Method for confirming partial discharge fault of GIS considering accumulation lag of SF6 decomposition products

Abstract

This invention discloses a method for confirming partial discharge faults in GIS (Gas-Insulated Geological Components) considering the cumulative lag of SF6 decomposition products. The method includes: acquiring transient partial discharge signals from the GIS equipment and extracting fast-variable characteristic parameters of the partial discharge; acquiring SF6 decomposition product detection information within the GIS equipment's gas chamber and extracting slow-variable characteristic parameters of the decomposition products; establishing fast-variable and slow-variable time windows based on the difference in response time between the partial discharge signal and decomposition product changes, and establishing a mapping relationship between the two based on the lag time; constructing primary criteria, secondary criteria, and rejection criteria; calculating the partial discharge anomaly degree, decomposition product anomaly degree, and interference penalty term to obtain a comprehensive risk score; and outputting the fault confirmation result based on the comprehensive risk score and the primary, secondary, and rejection criteria. This invention can balance the rapid detection of partial discharge and the delayed confirmation of decomposition products, reducing false alarms and improving the accuracy and engineering practicality of GIS fault confirmation.

Description

Technical Field

[0001] This invention relates to the field of condition monitoring and fault diagnosis technology for gas-insulated metal-enclosed switchgear, and in particular to a method for confirming partial discharge faults in GIS that takes into account the lag of SF6 decomposition product accumulation. Background Technology

[0002] GIS equipment is widely used in power transmission, substation, and distribution systems due to its small footprint, high insulation performance, and good operational reliability. Problems such as floating potential, metal particles, sharp defects, insulation aging, and poor contact within the GIS can typically induce partial discharge. Over time, partial discharge can lead to insulation degradation and electric field distortion, and in severe cases, even insulation breakdown and power outages. Therefore, timely and accurate detection and confirmation of partial discharge faults in GIS is of great importance.

[0003] Existing methods for detecting partial discharge in GIS mainly include ultra-high frequency (300 MHz to 3 GHz) detection, ultrasonic detection, and high frequency (300 kHz to 30 MHz) current detection. These methods can quickly capture transient abnormal signals generated by partial discharge, making them suitable for anomaly detection and online monitoring. However, in engineering sites, these signals are easily affected by external electromagnetic interference, mechanical vibration, changes in propagation paths, and differences in installation location, leading to false alarms, missed alarms, or difficulties in confirming the authenticity of the fault even when an anomaly is detected.

[0004] On the other hand, SF6 gas in GIS decomposes under the influence of partial discharge, electric arc, and overheating, generating decomposition products such as SO2, SOF2, SO2F2, H2S, and HF. Detecting these decomposition products can chemically reflect the presence of abnormal discharge or insulation degradation within the GIS, thus possessing strong value in fault evidence and severity characterization. However, the generation, diffusion, mixing, and sampling detection of SF6 decomposition products typically involve time delays, and the detection results are mostly expressed as cumulative amounts, rates of change, or trends, making it difficult to provide a rapid response to single transient partial discharge events.

[0005] While existing technologies offer solutions combining partial discharge detection and SF6 decomposition product detection, most remain at the level of parallel acquisition of multi-source information, simple weighted fusion, or device-level integration. They lack a tiered confirmation logic addressing the fact that "partial discharge transient signals are fast variables, SF6 decomposition product changes are slow variables, and there is a significant timescale difference and cumulative lag between the two." Particularly in engineering applications, the following scenarios frequently occur: partial discharge pulses are detected, but decomposition products have not yet changed significantly; decomposition products exhibit abnormal trends, but partial discharge pulses only appear intermittently; normal opening and closing operations generate background decomposition products, increasing the risk of misjudgment. Therefore, existing technologies struggle to simultaneously address early anomaly detection, subsequent fault confirmation, and elimination of interference from normal operations.

[0006] In summary, the existing technology has the following problems:

[0007] (1) When relying solely on the transient signal of partial discharge, it is easily affected by noise and interference, making it difficult to confirm the authenticity of the fault;

[0008] (2) When relying solely on SF6 decomposition products for detection, there is a response lag, making it difficult to reflect early partial discharge anomalies in a timely manner;

[0009] (3) Existing joint detection schemes lack confirmation logic for time-series differences between fast and slow variables;

[0010] (4) There is a lack of engineering methods that can output multi-level results such as transient suspected, persistent suspected, partial discharge fault confirmation and serious fault warning.

