Method for identifying and analyzing resonance overvoltage in different types of frequent grounding forms

Abstract

The invention discloses a method for identifying and analyzing resonance overvoltage in different types of frequent grounding forms. The method comprises the following steps: S1, establishing a high-quality and synchronous data foundation for subsequent analysis through data acquisition, synchronization and preprocessing analysis; s2, automatically judging the current grounding operation mode of the system at the initial stage of a fault through feature extraction and mode matching analysis; s3, fault types are diagnosed and judged through transient state and steady state characteristic conjoint analysis, and specific fault attributes causing resonance risks are deeply diagnosed in a known grounding form; s4, through dynamic time-frequency analysis and key characteristic quantity calculation, frequency spectrum and waveform characteristics possibly representing resonance are extracted from the voltage waveform; s5, associating the analysis results through multi-source information fusion, judging whether resonance occurs or not, determining the type of the resonance, and explaining the causal relationship between the resonance and the grounding form; s6, constructing an event time axis, forming a structured analysis conclusion, and tracing the full view of the event; the method has the advantages of process dominance, logic interpretability and conclusion three-dimensional.

Description

Technical Field

[0001] This invention belongs to the field of fault diagnosis technology, specifically relating to a method for identifying and analyzing resonant overvoltage under different types of frequent grounding. Background Technology

[0002] Resonant overvoltage occurs when the parameters of inductance (such as transformers and arc suppression coils) and capacitance (such as line-to-ground capacitance) in a power system match under specific conditions (such as during a ground fault), forming a resonant circuit. This results in overvoltages with high amplitude and long duration. Resonant overvoltages pose significant safety hazards to the insulation of various devices (such as cable heads and instrument transformers), causing problems such as insulation damage and breakdown, secondary system and protection malfunctions, and severe deterioration of power quality. Furthermore, different types of frequent grounding will create drastically different resonant conditions. Currently, there are the following problems in identifying and analyzing resonant overvoltages: 1. The monitoring system only issues grounding or overvoltage alarms, but whether the two are related and what kind of grounding causes what kind of overvoltage depends entirely on human experience and is very easy to misjudge. 2. It can only be dealt with after the resonant overvoltage has formed or even after the equipment has been damaged, which is too late. 3. Fault diagnosis relies heavily on expert experience, and the knowledge is difficult to quantify, preserve, and pass on, posing a risk of technical gaps due to personnel changes; 4. Prevention and control measures (such as the optional installation of harmonic suppression devices) are often based on general principles and lack data support for specific problems of the station, resulting in poor effectiveness and potential waste of investment; To address the aforementioned issues, it is essential to develop a method for identifying and analyzing resonant overvoltages under different types of frequent grounding conditions. Summary of the Invention

[0003] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for identifying and analyzing resonant overvoltage under different types of frequent grounding, which makes the process explicit, the logic interpretable, and the conclusions three-dimensional.

[0004] The objective of this invention is achieved as follows: a method for identifying and analyzing resonant overvoltage under different types of frequent grounding, comprising the following steps: S1, Synchronous Acquisition and Preprocessing Analysis of Multi-Source Heterogeneous Data: Through data acquisition, synchronization and preprocessing analysis, a high-quality and synchronized data foundation is laid for subsequent analysis; S2, Adaptive Grounding Mode Identification and Analysis: Through feature extraction and pattern matching analysis, the current grounding operation mode of the system is automatically determined in the early stage of a fault. S3, Fine-grained diagnostic analysis of grounding fault types: Through joint analysis of transient and steady-state characteristics, the fault type is diagnosed and determined. Under known grounding conditions, the specific fault attributes that cause resonance risk are diagnosed in depth. S4, Resonant Overvoltage Feature Extraction and Mode Screening: Through dynamic time-frequency analysis and key feature quantity calculation, the spectrum and waveform features that may characterize resonance are extracted from the voltage waveform. S5, Resonance Mechanism Correlation and Comprehensive Judgment Analysis: By fusing multi-source information, the aforementioned analysis results are correlated to determine whether resonance occurs, identify its type, and explain its causal relationship with the grounding form; S6, Full Event Tracing and Analysis Report Generation: Constructs an event timeline, generates structured analytical conclusions, and traces the entire event; S7, Case Library Iteration and Self-Optimization of Analysis Methods: Construct a case library for incremental learning of the model and dynamically update the rule library, enabling the system to have continuous learning capabilities and improve the accuracy of analysis.

