High-precision sensing device capable of being used for GIS accidental partial discharge positioning and type diagnosis

Abstract

The invention relates to a high-precision sensing device capable of being used for GIS accidental partial discharge positioning and type diagnosis. The high-precision sensing device comprises an ultrahigh frequency sensor and a data acquisition system. The ultrahigh frequency sensor is responsible for capturing weak accidental partial discharge signals generated in the GIS equipment; the captured original signal is then sent to a filtering and amplifying module and a synchronization module for hardware preprocessing, and is digitalized through a high-speed sampling assembly; the digitized mass data are transmitted to the high-performance GPU parallel computing unit, and the data are efficiently analyzed by utilizing the strong parallel processing capability of the high-performance GPU parallel computing unit. According to the invention, the equipment sampling rate is improved to a high sampling rate of 5GHz, so that the partial discharge pulse sampling interval of the GIS partial discharge monitoring equipment reaches 0.2 ns, and the sampling points and the data volume are greatly increased; gPU parallel computing is combined and utilized to greatly improve the pulse processing capacity, so that the system can effectively find and capture intermittent and accidental partial discharge signals, and the detection rate of the intermittent and accidental signals is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring technology, and in particular to a high-precision sensing device that can be used for the location and type diagnosis of sporadic partial discharges in GIS. Background Technology

[0002] Gas-insulated substations (GIS), as a crucial component of power systems, play a vital role in load switching, system dispatching, and system protection, ensuring the safe and economical operation of the power grid. Due to their small footprint, strong environmental adaptability, and high overall reliability, GIS is widely used in power systems of all voltage levels. However, the compact insulation design, high manufacturing requirements, and complex transportation and installation of GIS equipment can lead to potential insulation defects, posing a threat to long-term operational safety. To ensure the safe operation of GIS, live-line monitoring technology has enabled the effective acquisition of switchgear status data. Equipment condition monitoring, condition assessment, and fault diagnosis technologies have been widely applied, playing a significant role in ensuring the safe and reliable operation of the equipment.

[0003] However, although online monitoring and live detection technologies, mainly based on UHF partial discharge detection, have been widely applied to the operation and maintenance of GIS equipment, most GIS equipment uses electronegative SF6 gas as the gaseous insulating medium. The partial discharges caused by its defects are highly random and sporadic, and cannot form statistical characteristics in a short period of time. Existing detection technologies are difficult to effectively capture partial discharge events on a long time scale, making it difficult to grasp and analyze the existence and development of defects in an overall manner.

[0004] Existing GIS UHF partial discharge detection equipment achieves discharge signal sensing, analysis, and feedback based on the collaborative operation of multiple modules, such as... Figure 1 As shown, its system structure mainly consists of an ultra-high frequency sensor, a data acquisition unit, a data processing unit, a display unit, a control unit, and a charging unit. The functions of each module are as follows:

[0005] The ultra-high frequency sensor, as the core sensing unit, is integrated into the surface of the cavity of gas-insulated switchgear, either internally or externally, to capture electromagnetic waves in the 300MHz to 3GHz frequency band generated during partial discharge. This sensor employs a wideband antenna design, effectively distinguishing discharge signals from background noise. Its non-contact detection characteristics avoid invasive modifications to the internal structure of the equipment, while also possessing anti-electromagnetic interference capabilities, making it suitable for early discharge monitoring under complex operating conditions.

[0006] The data acquisition unit is responsible for converting the analog signals output by the sensors into digital signals. It integrates a high-speed analog-to-digital converter and utilizes a pre-filter circuit to suppress low-frequency interference such as power frequency harmonics. This unit is equipped with a buffer mechanism to temporarily store sudden discharge pulse data, providing continuous timing signal input for subsequent processing.

[0007] The data processing unit employs embedded algorithms to perform in-depth analysis of the acquired signals. First, environmental noise is eliminated through digital filtering and wavelet denoising techniques. Then, characteristic parameters such as the amplitude, phase, and repetition frequency of the discharge pulse are extracted, and typical discharge types are matched based on pattern recognition algorithms.

[0008] The display unit serves as the human-machine interface, presenting real-time discharge spectra, historical trend curves, and alarm information in a graphical manner. Its interface design is compatible with both two-dimensional spectrum and three-dimensional time-frequency analysis results, supports touch operation and multi-view switching, facilitating maintenance personnel to quickly locate abnormal discharge sites and assess equipment health status.

[0009] The control unit coordinates the system's operation via a programmable logic controller (PLC) or microprocessor. Its functions include setting the acquisition frequency, configuring algorithm parameters, managing data storage strategies, and adapting to external communication protocols. This unit can also trigger multi-level alarms based on preset thresholds and coordinate with the substation automation system to perform protective actions.

[0010] The charging unit features a wide-voltage input design, supporting AC / DC power compatibility. It incorporates a built-in lithium-ion battery pack and power management chip, ensuring continuous operation even during grid failures or extreme outdoor environments. Its intelligent charge / discharge control module monitors the battery status in real time and extends battery life through a low-power mode.