[0011] Therefore, it is necessary to propose a new method for confirming partial discharge faults in GIS, so that the fast variables of partial discharge and the slow variables of SF6 decomposition products can work together within a unified confirmation framework, thereby improving the accuracy and engineering applicability of GIS fault identification. Summary of the Invention

[0012] The purpose of this invention is to provide a GIS partial discharge fault confirmation method that takes into account the lag in the accumulation of SF6 decomposition products, so as to solve the problems in the prior art. This method can balance the rapid detection of partial discharge and the lag in the confirmation of decomposition products, reduce false alarms, and improve the accuracy and engineering practicality of GIS fault confirmation.

[0013] To achieve the above objectives, the present invention adopts the following technical solution:

[0014] A method for confirming partial discharge faults in GIS that takes into account the hysteresis of SF6 decomposition product accumulation includes:

[0015] Step S1: Obtain the transient signal of partial discharge from the GIS equipment and extract the characteristic parameters of the rapid variable of partial discharge;

[0016] Step S2: Obtain the SF6 decomposition product detection information in the corresponding gas chamber of the GIS equipment, and extract the slow variable characteristic parameters of the decomposition products;

[0017] Step S3: Based on the difference in response time between the partial discharge transient signal and the SF6 decomposition product detection information, establish a fast variable time window and a slow variable time window respectively, and establish the temporal correlation between the two.

[0018] Step S4: Construct a main criterion based on the fast variable characteristic parameters of partial discharge, construct an auxiliary criterion based on the slow variable characteristic parameters of decomposition products, and construct a rejection criterion for excluding non-fault factors;

[0019] Step S5: Calculate the partial discharge anomaly degree based on the partial discharge fast variable characteristic parameters, calculate the decomposition anomaly degree based on the decomposition slow variable characteristic parameters, and calculate the interference penalty term based on the rejection criterion.

[0020] Step S6: Calculate the comprehensive risk score based on the partial discharge anomaly degree, decomposition anomaly degree, and interference penalty term;

[0021] Step S7: Based on the comprehensive risk score and the combined results of the main criteria, auxiliary criteria and rejection criteria, the status of GIS equipment is classified and confirmed, and the classification and confirmation results are output.

[0022] Furthermore, the transient signal of partial discharge includes one or more of ultra-high frequency signals, ultrasonic signals, and high frequency current signals, and the rapid variable characteristic parameters of partial discharge include one or more of pulse amplitude, pulse repetition rate, number of events per unit time, pulse energy, and multi-channel consistency characteristics.

[0023] The SF6 decomposition products include one or more of SO2, SOF2, SO2F2, H2S, and HF. The slow variable characteristic parameters of the decomposition products include one or more of the following: current concentration value, rate of change per unit time, cumulative change over multiple detection cycles, and characteristic component ratio.

[0024] Further, in step S3, the fast variable time window includes an instantaneous window W1 for identifying a single partial discharge anomaly and a short-time window W2 for identifying recurring partial discharge anomalies. The slow variable time window includes a medium-time window W3 for identifying the trend of decomposition product changes and a long-time window W4 for identifying the degree of decomposition product accumulation. Let the time when the partial discharge anomaly is first detected be t0. The instantaneous window W1 is defined as W1=[t0-Δt1, t0+Δt1], the short-time window W2 is defined as W2=[t0, t0+T2], the medium-time window W3 is defined as W3=[t0+τ, t0+τ+T3], and the long-time window W4 is defined as W4=[t0+τ, t0+τ+T4]. Wherein, Δt1 is the single abnormal pulse identification tolerance time, T2 is the continuous review duration, T3 is the decomposition product trend identification duration, T4 is the decomposition product accumulation confirmation duration, and T4 is greater than T3. τ is the decomposition product response lag time.

[0025] Further, the decomposition product response lag time τ = k1·V / Q + k2·P + k3; where V is the equivalent volume of the target gas chamber, Q is the sampling flow rate or equivalent gas exchange flow rate, P is the gas chamber operating pressure or diffusion influence parameter, and k1, k2, and k3 are empirical coefficients.

