GIS partial discharge defect positioning method and system

Abstract

The invention discloses a GIS partial discharge defect positioning method and system. The method comprises the following steps: respectively acquiring two modal signals of different nodes of a GIS; calculating the peak amplitude, the arrival time, the signal-to-noise ratio and the waveform similarity of the partial discharge electromagnetic pulse signal and the transient current signal of the ith node; then calculating a quality index for each mode, carrying out weight setting on the two modes, finally calculating the fusion credibility of the node i, and obtaining the initial defect position estimation of each node according to the fusion credibility; each node updates the defect position through multi-round interaction of a collaborative game, the influence weight of each node on a final positioning result is dynamically adjusted according to the credibility in the game process, and the final defect position is output. The invention aims to provide a brand new technical scheme with high precision, high robustness and adaptive evolution for GIS partial discharge defect positioning, and provides more reliable technical support for state monitoring and preventive maintenance of power equipment.

Description

Technical Field

[0001] This invention relates to the field of power equipment defect detection technology, specifically to an environmentally friendly GIS (Gas Insulated Switchgear) partial discharge defect location method and system that combines multi-node distributed detection units with a collaborative game decision-making mechanism. This method achieves high-precision location of partial discharge sources by deploying detection units within the GIS that possess adaptive signal processing, information exchange, and collaborative game optimization functions. Background Technology

[0002] Gas-insulated switchgear (GIS) is widely used in various high-voltage and ultra-high-voltage power transmission and transformation systems due to its advantages such as small footprint, high operational reliability, and low maintenance. However, during long-term operation, GIS is highly susceptible to partial discharge defects due to manufacturing defects, installation stress, operating voltage stress, and environmental factors. Partial discharge not only accelerates insulation aging but may even induce serious accidents such as phase-to-phase short circuits and gas chamber breakdown. Therefore, timely and accurate detection and location of partial discharge in GIS is crucial for ensuring the safe operation of power systems. Existing GIS partial discharge detection mainly relies on single-point ultra-high frequency (UHF) sensing or multi-point time difference of arrival (TDOA) methods. While single-point detection methods are simple to deploy, they cannot effectively eliminate on-site noise and interference signals, resulting in extremely limited positioning accuracy. Although multi-point TDOA methods can theoretically achieve higher positioning accuracy, in practical engineering, they require extremely high time synchronization accuracy between detection nodes, typically necessitating expensive synchronization clock systems and complex wiring schemes. Furthermore, they are highly dependent on the integrity of node signals; if some nodes experience signal loss due to faults, obstruction, or channel loss, the positioning accuracy will significantly decrease. Furthermore, most existing methods employ a centralized data processing architecture, requiring all detection data to be transmitted to a central processing unit for computation. This not only increases communication latency and data transmission pressure but also causes the system to completely lose its positioning capability when the central node fails, lacking robustness and adaptability. In recent years, although some research has attempted to introduce multi-sensor information fusion and machine learning techniques to improve positioning performance, these methods often rely on extensive historical data training and offline model updates, making it difficult to meet the requirements of real-time performance and dynamic adaptation in the field. Especially under complex operating conditions such as changing environmental conditions, fluctuating gas composition, and unstable states of some detection nodes, existing methods lack a mechanism capable of autonomous collaboration among multiple nodes, dynamic weight adjustment, and maintaining high positioning accuracy even when some nodes fail. Therefore, a novel GIS partial discharge positioning method is urgently needed. This method can construct multiple distributed detection nodes into units with autonomous computing and collaborative decision-making capabilities, enabling them to automatically converge to a consistent defect location estimate through multiple rounds of interaction and optimization game without centralized dependence. This significantly improves the system's robustness and adaptability while maintaining high accuracy. Summary of the Invention

[0003] The technical problem to be solved by this invention is how to overcome the shortcomings of existing GIS partial discharge location methods in terms of accuracy, robustness and real-time performance.

[0004] The present invention solves the above-mentioned technical problems through the following technical means: The method for locating partial discharge defects in GIS includes the following steps: Two mode signals were acquired from different compartments, busbar sections, and key electrical nodes of the GIS, respectively, including partial discharge electromagnetic pulse signals and transient current signals generated by partial discharge. Calculate the first i The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity of the node partial discharge electromagnetic pulse signal and transient current signal are calculated. Then, the quality index is calculated for each mode, and weights are set for the two modes. Finally, the node partial discharge electromagnetic pulse signal and transient current signal are calculated. i The fusion confidence level is used to obtain the initial defect location estimate for each node. Each node updates the defect location through multiple rounds of collaborative game interaction. During the game, the influence weight of each node on the final location result is dynamically adjusted according to the credibility, and the final defect location is output.