[0011] Intermittent partial discharge events occur infrequently and with low intensity. To detect and identify UHF intermittent partial discharge signals inside UHV GIS (Ultra-High Voltage GIS), practical applications typically employ multi-channel detection using both built-in and external spatial UHF sensors. Each signal pulse is manually analyzed, and the amplitude and time difference of the signals from the built-in and external spatial UHF sensors are compared to identify the partial discharge signals inside the UHV GIS. When there is significant interference, the identification workload is enormous and cannot meet the needs of rapid on-site diagnosis. In noisy environments, detection systems require sufficient sensitivity to capture these weak signals and an extremely high sampling rate to capture the raw pulse waveform for localization. Furthermore, the large amount of monitoring data necessitates efficient processing algorithms and sufficient storage space. Summary of the Invention

[0012] The technical problem to be solved by this invention is: In order to overcome the above-mentioned technical problems, this invention provides a high-precision sensing device that can be used for the location and type diagnosis of sporadic partial discharges in GIS. The goal is to develop a low-cost, high-precision sporadic partial discharge sensing device by introducing high-sampling-rate hardware and high-computing-power software through hardware and software collaboration, so as to achieve accurate capture and effective identification of sporadic pulse signals. This enables the device to overcome the limitations of traditional methods in processing sparse and intermittent signals, significantly improve the detection capability of sporadic partial discharges, and thus achieve high-precision detection.

[0013] The technical solution adopted by this invention to solve its technical problem is: a high-precision sensing device that can be used for the location and type diagnosis of sporadic partial discharges in GIS, comprising:

[0014] Ultra-high frequency sensor: It is a high-sensitivity wideband antenna with a planar butterfly antenna structure. The operating frequency band is 300MHz~1500MHz. It is non-contactly installed on the surface of the GIS equipment cavity and is used to capture electromagnetic signals generated by partial discharge of GIS equipment.

[0015] Data acquisition system: includes high-speed sampling components and GPU parallel computing unit, with a 5GSPS sampling rate and multi-channel synchronous acquisition capability. Based on the GPU parallel computing unit, it performs real-time analysis and processing of the acquired signals for the capture, spatial positioning and discharge type identification of partial discharge signals.

[0016] The high-speed sampling component is used to digitize the signal from the ultra-high frequency sensor and achieve an equivalent sampling rate of 20 GSPS through multi-channel time interleaving.

[0017] The GPU parallel computing unit is used to perform spatial positioning and discharge type identification on digital signals in real time, achieving a spatial positioning accuracy of ≤0.06m and a pulse real-time processing capability of ≥200,000 pulses / second.

[0018] The ultra-high frequency sensor adopts a planar butterfly antenna structure and is manufactured using printed circuit board technology. It has wide bandwidth characteristics, with an absolute working bandwidth of not less than 8:1, a standing wave ratio of less than 2, linear polarization, and directional orientation. The antenna has a maximum gain of 10dB±0.5dB and a small size, which can effectively distinguish the discharge signal from the background noise and achieve high-sensitivity detection of partial discharge signals of GIS equipment.

[0019] The data acquisition system includes a filtering and amplification module, a synchronization module, a power supply module, a control module, and a high-speed sampling component. Based on a GPU parallel computing unit, it performs real-time spatial positioning and discharge type identification on the acquisition signal output by the high-speed sampling component. The filtering and amplification module and the synchronization module are used to preprocess the signal output by the UHF sensor. The power supply module is used to provide power to each module of the data acquisition system. The control module is used to coordinate the operation of each module of the data acquisition system. The high-speed sampling component is used to capture and digitize the partial discharge pulse signal of the GIS equipment.

[0020] The filtering and amplification module includes a bandpass filter and a preamplifier. The bandpass filter is composed of a cascaded high-pass filter and a low-pass filter, with a bandwidth of 300MHz to 1500MHz. It is used to suppress low-frequency interference signals such as power frequency harmonics and corona discharge below 300MHz, thereby improving the anti-interference capability of the device. The preamplifier is used to amplify the weak signal output by the UHF sensor. The input impedance of the preamplifier is matched with the output impedance of the UHF sensor, and it has a stable gain and low noise characteristics. It has low input offset voltage, low input offset current, and small temperature drift, and has sufficient bandwidth and slew rate to ensure distortion-free amplification of partial discharge pulse signals. The preamplifier also has a high common-mode rejection ratio and a wide common-mode input range. The closed-loop gain can be adjusted to achieve gain switching in the range of 20 to 40dB, and it has high linearity and high amplification accuracy.

[0021] The synchronization module includes an external synchronization circuit and a power synchronization circuit. The external synchronization circuit obtains a synchronization signal from the high-voltage power grid, uses a voltage transformer to convert the high-voltage signal (AC 1V~400V, 30Hz~300Hz) into a low-voltage signal, and then converts it into a square wave synchronization signal for processing by the high-speed sampling component of the data acquisition system. The power synchronization circuit extracts a synchronization signal from the device's AC power supply, uses a voltage transformer to convert the AC 85V~240V, 40Hz~70Hz power supply signal into a square wave synchronization signal for processing by the high-speed sampling component. Through the reference synchronization signals provided by the external synchronization circuit and the power synchronization circuit, the data acquisition reference frequency of the partial discharge detection device is precisely matched with the power grid frequency to distinguish between real partial discharge signals and external noise, and to facilitate the classification and diagnosis of partial discharge types.