[0005] Preferably, step S1 includes the following steps: S11, collect three types of data: Electrical transient / steady-state quantities: three-phase voltage, zero-sequence voltage, zero-sequence current of each branch, including lines, arc suppression coils, and capacitor banks, with a sampling rate of not less than 10kHz; System status variables: neutral grounding mode switching status, arc suppression coil position, and related switch position signals; Historical and parameter data: power grid structure parameters, historical fault recording data, equipment nameplate parameters; S12, Synchronization and Preprocessing Analysis: Using BeiDou / GPS to achieve microsecond-level synchronization of network data, and performing filtering, noise reduction, and per-unit preprocessing on the raw data, analyzing its basic statistical characteristics, and determining the ground fault initiation time T0.

[0006] Preferably, step S2 includes the following steps: S21, Feature Extraction: Analyze the data within one power frequency cycle before and after T0, and extract key features, including the steady-state amplitude and phase of the neutral point displacement voltage U0, the amplitude ratio and phase difference between the arc suppression coil current IL and the total zero-sequence current IΣ of the system, and the consistency of the capacitive characteristic direction of the zero-sequence current 3I0 of each line. S22, Pattern Matching Analysis: Match the features extracted in step S21 with the pre-set grounding form feature library, and use a clustering algorithm or lightweight classification model to achieve online adaptive identification of grounding forms and output the confidence level.

[0007] Preferably, the grounding form feature library includes ungrounded, grounded via arc suppression coil, and grounded via resistor.

[0008] Preferably, step S3 includes the following steps: S31, Joint analysis of transient and steady-state characteristics: Transient analysis: Analyze the polarity of the first half-wave, the high-frequency oscillation frequency, and the decay rate of the zero-sequence current within 5ms after T0; Steady-state waveform analysis: Analyze the stability, periodic fluctuations and distortion patterns of the zero-sequence voltage within hundreds of milliseconds after T0; S32, Diagnosis and Judgment: Construct a multi-dimensional decision tree or support vector machine (SVM) model with transient and steady-state characteristics as input, and diagnose the fault types as follows: metallic grounding, which is stable and has clear characteristics; intermittent arc grounding, which is fluctuating and has high-frequency reignition; and high-resistance grounding, which is low amplitude and has weak characteristics.

[0009] Preferably, step S4 includes the following steps: S41, Dynamic Time-Frequency Analysis: Perform continuous wavelet transform (CWT) or short-time Fourier transform (STFT) on the three-phase voltage data after T0 to generate a time-frequency spectrum and observe the evolution of frequency components over time. S42, Calculation of key feature quantities: Dominant frequency analysis: Identify frequency sub-frequency or high-frequency components with significantly prominent energy, other than the fundamental frequency; Amplitude relationship analysis: Calculate the peak voltage of each phase, the effective value of U0, and its ratio to the rated value; Waveform distortion analysis: Calculate the total harmonic distortion (THD) and the content of each harmonic.

[0010] Preferably, step S5 includes the following steps: S51, Multi-source information fusion: The grounding form of step S2, the fault type of step S3, and the electrical characteristics of step S4 are combined into a joint feature vector. S52, rule-based mechanism inference: invoke a pre-built expert rule base to perform logical inference; S53, Model-Based Intelligent Judgment: Employs ensemble learning algorithms such as random forest or gradient boosting tree to train the model to learn the complex mapping relationship between joint feature vectors and resonance type / occurrence, outputs probabilistic judgment results, and cross-validates with rule inference results.

