnetic interference (such as space radiation) and discards it without proceeding to the next processing step.

[0118] Two challenges exist in the process of the central processing unit executing the multimodal data fusion and discrimination algorithm;

[0119] Challenge 1: Missed Reports Due to Signal Attenuation

[0120] In large equipment, a real but weak partial discharge may generate an ultrasonic signal that attenuates significantly during propagation, causing it to fall below the detection threshold by the time it reaches the sensor. In this case, although the UHF signal is captured, the algorithm may misclassify it as noise because a valid AE signal cannot be obtained for confirmation, resulting in a missed detection.

[0121] For example, this embodiment introduces a hierarchical confirmation mechanism to further optimize step three above, specifically including:

[0122] High-confidence event: Strong UHF signal, and clear signal detected by multiple AE sensors. Directly confirmed as a combined discharge event;

[0123] Medium Confidence Events (Suspected Events): The UHF signal is strong, but only one or two AE sensors detect a weak signal. Instead of immediately discarding the event, the system marks it as a suspected discharge event and performs time-series correlation analysis to observe whether such events will recur in the vicinity of the location within a short period. If they recur, its confidence level is increased.

[0124] This flexible processing strategy not only maintains the system's immunity to strong noise, but also greatly improves the detection sensitivity to early weak discharges, solving the problem of missed detection that may be caused by rigid criteria.

[0125] Challenge 2: Correlation and confusion issues under dense discharge or continuous background noise

[0126] In certain severe stages of a fault's development, multiple partial discharges may occur consecutively within a very short period (milliseconds); or, the equipment may have a continuous, pulse-like source of mechanical noise. This can lead to [the following issues] within the same [fault / fault]. Within the time window, multiple UHF pulses and multiple AE signals appeared, causing incorrect correlation.

[0127] For example, in this embodiment, a more advanced matching strategy may be employed, specifically including:

[0128] ① Preliminary positioning guidance: For the first UHF pulse within the window, the system can first use the time difference of its arrival at different UHF sensors to quickly estimate a rough discharge area;

[0129] ② Time of Arrival Prediction and Matching: Based on this rough area, the system can predict approximately when the ultrasonic signal should arrive at each AE sensor;

[0130] ③ Best match search: Then, the system will search for the set of AE signals that best matches this "predicted arrival time series" among the multiple AE signals actually detected in the window.

[0131] This refined correlation method based on prediction followed by matching can accurately find the unique multi-mode signal group corresponding to each discharge event, even in situations with dense signals. This greatly improves the algorithm's analytical capability and accuracy under complex working conditions, and each event points to high-value intelligence of real physical events, laying a solid and reliable data foundation for subsequent precise positioning and state assessment.

[0132] In some embodiments of the present invention, the joint localization algorithm includes:

[0133] First, based on the time difference of arrival of signals received by each sensor in the ultra-high frequency sensing array... A preliminary discharge location region was calculated. ;

[0134] Then, based on the time difference of arrival of the signals received by each sensor in the ultrasonic sensing array... The velocity of sound is corrected by combining the temperature data provided by the thermal imaging sensor unit. In the initial discharge location area A precise search was performed within the area to obtain the final discharge location coordinates. .

[0135] In the above, the initial discharge location region This is not the final result, but rather a preliminary search space calculated from the signals of an ultra-high frequency (UHF) sensor array: once a joint discharge event is confirmed, the approximate area of ​​the discharge is quickly calculated using the Time Difference of Arrival (TDOA) algorithm, taking advantage of the near-light speed propagation and minimal influence of the medium on UHF signals. This step is crucial, as it significantly reduces the computational complexity and ambiguity of subsequent acoustic localization. It's equivalent to using "radar" to first lock onto the approximate target range, avoiding aimless searching and effectively eliminating ghost solutions commonly found in acoustic localization.

[0136] Sound speed calculation based on temperature compensation This addresses the biggest source of error in traditional acoustic positioning, namely the speed of sound. It is considered a fixed constant. However, in actual large-scale power equipment, the temperature distribution of the internal insulating medium (such as transformer oil) is extremely uneven, resulting in significant differences in sound velocity at different locations. Ignoring the influence of temperature is the fundamental reason for the poor positioning accuracy of traditional methods.