[0022] The power module adopts a wide voltage input design, capable of connecting to AC power supplies ranging from 85 to 240V or corresponding DC power supplies. The power module includes an isolated AC / DC converter unit that converts AC 220V mains power into the DC voltage required by various parts of the data acquisition system. The AC / DC converter unit has an isolation withstand voltage of at least 4200VAC and an energy conversion efficiency of no less than 85%, and incorporates output short-circuit protection, overcurrent protection, and overvoltage protection circuits. The power module also includes a low-dropout linear regulator that regulates and steps down the 5V DC voltage output from the AC / DC converter unit to 3V DC voltage to power the control module.

[0023] The control module includes a microcontroller for coordinating the operation of various modules in the data acquisition system. The microcontroller is based on a Cortex-M4 core with a maximum clock frequency of 150MHz. It features a DSP instruction set and memory protection unit, and has peripheral resources such as a USB OTG interface, multiple timers, an analog-to-digital converter interface, a comparator, and a multi-channel communication interface, and supports low-power mode. The control module receives and processes the synchronization signal output by the synchronization module, communicates with the high-speed sampling component via the USB interface, and controls the gain of the preamplifier to switch between 20 and 40dB via the gain control pin (GCTRLx).

[0024] The high-speed sampling component includes a multi-channel high-speed analog-to-digital converter array with eight 14-bit ADC channels, each with a sampling rate of 5 GSPS, supporting multi-channel synchronous sampling. The high-speed sampling component achieves an equivalent sampling rate of 20 GSPS through multi-channel time interleaving to capture partial discharge pulse signals. The analog input terminal of the high-speed sampling component is connected to the filtering and amplification module via a high-performance miniature RF connector and is equipped with a high-speed cache memory for buffering the acquired data.

[0025] The high-speed sampling component adopts a radio frequency system-on-a-chip (RFSoC) architecture, integrating an FPGA programmable logic unit and a general-purpose processor, and is equipped with on-chip DDR4 memory. The high-speed sampling component has a 10 / 100 / 1000M Ethernet interface and at least two 100Gbps high-speed fiber optic interfaces for real-time output or offloading of acquired partial discharge data to external storage and computing devices. Through this RFSoC integrated architecture, the integration of multi-channel pulse signal acquisition is improved, the complexity and latency of the radio frequency signal chain are reduced, and hardware support for parallel digital signal processing in specific applications is provided. The external storage and computing devices mentioned here correspond to GPU parallel computing units, which are part of the high-speed sampling component.

[0026] The GPU parallel computing unit utilizes the massively parallel computing capabilities of the GPU (Graphics Processing Unit) to process the acquired partial discharge data in real time. The parallel computing unit executes signal analysis algorithms, including digital filtering and wavelet denoising of the partial discharge signal, as well as extracting the characteristic parameters of the discharge pulse, and matching typical discharge types with the built-in database, thereby realizing real-time detection, localization, and type diagnosis of partial discharge.

[0027] The GPU parallel computing unit improves pulse data processing efficiency by optimizing the memory access computation ratio (RMtoC) and memory access thread bundle limit (MWP) model.

[0028] (a) Memory access computation ratio:

[0029] (1)

[0030] (2)

[0031] Where: MemCycles refers to the average number of clock cycles in a thread-wrap memory access cycle, and memory access cycle refers to the number of clock cycles spent on a global memory access instruction; GlobalMemLatency is the global memory access latency, which refers to the number of clock cycles spent from when a thread initiates a global memory access to when the data is returned to the register (which can be used for calculation). CacheMissRate is the cache miss rate, which indicates what percentage of all cache accesses miss, requiring access to a higher level of memory than the currently accessed memory, i.e., slower memory. CacheLatency is the cache access latency, which refers to the average number of clock cycles required to complete a cache access. CompCycles refers to the average number of clock cycles in a thread-wrap computation cycle, and computation cycle refers to the number of clock cycles spent on all computation instructions between two adjacent memory access instructions, obtained through PTX code analysis; RMtoC represents the memory access-to-computation ratio, reflecting the proportional relationship between memory access latency and computation latency. The calculation formula is shown in Equation (2). The larger RMtoC is, the more difficult it is to mask memory access latency, and the algorithm tends to be memory-restricted; otherwise, it tends to be computation-restricted.

[0032] (b) Memory access thread bundle limit (MWP):

[0033] (3)

[0034] (4)

[0035] Where Bandwidth, f, and #SM are GPU hardware parameters, and LoadBytesPerWarp refers to the number of bytes that each thread bundle accesses from global memory in MemCycles. This refers to the number of base bytes that each thread bundle accesses from global memory in MemCycles.