[0011] Preferably, step S6 includes the following steps: S61, Build an event timeline: Automatically generate a refined timeline starting from T0, building key nodes; S62 generates intelligent analysis reports, which automatically include: Event summary: time, location, grounding method, fault type, resonance type; Analysis process: key feature data, spectrogram, and logical chain of judgment criteria; Conclusions and mechanism explanation: Clearly state "what kind of grounding fault, under what system, induced what kind of resonance"; Impact assessment: Analysis of potential or actual impacts on equipment insulation, protection operation, and power quality; Preliminary recommendations: Based on the analysis findings, we recommend targeted measures.

[0012] Preferably, the key nodes include grounding occurrence, fault type determination, resonance excitation, protection action, and voltage recovery.

[0013] Preferably, step S7 includes the following steps: S71, Case Library Construction: Store each accurate analysis report confirmed manually into the typical case library according to the three-dimensional tags of "grounding type - fault type - resonance type"; S72, Incremental Model Learning: Periodically use new case data to incrementally train the intelligent judgment model in step S5, optimize its parameters, and make it adapt to changes in the power grid structure. S73, Dynamic Rule Base Updates: Based on newly discovered special cases or misjudgment analysis, the expert rule base is supplemented and revised to keep the analysis logic up-to-date.

[0014] Due to the adoption of the above technical solutions, the beneficial effects of this invention are as follows: This invention emphasizes a complete analytical chain from raw data to mechanism explanation, rather than a single identification result; This invention, through rule inference and feature association, enables the judgment result to have traceable and interpretable physical or logical basis, which is in line with the thinking of engineers; The output of this invention is not only a resonance tag, but also an in-depth analysis report containing causes, processes, consequences and suggestions, directly supporting high-level decision-making; In summary, this invention has the advantages of explicit process, interpretable logic, and comprehensive conclusions. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow of the present invention. Detailed Implementation

[0016] The technical solution of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0017] like Figure 1 As shown, this invention provides a method for identifying and analyzing resonant overvoltage under different types of frequent grounding, including the following steps: S1, Synchronous Acquisition and Preprocessing Analysis of Multi-Source Heterogeneous Data: Through data acquisition, synchronization and preprocessing analysis, a high-quality and synchronized data foundation is laid for subsequent analysis.

[0018] S11, collect three types of data: Electrical transient / steady-state quantities: three-phase voltage, zero-sequence voltage, and zero-sequence current of each branch (including lines, arc suppression coils, and capacitor banks), with a sampling rate of not less than 10kHz; System status variables: neutral grounding mode switching status, arc suppression coil position, and related switch position signals; Historical and parameter data: power grid structure parameters, historical fault waveform data, equipment nameplate parameters.

[0019] S12, Synchronization and Preprocessing Analysis: Using BeiDou / GPS to achieve microsecond-level synchronization of data across the entire network, and performing filtering, noise reduction, and per-unit preprocessing on the raw data, analyzing its basic statistical characteristics (such as abrupt change points and RMS value changes), and determining the ground fault initiation time T0.

[0020] S2, Adaptive Grounding Mode Identification and Analysis: Through feature extraction and pattern matching analysis, the current grounding operation mode of the system is automatically determined in the early stage of a fault.

[0021] S21, Feature Extraction: Analyze the data within one power frequency cycle before and after T0, and extract key features, including the steady-state amplitude and phase of the neutral point displacement voltage U0, the amplitude ratio and phase difference between the arc suppression coil current IL and the total zero-sequence current IΣ of the system, and the consistency of the capacitive characteristics of the zero-sequence current 3I0 of each line.

[0022] S22, Pattern Matching Analysis: The features extracted in step S21 are matched with a pre-set grounding form feature library (ungrounded, grounded through an arc suppression coil, grounded through a resistor). A clustering algorithm (such as K-means) or a lightweight classification model is used to achieve online adaptive identification of the grounding form and output the confidence level.

[0023] S3, Fine-grained diagnostic analysis of grounding fault types: Through joint analysis of transient and steady-state characteristics, the fault type is diagnosed and determined. Under known grounding conditions, the specific fault attributes that cause resonance risk are diagnosed in depth.