[0137] Therefore, a thermal imaging sensing unit was introduced to provide a dynamic and accurate physical parameter for acoustic localization. It no longer relies on an inaccurate average sound velocity, but rather on... Real-time temperature of the area The physical formula is used to calculate a local sound velocity that perfectly matches the current operating conditions. .

[0138] Precise positioning: obtaining precise local sound velocity Afterwards, the system will be within the specified limits. Within the area, the arrival time difference data of the ultrasonic AE sensor array is used for final iterative search or analytical solution to obtain the centimeter-level discharge location coordinates.

[0139] For example, the corrected speed of sound mentioned above It can be determined using the following formula:

[0140]

[0141] in, This is the corrected speed of sound. The speed at which ultrasound travels in an insulating medium at 0 degrees Celsius. The temperature in Celsius is the discharge location region measured by the thermal imaging sensor unit.

[0142] As an example, first set the following device parameters:

[0143] Equipment to be monitored: 500kV gas-insulated switchgear (GIS), tank length 15 meters, bushing height 8 meters, rated current 3150A;

[0144] Ultra-high frequency sensor array: 4 sensors (2 built-in + 2 external), frequency band 300MHz-3GHz, sensitivity ≤1pC, sampling rate 2GSa / s;

[0145] Ultrasonic sensor array: 4 sensors, frequency 20-300kHz, sampling rate 5MSa / s, amplitude measurement range 0-80dBμV, strong magnetic adsorption installation;

[0146] Thermal imaging sensor unit: resolution 640×512, temperature measurement range -40℃-125℃, accuracy ±0.5℃, sampling period 1 minute;

[0147] Data acquisition and synchronization unit: BeiDou PPS synchronization, timestamp accuracy 1ns, data transmission rate 100Mbps.

[0148] Then, the localization calculation for partial discharge is performed:

[0149] Step 1: Multimodal Data Acquisition

[0150] UHF sensor: Detected 1 candidate discharge pulse, arrival time Amplitude 6pC;

[0151] Ultrasonic sensor: In Within the window (14:30:00.123456789s~14:30:00.132256789s), all four sensors detected valid signals, with sensor 1 arriving at [time value missing]. Amplitude 45 dBμV;

[0152] Thermal imaging unit: Acquired the temperature of the candidate discharge area (near the GIS bushing). .

[0153] Step 2: Data Synchronization and Verification

[0154] Synchronization verification: The time stamp deviation between the UHF and ultrasonic signals is ≤1ns, which meets the synchronization requirements;

[0155] Spatiotemporal correlation determination: UHF signal exceeds threshold (6pC > 5pC), AE signal exceeds threshold (45dBμV > 30dBμV), and AE is in The event was confirmed to be a combined discharge event.

[0156] Step 3: Joint Positioning Calculation

[0157] Preliminary localization (UHF): Calculate the TDOA of sensor A (built-in, coordinates (2,3,4)m) and B (external, coordinates (10,3,4)m). Combined with electromagnetic wave speed To obtain the preliminary area “X∈[8,10]m,Y∈[2,4]m,Z∈[3,5]m” (error ≤ 1 meter);

[0158] Sound speed correction: hour, ;

[0159] Precise positioning: Calculate the TDOA of sensor 1 (coordinates (2,3,4)m) and sensor 2 (coordinates (5,3,4)m). , combined The discharge coordinates are obtained by solving. The positioning error is ±0.08 meters, meaning the discharge is located 8.5 meters inside the GIS sleeve.

[0160] Figure 4 shows a schematic diagram (e) of the UHF coarse positioning area superimposed with the AE precise positioning result and a positioning error distribution diagram (f). In the left figure (e), the blue circular area represents the coarse positioning result based on the UHF signal, with a radius of approximately 1.2m, reflecting that the UHF method can pinpoint the approximate area of ​​partial discharge, but its accuracy is limited. The red scatter dots represent the precise positioning result calculated by the AE algorithm, distributed near the actual discharge source, indicating that the AE positioning has achieved secondary optimization within the UHF region. The yellow pentagram marks the coordinates of the actual discharge source (2.5m, 3.5m), and it can be intuitively seen that the AE positioning result is significantly closer to the actual location.