[0036] The beneficial effects of this invention are as follows: This invention provides a high-precision sensing device for the location and type diagnosis of sporadic partial discharges in GIS (Geographic Information System). An ultra-high frequency sensor is responsible for capturing weak sporadic partial discharge signals generated inside the GIS equipment. The captured raw signals are then sent to a filtering and amplification module and a synchronization module for hardware preprocessing, and digitized by a high-speed sampling component. The massive amount of digitized data is transmitted to a high-performance GPU parallel computing unit, utilizing its powerful parallel processing capabilities for efficient data analysis. This invention increases the sampling rate of the device to a high sampling rate of 5GHz, enabling the partial discharge pulse sampling interval of the GIS partial discharge monitoring equipment to reach 0. The 0.2ns speedup significantly increases the number of sampling points and data volume. Combined with the use of GPU parallel computing, the pulse processing capability is greatly improved, increasing the processing capacity from only 10-20 pulses per second for traditional CPUs to 200,000 pulses per second. This pulse processing speedup of up to 666.67 times enables the system to effectively detect and capture intermittent partial discharge signals, thereby significantly improving the detection rate of intermittent partial discharge signals. It can not only accurately capture each intermittent partial discharge pulse, but also enhance the computing power and response speed of the equipment. In partial discharge localization, it greatly improves the spatial positioning accuracy, increasing the positioning accuracy from about 0.5m to 0.06m. Attached Figure Description

[0037] Figure 1 This is a block diagram of a GIS UHF detection system.

[0038] Figure 2 This is a circuit diagram of the filtering and amplification module of the present invention.

[0039] Figure 3 This is a circuit diagram of the external synchronization circuit of the present invention.

[0040] Figure 4 This is a circuit diagram of the power synchronization circuit of the present invention.

[0041] Figure 5 This is a design block diagram of the power module of the present invention.

[0042] Figure 6 This is a circuit diagram of the control module of the present invention.

[0043] Figure 7 This is a diagram of a single-node, multi-GPU parallel computing framework.

[0044] Figure 8 This is a flowchart of the operation of the high-precision sensing device of the present invention, which can be used for the location and type diagnosis of sporadic partial discharges in GIS.

[0045] Figure 9 This is the original waveform.

[0046] Figure 10 It is the cumulative PRPD over a long period of time in occasional local releases. Detailed Implementation

[0047] The invention will now be described in further detail with reference to the accompanying drawings. It should be emphasized that the following description is merely exemplary and not intended to limit the scope or application of the invention.

[0048] The high-precision sensing device of the present invention, applicable to the location and type diagnosis of sporadic partial discharges in GIS (Gas Insulation System), aims to significantly improve the application efficiency of partial discharge detection technology and achieve accurate detection of sporadic pulses caused by insulation defects in GIS equipment. The core components of this device include:

[0049] (1) High-sensitivity, wide-bandwidth ultra-high frequency sensor;

[0050] (2) A data acquisition system with high sampling rate and multi-channel synchronous acquisition capability.

[0051] Ultra-high frequency (UHF) sensor: As the signal front-end for detecting intermittent partial discharge pulses, selecting a high-sensitivity, wide-bandwidth UHF sensor is crucial. After comprehensively comparing typical broadband antennas currently used in the UHF band, including biconical antennas, disc-conical antennas, butterfly antennas, horn antennas, dipole antennas, helical antennas, and log-periodic antennas, this invention ultimately selected a butterfly antenna as an external UHF sensor specifically for receiving electromagnetic wave signals radiated from the casting hole.

[0052] The reason for choosing a butterfly antenna is its planar structure, which is particularly suitable for printing using printed circuit board technology. This makes its manufacturing process simple, inexpensive, and easy to mass-produce. This cost-effectiveness achieved at the core component level gives the invention a significant advantage in commercial promotion, enabling the provision of high-performance solutions at a lower cost. Furthermore, the butterfly antenna, with its application in the ultra-high frequency band, absolute operating bandwidth of 8:1, VSWR of less than 2, linear polarization, directional orientation, maximum gain of 10dB, and small size, meets the requirements for high-precision detection in all aspects.

[0053] Data acquisition system: includes filtering and amplification modules, synchronization modules, power supply modules, control modules, and high-speed sampling components, and achieves high computing power based on GPU parallel computing.

[0054] The filtering and amplification module mainly consists of two parts: anti-interference and amplification, to better capture weak partial discharge signals and avoid the influence of complex electromagnetic environments on partial discharge signal acquisition. The interference signals generated by corona discharge are mainly concentrated below 300MHz. Therefore, this invention uses a bandpass filter composed of cascaded high-pass and low-pass filters, with a frequency range of 300MHz to 1500MHz, to improve the anti-interference capability of the device. The bandpass filter adopts a three-stage passive filter structure, each stage consisting of an LC network of inductors and capacitors, implemented using microstrip line technology to ensure the filter's stability and low insertion loss characteristics at high frequencies. The preamplifier circuit uses a low-noise operational amplifier chip (such as the OPA series or equivalent chips). Its input is connected in parallel with a matching resistor and a coupling capacitor to achieve impedance matching and AC coupling, thereby ensuring the integrity of the signal before amplification. The output is configured with an adjustable potentiometer and a feedback resistor network for closed-loop gain adjustment. Considering that the signals output by sensors are usually weak, the connections between circuits are at a certain distance, and the noise generated by the amplification circuit can also cause interference, the following approach is adopted... Figure 2 The preamplifier of the present invention shown satisfies:

[0055] (a) The input impedance of the measuring amplifier circuit should match the output impedance of the sensor;

[0056] (b) Stable magnification;

[0057] (c) Low noise;

[0058] (d) Low input offset voltage and input offset current, as well as low drift, sufficient bandwidth and slew rate to ensure distortion-free amplified signal;

[0059] (e) High common-mode input range and high common-mode rejection ratio;

[0060] (f) Adjustable closed-loop gain;

[0061] (g) Good linearity and high accuracy;

[0062] (h) Low cost.