[0024] S31, Joint analysis of transient and steady-state characteristics: Transient analysis: Analyze the polarity of the first half-wave, the high-frequency oscillation frequency, and the decay rate of the zero-sequence current within 5ms after T0; Steady-state waveform analysis: Analyze the stability, periodic fluctuations and distortion patterns of the zero-sequence voltage within hundreds of milliseconds after T0.

[0025] S32, Diagnostic Decision: Construct a multi-dimensional decision tree or support vector machine (SVM) model with transient and steady-state features as input to diagnose the fault type as follows: Metallic grounding is stable and has clear characteristics; Intermittent arc grounding is characterized by fluctuations and high-frequency reignition. High-resistance grounding is characterized by low amplitude and weak characteristics.

[0026] Thus, through the correlation analysis of steps S1 to S3, this invention establishes a clear diagnostic chain, which can clearly analyze that: for example, the overvoltage was caused by a third high-frequency ferroresonance triggered by an intermittent arc grounding fault occurring under the 10kV busbar grounding system via the arc suppression coil. This provides an accurate diagnostic basis for all subsequent operations.

[0027] S4, Resonant Overvoltage Feature Extraction and Pattern Screening: Through dynamic time-frequency analysis and key feature quantity calculation, the spectrum and waveform features that may characterize resonance are extracted from the voltage waveform.

[0028] S41, Dynamic Time-Frequency Analysis: Perform continuous wavelet transform (CWT) or short-time Fourier transform (STFT) on the three-phase voltage data after T0 to generate a time-frequency spectrum and observe the evolution of frequency components over time.

[0029] S42, Calculation of key feature quantities: Dominant frequency analysis: Identify frequency sub-frequency components (e.g., 25Hz) or high-frequency components (e.g., 150Hz) that have significantly prominent energy, other than the fundamental frequency; Amplitude relationship analysis: Calculate the peak voltage of each phase, the effective value of U0, and its ratio to the rated value; Waveform distortion analysis: Calculate the total harmonic distortion (THD) and the content of each harmonic.

[0030] S5, Resonance Mechanism Correlation and Comprehensive Judgment Analysis: By fusing multi-source information, the aforementioned analysis results are correlated to determine whether resonance occurs, identify its type, and explain its causal relationship with the grounding form.

[0031] S51, Multi-source information fusion: The grounding form of step S2, the fault type of step S3, and the electrical characteristics of step S4 are combined into a joint feature vector.

[0032] S52, rule-based mechanism inference: invokes a pre-built expert rule base for logical inference.

[0033] For example: IF (Grounding type = grounded via arc suppression coil) AND (Fault type = intermittent arc grounding) AND (Dominant frequency ≈ 25Hz) >THEN (Judgment result = High probability of frequency-division ferroresonance; Mechanism = Arc reignition excites PT core saturation, which resonates with the system's capacitance to ground).

[0034] S53, Model-Based Intelligent Judgment: Employs ensemble learning algorithms such as random forest or gradient boosting tree to train the model to learn the complex mapping relationship between joint feature vectors and resonance type / occurrence, outputs probabilistic judgment results, and cross-validates with rule inference results.

[0035] Thus, the present invention can capture the characteristic frequency components of resonance in the early stage of development through real-time time-frequency analysis (such as wavelet transform) in step S4, and combined with the fusion judgment in step S5, the system can issue a "resonance risk warning" tens to hundreds of milliseconds before the overvoltage amplitude reaches the dangerous threshold, providing operators with a critical decision window for deploying harmonic suppression devices or adjusting the operating mode, turning passive into active.

[0036] S6, Full Event Tracing and Analysis Report Generation: Constructs an event timeline, generates structured analytical conclusions, and traces the entire event.

[0037] S61, Construct Event Timeline: Automatically generate a refined timeline starting from T0, including key nodes such as grounding occurrence, fault type establishment, resonance excitation, protection action, and voltage recovery.