[0005] Furthermore, the two modal signals are obtained by UHF antennas and HFCT sensors installed at compartments, busbar sections, and key electrical nodes, respectively.

[0006] Furthermore, the peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity are calculated as follows: Peak amplitude: ; Arrival time: ; Signal-to-noise ratio: ; Waveform similarity: ,in r For normalized cross-correlation, This is the aligned group template.

[0007] Furthermore, the quality-driven multimodal adaptive fusion is specifically implemented as follows: First, calculate the quality index for each mode:

[0008] in For impulse kurtosis, norm( ) represents the Min-Max normalization across nodes within this event, β1+β2+β3=1; Set the weights of the two modes to

[0009] node i The fusion credibility is defined as:

[0010] in ;λ1,λ2,λ3≥0,λ1+λ2+λ3=1.

[0011] Furthermore, the methods for cooperative game theory and position update are as follows: Step 1. Each node calculates its initial position based on its own signal. ; Step 2: Nodes exchange information, including: 1) Position estimate for the current round. 2) The credibility γ of this node i ; Step 3: Calculate the weighted average of the other node positions:

[0012] in Indicates the node i Other nodes, Indicates the first j The node at the th k The estimated position of the wheel, γ j Indicates the first j The credibility of each node is calculated by dividing the numerator by the position of the credible node and its "discourse power", and the denominator by the total discourse power. Step 4: Update the defect location of the node based on the difference. Each node compares its own viewpoint with the combined viewpoints of other nodes, and leans slightly towards the latter:

[0013] This represents the difference between other nodes and itself, where α is the update step size; Step 5: Determine if convergence has occurred Calculate the positional change of all nodes between the two rounds:

[0014] If the maximum change is less than the threshold ε, then stop the iteration; or the number of iterations reaches the maximum value. K max Stop iteration when the time is right; Step 6: Output the final result Calculate the average again using credibility weighting:

[0015] P final This indicates the final defect location.

[0016] The present invention also provides a GIS partial discharge defect location system, comprising: Partial Discharge Detection Unit: Partial discharge detection units are respectively arranged in different compartments, busbar sections, and key electrical nodes of the GIS; the partial discharge detection unit is used to collect two mode signals: the partial discharge electromagnetic pulse signal and the transient current signal generated by partial discharge at its respective node. Feature calculation unit: calculates the first i The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity of the node partial discharge electromagnetic pulse signal and transient current signal are calculated. Then, the quality index is calculated for each mode, and weights are set for the two modes. Finally, the node partial discharge electromagnetic pulse signal and transient current signal are calculated. i The fusion confidence level is used to obtain the initial defect location estimate for each node. Collaborative game and location update unit: Each node updates the defect location through multiple rounds of collaborative game interaction. During the game, the influence weight of each node on the final location result is dynamically adjusted according to the credibility, and the final defect location is output.

[0017] Furthermore, the two modal signals are obtained by UHF antennas and HFCT sensors installed at compartments, busbar sections, and key electrical nodes, respectively.

[0018] Furthermore, the peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity are calculated as follows: Peak amplitude: ; Arrival time: ; Signal-to-noise ratio: ; Waveform similarity: ,in r For normalized cross-correlation, This is the aligned group template.

[0019] Furthermore, the quality-driven multimodal adaptive fusion is specifically implemented as follows: First, calculate the quality index for each mode:

[0020] in For impulse kurtosis, norm( ) represents the Min-Max normalization across nodes within this event, β1+β2+β3=1; Set the weights of the two modes to

[0021] node i The fusion credibility is defined as:

[0022] in ;λ1,λ2,λ3≥0,λ1+λ2+λ3=1.