[0063] The signal acquired by the sensor is capacitively coupled and then passed through a bandpass filter to remove interference signals before being amplified by 20dB to 40dB. The selected amplifier has low noise interference and can increase the average effective height to 100mm.

[0064] The synchronization module is designed with two synchronization methods: external synchronization and power supply synchronization. The circuit diagrams are shown below. Figure 3 and Figure 4As shown. External synchronization involves obtaining a synchronization signal from the high-voltage power grid at the measurement site. Power synchronization involves extracting a synchronization signal from the AC power supply to the equipment. The design of these two synchronization methods ensures that the reference frequency of the partial discharge detection device is precisely matched with the system frequency of the partial discharge data recording. This matching helps to more accurately distinguish whether the received signal is a true partial discharge or external noise, and allows for better classification of the partial discharge type. The input section of the external synchronization circuit is equipped with a voltage transformer and a current-limiting resistor, and the output is shaped by a Schmitt trigger circuit to ensure a steep edge and strong anti-interference capability of the synchronization signal. The power synchronization circuit connects an RC filter network in series at the transformer output to filter out high-frequency spikes. Subsequently, a comparator circuit converts the AC signal into a stable square wave signal, which is then input to the high-speed sampling component.

[0065] The external synchronization circuit employs a 2mA / 2mA voltage transformer design. The voltage transformer converts the high-voltage signal into a low-voltage signal, which is then input into the synchronization circuit and converted into a square wave signal for output to the high-speed sampling component for signal synchronization. The design parameters are: AC: 1V~400V; 30Hz~300Hz. The power supply synchronization also uses a 2mA / 2mA voltage transformer design, converting the AC220V signal into a square wave signal for output to the high-speed sampling component for signal synchronization. The design parameters are: AC85~240V; 40Hz~70Hz.

[0066] The function of the power module is to convert AC220V alternating current into DC current of various voltages required by the system. Power supply design as follows: Figure 5 As shown, the power module's input is equipped with an EMI filter and surge protection circuit to prevent the impact of transient power grid interference on the detection system. A DC-DC isolated power supply chip is added to the module's power output to ensure isolated power supply between analog and digital circuits and reduce mutual coupling noise. Each power output is equipped with a voltage regulator chip and decoupling capacitor to ensure stable power supply voltage and low ripple.

[0067] The AC / DC converter uses the LHE15-23BXX series modular power supply, with an isolation voltage of 4200V AC, an efficiency of 85%, and output short-circuit protection, overcurrent protection, and overvoltage protection. An LDO chip converts DC 5V to 3V to power the control module.

[0068] The control module of this invention has the following functions:

[0069] (a) To control the gain of the preamplifier module and achieve gain switching from 20 to 40 dB;

[0070] (b) Synchronization signal processing;

[0071] (c) Communicate with the high-speed sampling component via USB interface.

[0072] The core processor of the control module is the AT32F415KBU7-4 with a Cortex-M4 core. This microcontroller has a maximum operating frequency of 150MHz, a memory protection unit, and a DSP instruction set. It also features rich peripheral interfaces: USB OTG, 11 timers, 1 ADC, 2 comparators, and 12 communication interfaces. Furthermore, the microcontroller has a low-power mode. The MCU receives the output signal from the synchronization circuit, processes it, and sends the data to the high-speed sampling component via the USB communication interface. The preamplifier gain is controlled via the gain control pin GCTRLx. The circuit design is as follows: Figure 6 As shown. To ensure the signal integrity of the control module, the USB interface circuit adds a common-mode inductor and ESD protection devices to resist external electrostatic interference; a level conversion circuit is added between the gain control pin and the preamplifier to ensure compatibility and reliability across different voltage domains.

[0073] The high-speed sampling component employs a PDS (Position Sensitive Detector) high-speed acquisition card, consisting of eight 14-bit ADCs with a sampling rate of 5 GSPS, achieving a sampling rate of 20 GSPS, suitable for various programmable applications. The ADC port is supported via a high-performance miniature RF (Radio Frequency) connector. Onboard DDR4, Micro SD card (supports UHS), 10 / 100 / 1000 Ethernet Gigabit Ethernet port, USB JTAG / UART support, shared UART output in JTAG loading mode, and two 100Gbps QSFP28 connectors provide optical ports, enabling efficient high-speed data processing in parallel with analog and digital conversion. This design has the following characteristics:

[0074] (a) Reduced RF signal chain complexity: RFSoC technology simplifies the design, addresses the cost and complexity challenges of pulse acquisition with multiple channels, and helps reduce latency;

[0075] (b) Maximize input / output channel density: The PDS high-speed acquisition card has 8-channel ADC, which can meet the needs of larger system applications;

[0076] (c) Heterogeneous processing capability: With the help of the PDS high-speed acquisition card, a streaming DSP with FPGA and a general-purpose processor for decision-making and control can be integrated into a single module by utilizing RFSoC technology, making more computing resources available for application-specific IP.