[0038] S62 generates intelligent analysis reports, which automatically include: Event summary: time, location, grounding method, fault type, resonance type; Analysis process: key feature data, spectrogram, and logical chain of judgment criteria; Conclusions and mechanism explanation: Clearly state "what kind of grounding fault, under what system, induced what kind of resonance"; Impact assessment: Analysis of potential or actual impacts on equipment insulation, protection operation, and power quality; Preliminary recommendations: Based on the analysis results, we recommend targeted measures such as "installing a harmonic suppression resistor" or "checking the insulation of a certain section of the line".

[0039] S7, Case Library Iteration and Self-Optimization of Analysis Methods: Construct a case library for incremental learning of the model and dynamically update the rule library, enabling the system to have continuous learning capabilities and improve the accuracy of analysis.

[0040] S71, Case Library Construction: Store each accurate analysis report confirmed manually in the typical case library according to the three-dimensional tags of "grounding type - fault type - resonance type".

[0041] S72, Incremental Model Learning: Periodically use new case data to incrementally train the intelligent decision model in step S5, optimize its parameters, and adapt it to changes in the power grid structure.

[0042] S73, Dynamic Rule Base Updates: Based on newly discovered special cases or misjudgment analysis, the expert rule base is supplemented and revised to keep the analysis logic up-to-date.

[0043] Thus far, the present invention encodes the expert's judgment logic into a "rule base" in step S5, automatically generates a structured report containing waveforms, spectra, and logic chains in step S6, and stores each verified analysis case into the knowledge base in step S7 for the system to continuously learn.

[0044] In summary, this invention integrates grounding pattern identification, fault type diagnosis, and resonance characteristic analysis through time-series correlation and logical fusion, ultimately outputting a complete analysis report that combines phenomenon description and mechanism explanation, thereby achieving accurate identification and in-depth analysis of the causes of resonant overvoltage.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of the claims of the present invention.

Claims

1. A method for identifying and analyzing resonant overvoltage under different types of frequent grounding, characterized in that... It includes the following steps: S1, Synchronous Acquisition and Preprocessing Analysis of Multi-Source Heterogeneous Data: Through data acquisition, synchronization and preprocessing analysis, a high-quality and synchronized data foundation is laid for subsequent analysis; S2, Adaptive Grounding Mode Identification and Analysis: Through feature extraction and pattern matching analysis, the current grounding operation mode of the system is automatically determined in the early stage of a fault. S3, Fine-grained diagnostic analysis of grounding fault types: Through joint analysis of transient and steady-state characteristics, the fault type is diagnosed and determined. Under known grounding conditions, the specific fault attributes that cause resonance risk are diagnosed in depth. S4, Resonant Overvoltage Feature Extraction and Mode Screening: Through dynamic time-frequency analysis and key feature quantity calculation, the spectrum and waveform features that may characterize resonance are extracted from the voltage waveform. S5, Resonance Mechanism Correlation and Comprehensive Judgment Analysis: By fusing multi-source information, the aforementioned analysis results are correlated to determine whether resonance occurs, identify its type, and explain its causal relationship with the grounding form; S6, Full Event Tracing and Analysis Report Generation: Constructs an event timeline, generates structured analytical conclusions, and traces the entire event; S7, Case Library Iteration and Self-Optimization of Analysis Methods: Construct a case library for incremental learning of the model and dynamically update the rule library, enabling the system to have continuous learning capabilities and improve the accuracy of analysis.

2. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S1 includes the following steps: S11, collect three types of data: Electrical transient / steady-state quantities: three-phase voltage, zero-sequence voltage, zero-sequence current of each branch, including lines, arc suppression coils, and capacitor banks, with a sampling rate of not less than 10kHz; System status variables: neutral grounding mode switching status, arc suppression coil position, and related switch position signals; Historical and parameter data: power grid structure parameters, historical fault recording data, equipment nameplate parameters; S12, Synchronization and Preprocessing Analysis: Using BeiDou / GPS to achieve microsecond-level synchronization of network data, and performing filtering, noise reduction, and per-unit preprocessing on the raw data, analyzing its basic statistical characteristics, and determining the ground fault initiation time T0.

3. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S2 includes the following steps: S21, Feature Extraction: Analyze the data within one power frequency cycle before and after T0, and extract key features, including the steady-state amplitude and phase of the neutral point displacement voltage U0, the amplitude ratio and phase difference between the arc suppression coil current IL and the total zero-sequence current IΣ of the system, and the consistency of the capacitive characteristic direction of the zero-sequence current 3I0 of each line. S22, Pattern Matching Analysis: Match the features extracted in step S21 with the pre-set grounding form feature library, and use a clustering algorithm or lightweight classification model to achieve online adaptive identification of grounding forms and output the confidence level.

4. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 3, characterized in that: The grounding type feature library includes ungrounded, grounded via arc suppression coil, and grounded via resistor.

5. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S3 includes the following steps: S31, Joint analysis of transient and steady-state characteristics: Transient analysis: Analyze the polarity of the first half-wave, the high-frequency oscillation frequency, and the decay rate of the zero-sequence current within 5ms after T0; Steady-state waveform analysis: Analyze the stability, periodic fluctuations and distortion patterns of the zero-sequence voltage within hundreds of milliseconds after T0; S32, Diagnosis and Judgment: Construct a multi-dimensional decision tree or support vector machine (SVM) model with transient and steady-state characteristics as input, and diagnose the fault types as follows: metallic grounding, which is stable and has clear characteristics; intermittent arc grounding, which is fluctuating and has high-frequency reignition; and high-resistance grounding, which is low amplitude and has weak characteristics.

6. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S4 includes the following steps: S41, Dynamic Time-Frequency Analysis: Perform continuous wavelet transform (CWT) or short-time Fourier transform (STFT) on the three-phase voltage data after T0 to generate a time-frequency spectrum and observe the evolution of frequency components over time. S42, Calculation of key feature quantities: Dominant frequency analysis: Identify frequency sub-frequency or high-frequency components with significantly prominent energy, other than the fundamental frequency; Amplitude relationship analysis: Calculate the peak voltage of each phase, the effective value of U0, and its ratio to the rated value; Waveform distortion analysis: Calculate the total harmonic distortion (THD) and the content of each harmonic.

7. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S5 includes the following steps: S51, Multi-source information fusion: The grounding form of step S2, the fault type of step S3, and the electrical characteristics of step S4 are combined into a joint feature vector. S52, rule-based mechanism inference: invoke a pre-built expert rule base to perform logical inference; S53, Model-Based Intelligent Judgment: Employs ensemble learning algorithms such as random forest or gradient boosting tree to train the model to learn the complex mapping relationship between joint feature vectors and resonance type / occurrence, outputs probabilistic judgment results, and cross-validates with rule inference results.

8. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S6 includes the following steps: S61, Build an event timeline: Automatically generate a refined timeline starting from T0, building key nodes; S62 generates intelligent analysis reports, which automatically include: Event summary: time, location, grounding method, fault type, resonance type; Analysis process: key feature data, spectrogram, and logical chain of judgment criteria; Conclusions and mechanism explanation: Clearly state "what kind of grounding fault, under what system, induced what kind of resonance"; Impact assessment: Analysis of potential or actual impacts on equipment insulation, protection operation, and power quality; Preliminary recommendations: Based on the analysis findings, we recommend targeted measures.

9. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that: The key nodes include grounding occurrence, fault type determination, resonance excitation, protection action, and voltage recovery.

10. The method for identifying and analyzing resonant overvoltage under different types of frequent grounding as described in claim 1, characterized in that, Step S7 includes the following steps: S71, Case Library Construction: Store each accurate analysis report confirmed manually into the typical case library according to the three-dimensional tags of "grounding type - fault type - resonance type"; S72, Incremental Model Learning: Periodically use new case data to incrementally train the intelligent judgment model in step S5, optimize its parameters, and make it adapt to changes in the power grid structure. S73, Dynamic Rule Base Updates: Based on newly discovered special cases or misjudgment analysis, the expert rule base is supplemented and revised to keep the analysis logic up-to-date.

Images