[0023] Furthermore, the methods for cooperative game theory and position update are as follows: Step 1. Each node calculates its initial position based on its own signal. ; Step 2: Nodes exchange information, including: 1) Position estimate for the current round. 2) The credibility γ of this node i ; Step 3: Calculate the weighted average of the other node positions:

[0024] in Indicates the node i Other nodes, Indicates the first j The node at the th k The estimated position of the wheel, γ j Indicates the first j The credibility of each node is calculated by dividing the numerator by the position of the credible node and its "discourse power", and the denominator by the total discourse power. Step 4: Update the defect location of the node based on the difference. Each node compares its own viewpoint with the combined viewpoints of other nodes, and leans slightly towards the latter:

[0025] This represents the difference between other nodes and itself, where α is the update step size; Step 5: Determine if convergence has occurred Calculate the positional change of all nodes between the two rounds:

[0026] If the maximum change is less than the threshold ε, then stop the iteration; or the number of iterations reaches the maximum value. K max Stop iteration when the time is right; Step 6: Output the final result Calculate the average again using credibility weighting:

[0027] P final This indicates the final defect location.

[0028] The advantages of this invention are: The purpose of this invention is to address the shortcomings of existing GIS partial discharge location methods in terms of accuracy, robustness, and real-time performance by proposing a location method and system based on a distributed collaborative and game-theoretic optimization mechanism involving multiple detection nodes. This method constructs partial discharge detection nodes distributed across different compartments of the GIS into functional units with autonomous computing, signal feature analysis, dynamic weight adjustment, and multi-round collaborative game capabilities. This enables them to achieve self-organized, consistent convergence to the defect location through mutual information exchange and reward-driven mechanisms between nodes, without relying on a central processor. This invention aims to solve the technical challenges of existing centralized location systems, such as susceptibility to single-point failures of central nodes, decreased location accuracy in the event of partial node failure or signal loss, and the inability to adjust location strategies in real-time according to the on-site operating status. By introducing mechanisms such as dynamic evaluation of node credibility, optimization of game-theoretic reward functions, and removal of abnormal nodes, this invention maintains high location accuracy and fast convergence speed under complex operating environments and different working conditions, while also possessing good fault tolerance. Furthermore, this method integrates multi-dimensional features such as amplitude, time difference of arrival, and waveform shape in signal acquisition, enabling the positioning result to be based on a comprehensive judgment of multiple features rather than a single parameter, thus further reducing the risk of misjudgment. In summary, this invention aims to provide a novel, highly accurate, robust, and adaptively evolving technical solution for GIS partial discharge defect positioning, offering more reliable technical support for the condition monitoring and preventative maintenance of power equipment. Attached Figure Description

[0029] Figure 1 This is a system architecture diagram in Embodiment 1 of the present invention. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Example 1 This embodiment provides a GIS partial discharge defect location system based on a multi-detection node distributed collaborative and game-theoretic optimization mechanism. The system consists of several partial discharge detection units deployed at different compartments, busbar sections, and key electrical nodes within the GIS system. Each detection unit has a signal acquisition module, a feature extraction module, a collaborative game-theoretic decision-making module, and a communication module. It can achieve high-precision, consistent estimation of partial discharge defect locations through multiple rounds of data exchange and strategy updates between nodes, without centralized control. Specifically: The system contains at least three partial discharge detection units, each of which includes: Signal acquisition module: includes a UHF antenna for receiving partial discharge electromagnetic pulse signals and an HFCT sensor for detecting transient current signals generated by partial discharge; Feature extraction module: used to extract peak amplitude from the acquired signal. A i Arrival time t i and waveform shape similarity x i Calculation; Collaborative game decision-making module: Based on an embedded processor, it realizes multi-round game iteration, payoff function calculation and position estimation update; Communication module: Supports low-latency broadcasting or point-to-point transmission for exchanging data and policies between nodes.

[0032] Example 2 Based on the system in Example 1, this example will specifically describe the GIS partial discharge defect location method.

[0033] Step 1: Signal Synchronization and Feature Calculation To ensure the comparability of multi-source features in the time domain, all nodes are synchronized to a synchronization error of ≤10 ns via IEEE-1588 (PTP). Each node simultaneously acquires UHF signals and HFCT transient current signals. Sampling rate: UHF ≥ 1 GS / s, HFCT ≥ 100 MS / s; Preprocessing: bandpass filtering (UHF: 300–800 MHz, HFCT: 100 kHz–20 MHz), detrending, normalization to [0,1].

[0034] Regarding a partial discharge incident, in the first i The node calculates the following basic characteristics: (The superscript (m) in the formulas below indicates the signal mode, i.e.: m =UHF indicates that this feature originates from ultra-high frequency electromagnetic signals; m=HFCT indicates that the feature originates from a transient current signal. In other words, this embodiment simultaneously acquires signals from two different sensors, calculates a set of features for each signal, and then adaptively fuses them based on quality.