[0161] Figure (f) on the right shows a histogram of the error distribution between the AE positioning results and the actual location. Most positioning errors are concentrated in the range of 0.1–0.3m, with an average error of approximately 0.18m and a maximum error not exceeding 0.5m. This figure visually demonstrates that the combined algorithm of UHF coarse positioning and AE precise positioning can balance range locking and accuracy optimization.

[0162] And such Figure 6 This demonstrates how positioning error changes with temperature. Figure 6The red curve represents the positioning error without temperature compensation, which increases significantly with rising temperature, reaching approximately 3.5m at 70°C. The green curve represents the result after applying the proposed temperature compensation algorithm, with the error consistently remaining within the range of 0.2–0.4m. This comparison fully demonstrates the effectiveness of the temperature compensation mechanism, resolving the positioning deviation problem caused by temperature changes in traditional methods, thereby significantly improving the reliability and accuracy of power supply positioning.

[0163] In the above, since the precise search is based solely on the Time Difference of Arrival (TDOA) of the AE signal, the following limitations have not been considered:

[0164] 1. Multipath effect: Ultrasonic waves are reflected on the surface of the metal tank and insulating parts of the equipment, resulting in the superposition of "reflected waves" and "direct waves", which leads to a TDOA calculation deviation of ≥15μs;

[0165] 2. Blurred boundaries: Initial discharge area The boundary of (UHF coarse positioning) is a "fuzzy range". The search domain is not narrowed by combining the equipment structure (such as the diameter of the GIS tank and the diameter of the casing hole), which leads to slow convergence of the search algorithm (such as the least squares method) (time > 500ms) and easy to produce "pseudo solutions". For example, the location corresponding to the reflected wave is misjudged as the actual discharge point, and the actual positioning accuracy drops to more than 0.2 meters. The real-time performance cannot meet the needs of online monitoring.

[0166] Based on the above limitations, this embodiment employs the following steps to further improve accuracy and efficiency, specifically including:

[0167] Step 1: Direct Wave Separation: Based on the "amplitude attenuation characteristics" and "phase consistency" of the AE signal, the direct wave, i.e. the direct propagation signal generated by the actual discharge, is extracted from the superimposed signal. The reflected wave (interference signal) is eliminated to ensure that the TDOA calculation is based on the effective signal.

[0168] Step 2: Dynamic Search Domain Optimization: Based on the equipment's three-dimensional structure, such as the inner wall coordinates of the GIS tank and the diameter range of the casing, the initial search area is optimized. The boundary was modified to "equipment physical constraint boundary" (such as excluding areas inside the metal tank where discharge cannot occur), thus narrowing the search range;

[0169] Step 3: Improve the particle swarm algorithm: With "minimum direct wave TDOA error" as the objective function, quickly optimize within the dynamic search domain to obtain the true discharge location, balancing accuracy and efficiency.

[0170] In some embodiments of the present invention, the central processing unit is further configured as follows:

[0171] First, discharge feature vectors are extracted from the multimodal signals corresponding to the joint discharge events. The original multimodal signals are complex time-domain waveforms containing a large amount of redundant information and noise. Directly inputting the original signals into the model is not only computationally intensive but also makes it difficult for the model to converge quickly and learn effective patterns. Therefore, it is necessary to first extract the identity information that can most significantly distinguish different discharge types, that is, to construct a highly condensed and representative discharge feature vector. The feature combination selected in this embodiment is based on a deep understanding of the physical mechanism of discharge and takes into account the characteristics of two different physical phenomena, electromagnetic waves and sound waves.

[0172] For example, the discharge feature vector includes:

[0173] Characteristics of ultra-high frequency signals:

[0174] Pulse amplitude: This is the peak voltage of the UHF signal pulse waveform. It is directly related to the energy intensity released in a single discharge event. Typically, internal air gap discharges have relatively large pulse amplitudes due to their short paths and concentrated energy; while corona discharges usually have smaller amplitudes and are more dispersed.