[0077] (d) More efficient data offloading: The PDS high-speed acquisition card can utilize optional 2 100G fiber optic interfaces to help offload data more efficiently.

[0078] The hardware technical parameters of the high-speed sampling component are as follows:

[0079] Table 1 Technical Parameters

[0080]

[0081] The physical diagram, interface layout, and description of the high-speed sampling component are as follows:

[0082] Table 2 Interface Description

[0083]

[0084] The PDS high-speed acquisition card is based on the Zynq UltraScale+ XCZU47DR-ffve1156 chip, which combines a Processing System (PS) and Programmable Logic (PL) on the same chip. The PS part of the Zynq UltraScale+ RFSoC uses a Cortex®-A53 64-bit quad-core processor and a Cortex-R5 dual-core real-time processor.

[0085] The high-performance computing of this invention is based on GPU parallel computing units, primarily utilizing NVIDIA's CUDA programming model and OpenMP programming technology. This GPU parallel computing forms the computational backbone for achieving high-precision and rapid diagnosis. High-sampling-rate hardware generates massive amounts of data (0.2ns sampling interval), and processing this data to achieve real-time detection, accurate localization, and rapid diagnosis of sporadic PD (problem-related malfunction) requires enormous computing power. GPUs, with their numerous computing cores, superior floating-point computing capabilities, and extremely high memory bandwidth, are well-suited for large-scale parallel data processing, providing strong support for high-precision and rapid diagnosis. Figure 7 It is a single-node multi-GPU parallel computing framework.

[0086] This invention is based on the working mechanism of GPUs achieving high-performance computing by masking the pipeline with a large number of threads, and optimizes GPU parallel algorithms. The performance optimization method mainly includes two models:

[0087] (a) Memory access computation ratio RMtoC

[0088] MemCycles=GlobalMemLatency×CacheMissRate+CacheLatency×(1-CacheMissRate)(1)

[0089] RMtoC = Memcycles / Compcycles(2)

[0090] Where: MemCycles refers to the average number of clock cycles in a thread bundle's "memory access cycle," and "memory access cycle" refers to the number of clock cycles spent on a global memory access instruction; GlobalMemLatency is the global memory access latency, which refers to the number of clock cycles spent from when a thread initiates a global memory access until the data is returned to the register (which can be used for calculation). CacheMissRate is the cache miss rate, which indicates the proportion of cache misses in all cache accesses, requiring access to a higher level of memory than the currently accessed memory, i.e., slower memory. CacheLatency is the cache access latency, which refers to the average number of clock cycles required to complete a cache access. CompCycles refers to the average number of clock cycles in a thread bundle's "computation cycle," and "computation cycle" refers to the number of clock cycles spent on all computation instructions between two adjacent memory access instructions, obtained through PTX code analysis; RMtoC represents the memory access-computation ratio, reflecting the proportional relationship between memory access latency and computation latency. The calculation formula is shown in Equation (2). The larger the RMtoC, the more difficult it is to mask the memory access latency, and the more the algorithm tends to be memory-restricted; otherwise, it tends to be computation-restricted.

[0091] (b) Memory access thread bundle limit (MWP):

[0092] MWP=(Bandwidth×MemCycles) / (#SM×f×LoadBytesPerWarp)(3)

[0093] LoadBytesPerWarp=LoadBytesPerWarpold×32×CacheMissRate(4)

[0094] Where Bandwidth, f, and #SM are GPU hardware parameters, and LoadBytesPerWarp refers to the number of bytes that each thread bundle accesses from global memory in MemCycles. This refers to the base number of bytes that each thread bundle accesses from global memory in MemCycles. Here, the base number of bytes is the number of bytes that each thread bundle accessed from global memory in MemCycles during the last access.

[0095] Combining RMtoC and MWP can help identify algorithm bottlenecks. RMtoC represents the number of thread bundles required to mask memory access latency. When MWP cannot meet this requirement, the algorithm is a memory-constrained problem, and its performance is mainly limited by memory bandwidth. RMtoC and MWP can also be used to optimize the number of thread bundles on each SM. If the algorithm is a memory-constrained problem, allocating MWP thread bundles to each SM is sufficient. Increasing the number of thread bundles further will not increase the algorithm's bandwidth but will instead cause performance degradation due to register usage and SM management overhead.

[0096] The present invention provides a high-precision sensing device for the location and type diagnosis of sporadic partial discharges in GIS, the overall workflow of which is as follows: Figure 8 As shown:

[0097] (a) The ultra-high frequency sensor is responsible for capturing the weak, occasional partial discharge signals generated inside the GIS equipment;

[0098] (b) The captured raw signal is then sent to the filtering and amplification module and the synchronization module for hardware preprocessing, and is digitized by the high-speed sampling component;

[0099] (c) The massive amount of digitized data is transmitted to the high-performance GPU parallel computing unit, which uses its powerful parallel processing capabilities to perform efficient data analysis.