[0035] 1. Peak amplitude: ; 2. Arrival Time: (or using the first-order threshold of the envelope) i (Definition of first arrival time). 3. Signal-to-noise ratio: ; 4. Waveform similarity: ,in r For normalized cross-correlation, The aligned "group template" (see below).

[0036] The "group template" is composed of the normalized waveforms of all nodes in this event. After coarse alignment using TDOA (Time Difference of Arrival) and taking the median point-by-point, the following is obtained:

[0037] Cross-node time difference is defined as:

[0038] The initial TDOA value is then obtained using the least squares arrival time solution, and the amplitude decay method is used to verify the consistency with the TDOA result.

[0039] To avoid positioning performance degradation due to single-mode failure, this embodiment employs quality-driven multimodal adaptive fusion. First, a quality index is calculated for each mode:

[0040] in For impulse kurtosis (to distinguish between real impulses and broadband noise), norm( ) represents the Min-Max normalization across nodes within this event, β1+β2+β3=1.

[0041] Set the weights of the two modes to

[0042] node i The fusion credibility is defined as:

[0043] in λ1,λ2,λ3≥0,λ1+λ2+λ3=1. When a certain mode is completely unusable (e.g., UHF is shielded), its As the value approaches zero, the weights naturally shift to another modality.

[0044] Step 2: Cooperative Game and Position Update After completing the initial position estimation of each detection node, the core innovation of this invention is that each node does not directly submit its results to the central processor, but gradually reaches a consensus and obtains a more accurate final position through a multi-round interactive optimization mechanism of "cooperative game".

[0045] This mechanism can be understood as follows: multiple nodes are like "holding a meeting to discuss" each other. In each round, each node puts forward its own judgment and makes adjustments by referring to the opinions of other nodes. In the end, everyone's opinions will get closer and closer to the true position.

[0046] The entire process includes the following steps: Step 11: Each node first proposes its own "initial judgment." This step provides the starting point for the game. Specifically, each node calculates an initial position based on its own signal, using either the Time Difference of Arrival (TDOA) or the amplitude decay method. Even if the initial estimate is inaccurate, it can still provide a reference starting point for subsequent optimization. Step 12: Nodes exchange information. This step lets each node know "how others see it." The exchanged information includes: 1) the estimated position for the current round. 2) The credibility γ of this node i (Used to indicate "how reliable what I say is"), this step is to let each node understand each other's views and credibility, in preparation for the subsequent "weighted negotiation".

[0047] Step 13: Calculate the weighted average of the other node positions. Each node calculates the "comprehensive opinion of other nodes" based on the current location information of other nodes, using credibility as a weight.

[0048]

[0049] Left side This represents all nodes other than i. γ represents the estimated position of the j-th node in the k-th round. j Let represent the credibility of the j-th node. The numerator represents the position of the credible node multiplied by its "discourse power," and the denominator represents the total discourse power. The overall formula is the "weighted average opinion" of all nodes. The significance of this setting is that nodes with high credibility have a greater influence, while nodes with low credibility have a smaller or even negligible influence, ensuring that "correct nodes drive global convergence."

[0050] Step 14: Nodes update their positions based on the differences. Each node compares its own viewpoint with the combined viewpoints of other nodes, and leans slightly towards the latter:

[0051] This represents the distance between other nodes and itself. α: update step size (between 0 and 1, determining how much closer to the node each time). For example, if α = 0.5, it means adjusting by half the distance each time. The purpose of this setting is to prevent jumping too far at once, which could lead to oscillations or errors, and to ensure stable and controllable convergence.

[0052] Step 15: Determine if convergence has occurred Calculate the positional change of all nodes between the two rounds:

[0053] If this maximum change value is less than a certain threshold ε, it means that everyone's views are very similar, and we can stop. Or, the iteration count has reached its maximum value. K max The process also stops when the time comes. Setting a maximum number of iterations is to prevent "infinite discussion" and ensure real-time performance.

[0054] Step 16: Output the final result The final position is not simply the result of the last node, but rather an average calculated again using confidence weighting:

[0055] The final output of the entire game-theoretic optimization algorithm is a unique coordinate of the discharge defect location (denoted as P). final ).