[0175] Pulse rise time: Defined as the time required for a pulse to rise from 10% to 90% of its peak value. This parameter reflects the speed of discharge development. For example, discharges induced by sharp electrodes (such as corona discharges) develop extremely rapidly, with rise times typically on the order of nanoseconds (ns), and are very steep. Surface discharges, on the other hand, develop along the surface of an insulator, with a relatively tortuous path, and their rise times may be slightly longer.

[0176] Pulse width: This typically refers to the duration when the pulse amplitude drops to 50% (half-width at half maximum). It characterizes the duration of a single discharge pulse. Different types of discharges have different plasma process durations, which are directly reflected in the pulse width.

[0177] Characteristics of ultrasonic signals:

[0178] Energy: The energy of an ultrasonic signal is usually obtained by calculating the sum (or integral) of the squares of the signal over a certain time window. It reflects the sound pressure intensity generated by the instantaneous expansion of gas during the discharge process. Internal gas discharge, due to being enclosed by a metal casing, allows the sound wave energy to be more easily coupled to the sensor, exhibiting a higher energy value. Surface discharge is next, while corona discharge produces the weakest sound waves.

[0179] Duration: This refers to the time it takes for an ultrasonic signal to decay to the level of background noise. This parameter is related to the continuity of the discharge process. For example, surface discharge may be accompanied by multiple consecutive micro-discharges, forming a relatively long envelope of acoustic events.

[0180] Dominant frequency: The frequency point with the highest energy is found by performing Fourier transform (FFT) or other spectral analysis methods on the ultrasonic signal. Different types of discharges have different spectral characteristics of their acoustic signals. For example, the acoustic frequencies generated by internal air gap discharges are usually concentrated in the lower frequency band (such as tens to over one hundred kHz), while the dominant frequency distribution of surface discharges may be wider or higher.

[0181] Then, the discharge feature vector is input into a pre-trained discharge type classification model to identify the type of partial discharge, including corona discharge, surface discharge, or internal air gap discharge.

[0182] To achieve effective classification of discharge feature vectors, various mature machine learning or deep learning models can be employed. For example:

[0183] Support Vector Machine (SVM): It performs well in classification problems with small samples and high dimensionality, and has strong generalization ability; Random Forest: It consists of multiple decision trees, has good robustness, and is not prone to overfitting; Multilayer Perceptron (MLP) or other neural network models: They can learn complex nonlinear relationships between features and have higher classification accuracy when there is sufficient data.

[0184] During actual operation, the central processing unit performs the following operations:

[0185] 1. Receive a real-time multimodal signal of a combined discharge event from the sensor;

[0186] 2. Using the aforementioned method, the six-dimensional discharge feature vector of the event is calculated in real time. ;

[0187] 3. As input, it is fed into the pre-trained classification model that has been embedded in the system;

[0188] 4. The model outputs a classification result, such as "surface discharge", based on the decision boundary it has learned internally, and may include a confidence score.

[0189] In some embodiments of the present invention, the data acquisition and synchronization unit uses the pulse-per-second (PPS) signal of the Global Positioning System (GPS) or the BeiDou Navigation Satellite System (BDS) to achieve nanosecond-level time synchronization of each sensing channel.

[0190] In some embodiments of the present invention, the system further includes a visualization and alarm unit, which is the core of user interaction and operation and maintenance decision-making of the system. Its core function is to transform the abstract data (such as three-dimensional coordinates, discharge type codes, and warning values) output by the central processing unit into intuitive and operable operation and maintenance information, while ensuring timely response to faults through a hierarchical alarm mechanism.

[0191] The visualization and alarm unit is configured as follows:

[0192] 1. Mark the calculated discharge position coordinates on the three-dimensional digital model of the power equipment. .

[0193] The location of the discharge can be marked by a red dynamic flashing icon (such as a solid dot with a diameter of 5mm and a flashing frequency of 2 times / second). The icon color changes with the warning level (red for Level I emergency, orange for Level II important, yellow for Level III general, and blue for Level IV attention) to intuitively distinguish the urgency of the fault.

[0194] At the same time, key information should be marked next to the icon: discharge coordinates (accurate to 0.01m, such as X:8.50m, Y:3.20m, Z:4.10m), positioning error range (such as ±0.08m), and positioning time (accurate to the second, such as 2024-08-25 14:30:00).