[0100] This invention increases the sampling rate of the device to a high sampling rate of 5GHz, enabling the partial discharge pulse sampling interval of the GIS partial discharge monitoring device to reach 0.2ns, which greatly increases the number of sampling points and the amount of data. Combined with the use of GPU parallel computing, it significantly improves the pulse processing capability, increasing the ability of the CPU to process only 10-20 pulses per second to 200,000 pulses per second. This pulse processing speedup ratio of up to 666.67 times enables the system to effectively detect and capture intermittent partial discharge signals, thereby significantly improving the detection rate of intermittent partial discharge signals. It can not only accurately capture each intermittent partial discharge pulse, but also enhance the computing power and response speed of the device. Combined with the partial discharge spatial positioning algorithm, it also greatly improves the spatial positioning accuracy, increasing the positioning accuracy from about 0.5m to 0.06m.

[0101] To verify the anti-interference capability, extreme working condition adaptability and detection performance of the device of the present invention under complex field conditions, multi-scenario field tests and comparative experiments were carried out.

[0102] Regarding anti-interference performance, the device of this invention was deployed within a 220kV GIS operating bay, and strong electromagnetic interference sources, including pulse interference, power frequency harmonics, and simulated corona discharge signals, were applied within its operating frequency band, with the highest interference source amplitude reaching 20dBμV. A standard partial discharge calibrator (pulse amplitude 5pC~500pC) was used as a reference in the experiment. The detection rate and false positive rate of the device for real partial discharge pulses were recorded under both the on and off states of the interference source, and compared with the detection results of the reference equipment. The results show that the device of this invention can still maintain a real signal detection rate of ≥95% and a false positive rate of ≤3% even under high low-frequency interference intensity, fully verifying the effectiveness of the bandpass filter design and high common-mode rejection.

[0103] Regarding adaptability to extreme operating conditions, a 500kV GIS substation at an altitude of 2800m was selected as the test environment. This environment is characterized by thin air, a complex electromagnetic environment, and frequent high-current switching operations. The device operated continuously for 168 hours, enduring multiple switching impacts, transient interference from thunderstorms, and local SF6 pressure fluctuations. Test results showed no data loss throughout the test, a sampling synchronization deviation of no more than ±0.5ns, stable positioning accuracy within 0.06m, and no significant attenuation in detection sensitivity. This demonstrates the reliability of the device under extreme conditions such as high altitude, high-impact electromagnetic interference, and pressure fluctuations.

[0104] In practical applications, the device of this invention has completed 240 hours of continuous online monitoring in the GIS equipment of a 500kV substation. Six UHF sensing channels were deployed, covering the busbar and circuit breaker bays. During the monitoring period, a total of 317 sporadic partial discharge pulses were captured, with a single pulse duration of 8ns to 30ns and a peak amplitude of 0.5mV to 2.1mV. The GPU parallel computing unit realized real-time waveform analysis and type identification, identifying 215 suspended particle discharges and 102 internal air gap discharges. Figure 9 Typical raw partial discharge pulse waveforms captured on-site are presented, showing a clear single-peak structure and high signal-to-noise ratio in the time domain. Using a time-difference-of-arrival (TDOA) localization algorithm, the defect is located at the flange connection point on the busbar side of the circuit breaker compartment, and a corresponding long-term cumulative phase distribution map (PRPD) is generated. Figure 10 As shown, the discharge event exhibits a concentrated distribution characteristic in phase. Subsequent disassembly and inspection by maintenance personnel revealed a minor discharge caused by a loose fastener, perfectly consistent with the device's detection and location results. This verification demonstrates that the device of this invention possesses high sensitivity, high accuracy, and strong anti-interference performance in real-time capture, precise location, and type diagnosis of intermittent partial discharges in high-voltage GIS equipment, meeting the practical needs of long-term online monitoring and rapid fault diagnosis in substations.

[0105] Based on the above-described preferred embodiments of the present invention, and through the foregoing description, those skilled in the art can make various changes and modifications without departing from the inventive concept. The technical scope of this invention is not limited to the contents of the specification, but must be determined according to the scope of the claims.

Claims

1. A high-precision sensing device that can be used for GIS accidental partial discharge positioning and type diagnosis, characterized in that, The utility model relates to a kind of GIS partial discharge detection device, including: Ultrahigh frequency sensor: it is high sensitivity wideband antenna, adopts planar bowtie antenna structure, and the working frequency band is 300MHz~1500MHz, non-contact installation is carried out in GIS device cavity surface, and it is used to capture the electromagnetic signal generated by GIS device partial discharge; Data acquisition system: including high-speed sampling component and GPU parallel computing unit, with 5GSPS sampling rate and multi-channel synchronous acquisition capability, based on GPU parallel computing unit, real-time analysis and processing are carried out to acquisition signal, for the capture of partial discharge signal, spatial positioning and discharge type identification; The high-speed sampling component is used to digitize the signal of the ultrahigh frequency sensor, and an equivalent sampling rate of 20GSPS is achieved through multi-channel time interleaving; The GPU parallel computing unit is used to perform real-time spatial positioning and discharge type identification on the digitized signal.

2. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 1, characterized in that, The absolute working bandwidth of the planar bowtie antenna structure is not less than 8:1, the standing wave ratio is less than 2, the polarization mode is linear polarization, and the directivity is directional. The maximum gain of the antenna is 10dB±0.5dB, which is used to distinguish discharge signals from background noise.

3. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 2, characterized in that, The ultrahigh frequency sensor is made by printed circuit board technology.

4. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 1, characterized in that, The data acquisition system also includes a filtering and amplification module, a synchronization module, a power module, and a control module. The data acquisition system performs real-time spatial positioning and discharge type identification on the acquisition signals output by the high-speed sampling component based on the GPU parallel computing unit. The filtering and amplification module and the synchronization module are used to preprocess the signals output by the ultrahigh frequency sensor. The power module is used to provide power for each module of the data acquisition system. The control module is used to coordinate the operation of each module of the data acquisition system. The high-speed sampling component is used to capture and digitize GIS partial discharge pulse signals.

5. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 4, characterized in that, The filtering and amplification module includes a band-pass filter and a preamplifier. The band-pass filter is composed of a high-pass filter and a low-pass filter in cascade. The passband range is 300MHz~1500MHz, which is used to suppress low-frequency interference signals below 300MHz. The preamplifier is used to amplify the signals output by the ultrahigh frequency sensor. The input impedance of the preamplifier matches the output impedance of the ultrahigh frequency sensor.

6. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 4, characterized in that, The synchronization module includes an external synchronization circuit and a power supply synchronization circuit. The external synchronization circuit obtains a synchronization signal from a high-voltage power grid. A voltage transformer is used to convert the high-voltage signal into a low-voltage signal and convert it into a square wave synchronization signal output to the high-speed sampling component of the data acquisition system for processing. The power supply synchronization circuit extracts a synchronization signal from the device AC power supply. A voltage transformer is used to convert the power supply signal into a square wave synchronization signal output to the high-speed sampling component for processing. The reference synchronization signal provided by the external synchronization circuit and the power supply synchronization circuit makes the data acquisition reference frequency of the partial discharge detection device match the power frequency of the power grid.

7. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 4, characterized in that, The control module comprises a microcontroller for coordinating operation of the modules of the data acquisition system; the microcontroller is based on a Cortex-M4 core, the control module is used for receiving and processing the synchronization signal output by the synchronization module, communicating data with the high-speed sampling component, and controlling the gain of the preamplifier to switch between 20-40 dB through a gain control pin.

8. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 4, characterized in that, The high-speed sampling component comprises a multi-channel high-speed analog-to-digital converter array, having 8 14-bit ADC channels, each with a sampling rate of 5 GSPS, supporting multi-channel synchronous sampling; the high-speed sampling component achieves an equivalent sampling rate of 20 GSPS through multi-channel time interleaving to capture partial discharge pulse signals; the analog input end of the high-speed sampling component is connected with the filtering and amplification module via a radio frequency connector, and is equipped with a cache memory for buffering the collected data.

9. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 8, characterized in that, The high-speed sampling component adopts a radio frequency system-on-chip architecture, integrating an FPGA programmable logic unit and a general-purpose processor, and is equipped with an on-chip DDR4 memory; the high-speed sampling component has a 10 / 100 / 1000M Ethernet interface and at least two 100Gbps high-speed optical fiber interfaces, for real-time output or unloading of the collected partial discharge data to external storage and computing devices.

10. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 9, characterized in that, The GPU parallel computing unit uses a GPU to perform real-time processing on the collected partial discharge data; the parallel computing unit executes a signal analysis algorithm, including digital filtering and wavelet denoising of the partial discharge signal, and extracting feature parameters of the discharge pulse, and combining with a built-in database to match typical discharge types, thereby realizing real-time detection, positioning and type diagnosis of the partial discharge.

11. The high-precision sensor device for GIS accidental partial discharge positioning and type diagnosis according to claim 9, characterized in that, The GPU parallel computing unit improves the efficiency of pulse data processing by optimizing the memory computation ratio and the upper limit model of the memory thread bundle: (a) Memory computation ratio: （1） （2） wherein MemCycles refers to the average number of clock cycles of a memory access period of a thread bundle, the memory access period refers to the number of clock cycles spent from initiating a global memory access by a thread to returning data to a register; GlobalMemLatency refers to the global memory access delay, which refers to the number of clock cycles spent from initiating a global memory access by a thread to returning data to a register; CacheMissRate refers to the cache miss rate; CacheLatency refers to the cache access delay, which refers to the average number of clock cycles required for a cache access to be completed; CompCycles refers to the average number of clock cycles of a computation period of a thread bundle, the computation period refers to the number of clock cycles spent by all computation instructions between adjacent memory instructions, which is obtained by analyzing PTX code; RMtoC represents the memory computation ratio, reflecting the proportional relationship between memory delay and computation delay, the larger the RMtoC, the more difficult it is to mask the memory delay, and the algorithm is biased towards memory-limited type, otherwise it is biased towards computation-limited type; (b) Upper limit of memory thread bundle MWP: （3） （4） wherein: Bandwidth, f, #SM are hardware parameters of the GPU, LoadBytesPerWarp refers to the number of bytes accessed from global memory by each thread warp in MemCycles, LoadBytesPerWarp refers to the number of bytes accessed from global memory by each thread warp in MemCycles.

Images