[0056] In summary, the core technical points and principles of this embodiment are as follows: 1. Method for constructing distributed autonomous detection nodes The partial discharge detection units deployed in different compartments of the GIS are designed as distributed nodes with autonomous functions such as signal acquisition, feature extraction, confidence calculation, game strategy update and data communication. They can independently complete local calculation and strategy adjustment, and complete defect location through inter-node interaction and collaboration.

[0057] 2. Node credibility assessment mechanism based on multi-dimensional feature fusion By fusing peak amplitude A i Arrival time difference Δt ij Waveform similarity x i and signal-to-noise ratio SNRi- and other multi-dimensional signal characteristics are used to calculate the node reliability index. c i The system dynamically adjusts the weight of each node's influence on the final location result based on credibility during the game process.

[0058] 3. Cooperative Game Theory Optimization Algorithm for GIS Localization In the defect localization process, each node is regarded as a player in a game, and a multi-round collaborative game strategy based on the payoff function is adopted to ensure that the positional information between nodes converges to a globally consistent defect location estimate within a finite number of rounds.

[0059] 4. Dynamic Abnormal Node Removal and Robustness Enhancement Mechanism During the game iteration process, if the node credibility c i If the value is continuously below a preset threshold, the node is marked as abnormal and removed from subsequent iterations to avoid interference from low-quality signals on positioning accuracy and to ensure that the system can still achieve stable positioning even if some nodes fail.

[0060] 5. Adaptive parameter adjustment method based on environmental changes When changes in the GIS operating environment (gas medium composition, pressure, temperature, etc.) cause changes in signal propagation characteristics, the system can adjust the game weight parameters online based on a comparison of historical positioning and actual maintenance results. and iteration step size This ensures positioning accuracy and stability during long-term operation.

[0061] 6. Low-latency distributed implementation architecture for engineering deployment By employing low-bandwidth inter-node communication and distributed parallel computing, real-time positioning can be achieved without relying on a central processing unit, reducing system deployment costs and complexity, and providing good scalability.

[0062] Example 3 This embodiment describes the application of Embodiments 1 and 2 in conjunction with a specific case. The details are as follows: A 110 kV GIS includes circuit breaker compartment A, busbar compartment B, and joint location C. Four detection nodes are arranged in A, B, C, and the adjacent compartment D. Each node is equipped with both UHF and HFCT sensors. The UHF sampling rate is 1.5 GS / s, and the HFCT sampling rate is 200 MS / s. The nodes are time-synchronized via fiber optic cable using IEEE 1588 PTP, with an error of less than 10 ns. When the discharge occurs, each node simultaneously acquires two signals. Nodes 1, 2, and 3 detect a clear UHF pulse (amplitude 0.8~1.0 V), and the HFCT signal is also relatively clear. Node 4, due to its installation location adjacent to a metal partition wall, has a UHF signal of only 0.15 V, and the HFCT signal is interfered with by noise, resulting in an SNR of less than 5 dB. Feature extraction and confidence calculation then begin (examples for nodes 1 and 4): Node 1: SNR(UHF) = 20 dB, ξ(UHF) = 0.92 → High confidence; Node 4: SNR(UHF) = 8 dB, ξ(UHF) = 0.45 → Low confidence. The confidence values ​​γi in this round are: γ1 = 0.32, γ2 = 0.29, γ3 = 0.28, γ4 = 0.11. Initial positioning (TDOA + amplitude attenuation method): The TDOA estimation result deviates from the actual position by 1.5 meters. Node 4 has a larger error due to a weak signal, but its low confidence limit its impact. Then, the collaborative game process begins. Cooperative game process (6 rounds of iteration): Round 1: Node swapping position estimation and γi; Round 2: Nodes 1, 2, and 3 tend to be consistent, while node 4 shows significant differences, and γ4 decreases to 0.07; Round 3: Node 4 is soft-suppressed due to continuous low credibility (weight multiplied by 0.2); Round 4: Core node location difference <0.6m; Round 5: The location difference of all high-confidence nodes is <0.3m; Round 6: The convergence threshold ε = 0.2 m is met, and the iteration stops.

[0063] Abnormal node handling: Node 4 was identified as an abnormal node and removed in the fourth round because its credibility score was less than 0.1 for three consecutive rounds.