[0195] Supports interactive query: Clicking the icon will bring up "Location Basis Details" (such as the sensor number involved in the location, the signal arrival time difference of each sensor, and the corrected sound velocity value), which makes it easy for maintenance personnel to trace the reliability of the location process.

[0196] like Figure 7 This diagram illustrates the calibration of discharge locations in a 3D model. The gray-white semi-transparent cube represents the 3D geometric model of the power equipment, and the red pentagram marks the local discharge locations (coordinate example: X=0.6, Y=0.4, Z=0.7, normalized relative to equipment dimensions). The diagram visually demonstrates that the positioning results accurately fall near the actual discharge location in 3D space. This method controls the positioning error in the 3D equipment model to within 0.2m, effectively improving the reliability of discharge calibration.

[0197] 2. Displays the assessment results of discharge type and severity.

[0198] (1) Discharge type display content:

[0199] Key findings: Discharge type, such as "internal air gap discharge of bushing", "corona discharge of outgoing line", and "discharge on insulation surface";

[0200] Confidence: The output confidence of the discharge type classification model. The confidence is calculated based on the matching degree between the discharge feature vector and the model training samples, such as the cosine similarity of 6-dimensional features such as UHF pulse amplitude and ultrasonic main frequency.

[0201] (2) Contents of the discharge severity assessment results:

[0202] Based on the aforementioned warning value W = location weight +Signal strength weight +Trend Weight The calculation logic displays the following quantization information:

[0203] Warning levels: Level I (urgent, W ≥ 80 points), Level II (important, 60 ≤ W < 80 points), Level III (general, 40 ≤ W < 60 points), Level IV (concerned, W < 40 points);

[0204] Scores for each weight: such as "position weight" =40 points (the inside of the casing is the critical area), signal strength weighting =25 points (UHF amplitude 6pC, AE amplitude 45dBμV), trend weight =22 points (discharge frequency increased by 15% within 10 minutes), total warning value W=87 points (Level I)”;

[0205] Trend curves: Display the "discharge frequency-time" and "signal amplitude-time" curves over the past hour (e.g., one data point every 5 minutes), visually presenting the discharge development trend. For example, "the discharge frequency increased from 5 times / minute to 18 times / minute, showing an accelerating upward trend."

[0206] 3. When the monitored indicators exceed the preset alarm threshold, an alarm is triggered.

[0207] Among them, the monitoring indicator types are: three core indicators directly related to discharge hazards are selected, and the threshold settings are based on the common technical requirements of the "GB / T11022-2020 High Voltage AC Switchgear and Controlgear Standard" and equipment operation and maintenance experience, as shown in the table below:

[0208] Monitoring indicators Threshold setting logic Example Discharge frequency Based on historical normal data of the device (e.g., ≤5 times / minute during normal operation), the threshold is set to 3 times the maximum normal value. Threshold = 15 times / minute Warning value Based on the warning level classification, the threshold for Level I alarms is 80 points, and for Level II it is 60 points. Level I alarm threshold = 80 points Discharge duration in critical areas If discharge in critical areas (such as bushings and windings) continues for more than 30 minutes, an alarm should be triggered to prevent further damage to the insulation. Threshold = 30 minutes

[0209] Threshold calibration mechanism: Supports dynamic adjustment of thresholds based on equipment operating years and environmental conditions (such as humidity and temperature). For example, for GIS equipment that has been operating for more than 10 years, the insulation performance has deteriorated, and the discharge frequency threshold can be lowered to 12 times / minute to ensure higher alarm sensitivity for older equipment.

[0210] Alarm levels and methods: Level I alarm (emergency): On-site audible and visual alarm (buzzer sounds continuously next to the equipment, red warning light stays on) + remote multi-channel notification (SMS to maintenance personnel, pop-up window on the maintenance platform, telephone alarm from the dispatch center), with a notification interval of 5 minutes (until confirmed receipt);

[0211] Level II Alarm (Important): On-site yellow warning light flashes (1 time / second) + remote SMS + platform pop-up, notification interval 15 minutes;

[0212] Level III / IV alarms (general / concern): Alarm information is only displayed on the operation and maintenance platform, without on-site audio and visual effects, to avoid interfering with normal operation and maintenance.

[0213] Alarm confirmation and closure: After an alarm is triggered, maintenance personnel need to "confirm receipt" and "feedback processing results" in the system (such as "maintenance has been arranged" or "false alarm (environmental interference)"). The system automatically records the entire process log of "alarm triggering-confirmation-processing-closure" (retention time ≥ 1 year) to meet the power industry's maintenance traceability requirements.

[0214] like Figure 8 A color-map of warning levels is displayed. Given different alarm levels for different areas, a color map (green / yellow / red) is used to visually represent the risk status. Green areas represent normal operation, yellow areas represent minor anomalies or alert status, and red areas represent alarm status with a serious risk of discharge. This map demonstrates that by overlaying warning levels onto different areas, maintenance personnel can quickly determine the overall health status of the equipment and high-risk locations. For example, a red area appears near (X=7, Y=3) in the map, indicating that priority action is needed. Example 2

[0215] like Figure 9 As shown in the above system embodiment, the present invention also proposes a method for monitoring and locating partial discharge of power equipment based on multimodal sensing, comprising the following steps:

[0216] First, multimodal sensing units, including ultra-high frequency sensing arrays and ultrasonic sensing arrays, are deployed around the power equipment to be monitored to simultaneously acquire ultra-high frequency electromagnetic wave signals and ultrasonic signals generated by partial discharge.

[0217] The data acquisition and synchronization unit uses BeiDou / GPS second pulse signals to achieve nanosecond-level time synchronization of each sensor channel, and performs high-precision synchronous sampling and timestamp marking on the signals to generate synchronous multimodal data;

[0218] After receiving the data, the central processing unit executes a multimodal data fusion and identification algorithm. Using the UHF signal as the trigger reference, it searches for the existence of a valid ultrasonic signal within a preset time window and identifies the real joint discharge event through spatiotemporal correlation.

[0219] After successful identification, the joint localization algorithm is executed. First, the initial discharge area is calculated based on the time difference of arrival of the UHF signal. Then, the sound velocity is corrected by combining the temperature data provided by the thermal imaging unit. Finally, the time difference of arrival of the ultrasonic signal is used to perform a precise search in the initial area, and the three-dimensional coordinates of the discharge location are calculated.

[0220] Simultaneously, discharge feature vectors are extracted from multimodal signals and input into a pre-trained classification model to identify the discharge type;

[0221] The final result is displayed on the device's 3D model by a visualization unit, indicating the location, display type, and severity. When the threshold is exceeded, a graded alarm is triggered. Example 3

[0222] Corresponding to the above embodiments, the present invention also proposes an electronic device.

[0223] like Figure 10 The diagram shows a structural schematic of an electronic device according to the present invention. The electronic device 200 includes a processor 201 and a memory 203. The processor 201 and the memory 203 are connected, for example, via a bus 202. Optionally, the electronic device 200 may further include a transceiver 204. It should be noted that in practical applications, the transceiver 204 is not limited to one unit, and the structure of this electronic device 200 does not constitute a limitation on the embodiments of the present invention.

[0224] Processor 201 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA, or other programmable logic device, transistor logic device, hardware component, or any combination thereof. It may implement or execute the various exemplary logic blocks, modules, and circuits described in connection with this disclosure. Processor 201 may also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0225] Bus 202 may include a path for transmitting information between the aforementioned components. Bus 202 may be a PCI bus or an EISA bus, etc. Bus 202 may be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 10 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0226] The memory 203 stores a computer program corresponding to the multimodal sensing-based power equipment partial discharge monitoring and location method of the above embodiments of the present invention. This computer program is controlled and executed by the processor 201. The processor 201 executes the computer program stored in the memory 203 to implement the content shown in the aforementioned method embodiments.