[0064] Final result: The final weighted position is (x: 1.2, y: 3.8, z: 1.1) m. Error 0.28 m < 0.4 m Global confidence level Γ = 0.86 Compared to traditional methods: Traditional TDOA error is 1.5 m The error of this invention is 0.28 μm, representing an improvement of over 80%. Furthermore, even if one node fails, it can still converge stably. This embodiment combines distributed partial discharge detection nodes with a multi-round collaborative game mechanism to achieve high-precision, adaptive localization of partial discharge defects in GIS under decentralized control, achieving the following significant technical effects: Firstly, regarding positioning accuracy, this embodiment integrates multi-dimensional features such as amplitude attenuation characteristics, time difference of arrival (TDOA), and waveform similarity during the signal processing stage. It also employs a confidence-weighted approach for feature fusion, ensuring that the positioning results do not rely on a single parameter and significantly reducing the impact of environmental noise and signal attenuation. Simulation and field tests have verified that in typical 110kV and 220kV GIS equipment, the positioning error can be controlled within 0.5 meters, improving accuracy by more than 30% compared to traditional single-point or centralized TDOA methods.

[0065] Secondly, regarding system robustness, this embodiment introduces a dynamic node credibility assessment and anomaly removal mechanism. This allows the system to automatically reduce the weight of some detection nodes, or even remove them from the game process, when their signals are distorted or lost due to channel attenuation, hardware failure, or shielding effects, ensuring that the remaining nodes can still achieve stable positioning. This mechanism significantly improves the system's fault tolerance in environments with node failure, signal loss, and high interference.

[0066] Third, regarding real-time performance and convergence speed, the game-theoretic optimization algorithm employs a distributed parallel computing model. Each detection node independently performs calculations and policy updates, requiring only the exchange of a small number of location estimates and confidence parameters to complete a round of the game. Testing showed that with six detection nodes participating simultaneously, the average time to converge to a stable positioning result is less than 1.5 seconds, meeting the requirements for online real-time monitoring.

[0067] Fourth, regarding adaptability, the game payoff function and weighting parameters of this invention can be dynamically adjusted according to changes in the operating environment. For example, when changes in the gas composition, pressure, or temperature of the GIS gas chamber cause alterations in signal propagation speed and attenuation characteristics, the system can adjust the weighting coefficients online based on the difference between historical positioning results and the actual confirmed maintenance location. and iteration step size This ensures stable positioning during long-term operation.

[0068] Finally, regarding engineering feasibility, this invention does not rely on high-cost centralized synchronization systems and high-capacity data transmission links. Instead, it utilizes low-bandwidth communication and distributed computing to reduce system costs and deployment complexity. Furthermore, since each node possesses autonomous decision-making capabilities, expanding the detection range only requires increasing the number of nodes and connecting them to the communication network, without requiring large-scale modifications to the original system architecture, thus exhibiting excellent scalability.

[0069] In summary, this embodiment not only significantly improves upon existing technologies in terms of positioning accuracy, robustness, and real-time performance, but also possesses high adaptability and engineering deployability, providing reliable technical support for online monitoring and preventive maintenance of partial discharge defects in GIS.

[0070] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for locating partial discharge defects in GIS, characterized in that, Includes the following steps: Two mode signals were acquired from different compartments, busbar sections, and key electrical nodes of the GIS, respectively, including partial discharge electromagnetic pulse signals and transient current signals generated by partial discharge. Calculate the first i The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity of the node partial discharge electromagnetic pulse signal and transient current signal are calculated. Then, the quality index is calculated for each mode, and weights are set for the two modes. Finally, the node partial discharge electromagnetic pulse signal and transient current signal are calculated. i The fusion confidence level is used to obtain the initial defect location estimate for each node. Each node updates the defect location through multiple rounds of collaborative game interaction. During the game, the influence weight of each node on the final location result is dynamically adjusted according to the credibility, and the final defect location is output.

2. The GIS partial discharge defect location method according to claim 1, characterized in that, The two modal signals were obtained by the UHF antenna and HFCT sensor installed at the compartment, bus section, and key electrical node, respectively.

3. The GIS partial discharge defect location method according to claim 1, characterized in that, The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity are calculated as follows: Peak amplitude: ; Arrival time: ; Signal-to-noise ratio: ; Waveform similarity: ,in ρ For normalized cross-correlation, This is the aligned group template.

4. The GIS partial discharge defect location method according to claim 1, characterized in that, The quality-driven multimodal adaptive fusion is specifically as follows: First, calculate the quality index for each mode: in For impulse kurtosis, norm( ) represents the Min-Max normalization across nodes within this event, β1+β2+β3=1; Set the weights of the two modes to node i The fusion credibility is defined as: among them ;λ1,λ2,λ3≥0,λ1+λ2+λ3=1.

5. The GIS partial discharge defect location method according to any one of claims 1 to 4, characterized in that, The methods for cooperative game theory and position update are as follows: Step 1. Each node calculates its initial position based on its own signal. ; Step 2: Nodes exchange information, including: 1) Position estimate for the current round. 2) The credibility γ of this node i ; Step 3: Calculate the weighted average of the other node positions: in Indicates the node i Other nodes, Indicates the first j The node at the th k The estimated position of the wheel, γ j Indicates the first j The credibility of a node is calculated by dividing the numerator by the position of the credible node and its "voice power", and the denominator by the total voice power. Step 4: Update the defect location of the node based on the difference. Each node compares its own viewpoint with the combined viewpoints of other nodes, and leans slightly towards the latter: This represents the difference between other nodes and itself, where α is the update step size; Step 5: Determine if convergence has occurred Calculate the positional change of all nodes between the two rounds: If the maximum change is less than the threshold ε, then stop the iteration; or the number of iterations reaches the maximum value. K max Stop iteration when the time is right; Step 6: Output the final result Calculate the average again using credibility weighting: P final This indicates the final defect location.

6. A GIS partial discharge defect location system, characterized in that, include: Partial Discharge Detection Unit: Partial discharge detection units are respectively arranged in different compartments, busbar sections, and key electrical nodes of the GIS; the partial discharge detection unit is used to collect two mode signals: the partial discharge electromagnetic pulse signal and the transient current signal generated by partial discharge at its respective node. Feature calculation unit: calculates the first i The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity of the node partial discharge electromagnetic pulse signal and transient current signal are calculated. Then, the quality index is calculated for each mode, and weights are set for the two modes. Finally, the node partial discharge electromagnetic pulse signal and transient current signal are calculated. i The fusion confidence level is used to obtain the initial defect location estimate for each node. Collaborative game and location update unit: Each node updates the defect location through multiple rounds of collaborative game interaction. During the game, the influence weight of each node on the final location result is dynamically adjusted according to the credibility, and the final defect location is output.

7. The GIS partial discharge defect location system according to claim 6, characterized in that, The two modal signals are obtained by the UHF antenna and the HFCT sensor installed at the compartment, bus section, and key electrical node, respectively.

8. The GIS partial discharge defect location system according to claim 6, characterized in that, The peak amplitude, arrival time, signal-to-noise ratio, and waveform similarity are calculated as follows: Peak amplitude: ; Arrival time: ; Signal-to-noise ratio: ; Waveform similarity: ,in ρ For normalized cross-correlation, This is the aligned group template.

9. The GIS partial discharge defect location system according to claim 1, characterized in that, The quality-driven multimodal adaptive fusion is specifically as follows: First, calculate the quality index for each mode: in For impulse kurtosis, norm( ) represents the Min-Max normalization across nodes within this event, β1+β2+β3=1; Set the weights of the two modes to node i The fusion credibility is defined as: among them ;λ1,λ2,λ3≥0,λ1+λ2+λ3=1.

10. The GIS partial discharge defect location system according to any one of claims 6 to 9, characterized in that, The methods for cooperative game theory and position update are as follows: Step 1. Each node calculates its initial position based on its own signal. ; Step 2: Nodes exchange information, including: 1) Position estimate for the current round. 2) The credibility γ of this node i ; Step 3: Calculate the weighted average of the other node positions: in Indicates the node i Other nodes, Indicates the first j The node at the th k The estimated position of the wheel, γ j Indicates the first j The credibility of a node is calculated by dividing the numerator by the position of the credible node and its "voice power", and the denominator by the total voice power. Step 4: Update the defect location of the node based on the difference. Each node compares its own viewpoint with the combined viewpoints of other nodes, and leans slightly towards the latter: This represents the difference between other nodes and itself, where α is the update step size; Step 5: Determine if convergence has occurred Calculate the positional change of all nodes between the two rounds: If the maximum change is less than the threshold ε, then stop the iteration; or the number of iterations reaches the maximum value. K max Stop iteration when the time is right; Step 6: Output the final result Calculate the average again using credibility weighting: P final This indicates the final defect location.

Images