[0227] Among them, electronic devices 200 include, but are not limited to: mobile terminals such as laptops and tablets, as well as fixed terminals such as desktop computers. Figure 10 The electronic device 200 shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0228] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, optimize, distribute, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, the computer-readable medium can even be paper or other suitable media on which the program can be printed, since the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0229] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0230] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0231] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0232] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A multi-modal sensor based partial discharge monitoring and location system for power equipment, the system comprising: Comprise: A multi-modal sensing unit configured to be arranged around the power equipment to be monitored for synchronously collecting multiple physical signals generated by partial discharge, the multi-modal sensing unit at least comprising: A UHF sensing array for receiving UHF electromagnetic wave signals generated by partial discharge; An ultrasonic sensing array for receiving ultrasonic signals generated by partial discharge; A thermal imaging sensing unit configured to periodically capture infrared radiation of the surface of the power equipment to be monitored and convert it into a two-dimensional temperature distribution map to obtain real-time temperature data of the surface of the equipment; The central processing unit is further configured to compensate the propagation speed of the ultrasonic signals using the temperature distribution data; A data acquisition and synchronization unit connected with the multi-modal sensing unit for high-precision synchronous sampling and time stamping of the UHF electromagnetic wave signals and the ultrasonic signals to generate synchronized multi-modal data; A central processing unit connected with the data acquisition and synchronization unit and configured to: Receive the synchronized multi-modal data; Execute a multi-modal data fusion discrimination algorithm to identify a real joint discharge event based on the spatio-temporal correlation of the UHF electromagnetic wave signals and the ultrasonic signals; When the joint discharge event is identified, execute a joint positioning algorithm to calculate the discharge location coordinates of the joint discharge event in three-dimensional space; the joint positioning algorithm comprises: Step one: based on the time difference of arrival of signals received by each sensor in the ultra-high frequency sensor array, a preliminary discharge location area is calculated Step two: acoustic velocity field correction based on heat conduction model is performed, specifically: according to the position of the preliminary discharge location area in the three-dimensional model of the device, the corresponding real-time surface temperature value of the area is queried and extracted from the two-dimensional temperature distribution map; based on the real surface temperature value, the preliminary discharge location area inside is estimated using a preset heat conduction model to estimate the local equivalent temperature inside the area; according to the local equivalent temperature, the corrected local acoustic velocity of the area is calculated Step three: based on the time difference of arrival of signals received by each sensor in the ultrasonic sensor array , combined with the corrected local acoustic velocity , the preliminary discharge location area is accurately searched to obtain the final discharge location coordinates .

2. The system of claim 1, wherein, The definition of the spatio-temporal correlation is: within a preset time window both the very high frequency sensor array and the ultrasonic sensor array detect signals that exceed respective preset energy thresholds and ​ 3. The system of claim 1, wherein, The specific steps of the central processing unit executing the multi-modal data fusion discrimination algorithm include: Step one: preliminary detection of the UHF electromagnetic wave signals to identify candidate discharge pulses; Step two: search whether there is valid ultrasonic signal collected by the ultrasonic sensor array within the preset time window based on the arrival time of each candidate discharge pulse Step two: search whether there is valid ultrasonic signal collected by the ultrasonic sensor array within the preset time window based on the arrival time of each candidate discharge pulse Step three: if there is, the corresponding UHF signals and ultrasonic signals of the group are confirmed as a joint discharge event.

4. The system of claim 1, wherein, the corrected sound velocity is determined by the equation: wherein, is the corrected sound velocity, is the propagation speed of the ultrasonic wave in the insulating medium at 0 degrees Celsius, is the Celsius temperature of the discharge position region measured by the thermal imaging sensor unit.

5. The system of claim 1, wherein, The central processing unit is further configured to: Extract a discharge feature vector from the multi-modal signals corresponding to the joint discharge event; Input the discharge feature vector into a pre-trained discharge type classification model to identify the type of partial discharge, the discharge type including corona discharge, surface discharge or internal air gap discharge.

6. The system of claim 5, wherein, The discharge feature vector includes: The pulse amplitude, pulse rise time and pulse width of the UHF signals, and the energy, duration and main frequency of the ultrasonic signals.

7. The system of claim 1, wherein, The data acquisition and synchronization unit uses the second pulse signal of the global positioning system or Beidou satellite navigation system to achieve nanosecond-level time synchronization of each sensing channel.

8. The system of claim 1, wherein, The system further comprises a visualization and alarm unit configured to: on a three-dimensional digital model of the electrical power device, indicating the computed discharge position coordinates ; Display the discharge type and discharge severity assessment results; Trigger an alarm when the monitoring index exceeds the preset alarm threshold.

Images