[0026] Further, in step S5, the partial discharge anomaly degree FPD = w1·An + w2·Rn + w3·Nn + w4·En + w5·Kn; where An, Rn, Nn, En, and Kn are the normalized values ​​of pulse amplitude, pulse repetition rate, number of events per unit time, pulse energy, and multi-channel consistency characteristics, respectively, and w1, w2, w3, w4, and w5 are the corresponding weighting coefficients, and w1+w2+w3+w4+w5=1;

[0027] The anomaly degree of the decomposition product FGas = v1·Cn + v2·Vn + v3·Mn + v4·Qn; where Cn, Vn, Mn, and Qn are the normalized values ​​of the concentration, rate of change, cumulative increment, and ratio of characteristic components of the decomposition product, respectively, and v1, v2, v3, and v4 are the corresponding weighting coefficients, and v1+v2+v3+v4=1;

[0028] The interference penalty term FInt = p1·O + p2·M + p3·S + p4·U; where O represents the normal opening and closing or normal interruption operation mark, M represents the background mark of residual decomposition products after maintenance, S represents the mark of partial discharge abnormality that only occurred once and did not recur, U represents the mark of external electromagnetic interference or mechanical noise, and p1, p2, p3, and p4 are penalty coefficients.

[0029] Furthermore, the weighting coefficients of the partial discharge anomaly degree FPD, the decomposition anomaly degree FGas, and the interference penalty term FInt are calculated using the improved CRITIC method, specifically including:

[0030] S51. Normalize the feature data corresponding to all weight coefficients;

[0031] S52. Calculate the volatility measure V_coff for each characteristic quantity. The volatility measure for the j-th characteristic quantity is: , where σj is the normalized standard deviation of the j-th feature and μj is its mean;

[0032] S53. Calculate the Pearson correlation coefficient between each characteristic quantity and construct the correlation matrix R;

[0033] S54. Calculate the average incoherence among the features as the conflict measure C_coff, and the conflict measure for the j-th feature is... , where n is the total number of features, r_jk is an element in the correlation matrix R, representing the Pearson correlation coefficient between the j-th feature and the k-th feature;

[0034] S55. Calculate the independence measure D_coff for each characteristic quantity. ,in This represents the proportion of variation that an indicator can explain using other indicators. It is obtained by performing multiple linear regression with one indicator as the dependent variable and the other indicators as independent variables. ;

[0035] S56. Take the geometric mean to obtain the comprehensive information content T_coff Then, the weights corresponding to each feature are obtained by normalization.

[0036] Further, in step S6, the comprehensive risk score R = α(t)·FPD + β(t)·FGas - γ·FInt; where FPD, FGas, and FInt are the partial discharge anomaly degree, the decomposition anomaly degree, and the interference penalty term, respectively, and α(t), β(t), and γ are weighting coefficients. The weighting coefficients α(t) and β(t) are dynamic weighting coefficients that satisfy: α(t) = α0·e^(-t / τ); β(t) = β0·(1-e^(-t / τ)); where t is the time difference from the first detection time t0 of the partial discharge anomaly, τ is the decomposition response lag time, and α0 and β0 are the set initial weighting coefficients.

[0037] Furthermore, it also includes: setting a preset protection period To for normal opening and closing or normal interruption operations; when the GIS equipment has a normal operation record within To, correcting the anomaly degree of the decomposed material: FGas' = λ·FGas; where 0<λ<1, FGas is the anomaly degree of the decomposed material; or correcting the comprehensive risk score: R' = R-δ; where δ is the operation interference correction amount, and R is the comprehensive risk score.

[0038] Further, step S7 specifically includes: when the primary criterion meets the initial screening condition for anomalies and the comprehensive risk score R is less than the first grading threshold R1, output transient suspected condition; when the primary criterion continuously meets the condition within multiple consecutive fast variable time windows, and the first grading threshold R1 is not greater than the comprehensive risk score R and R is less than the second grading threshold R2, output persistent suspected condition; when the primary criterion and auxiliary criterion simultaneously meet the fault confirmation condition, the rejection criterion is not established, and the second grading threshold R2 is not greater than the comprehensive risk score R and R is less than the third grading threshold R3, output partial discharge fault confirmation; when the primary criterion indicates abnormal enhancement of partial discharge, the auxiliary criterion indicates abnormal aggravation of decomposition products, the rejection criterion is not established, and the comprehensive risk score R is not less than the third grading threshold R3, output severe fault warning; wherein, R1 <R2<R3。

[0039] A GIS partial discharge fault confirmation system considering the hysteresis of SF6 decomposition product accumulation includes:

[0040] Partial discharge signal acquisition module, used to acquire transient partial discharge signals from GIS equipment;

[0041] The SF6 decomposition product sampling and analysis module is used to obtain SF6 decomposition product detection information in the gas chamber of GIS equipment;

[0042] The data processing module is used to extract fast variable characteristic parameters of partial discharge and slow variable characteristic parameters of decomposition products. Based on the difference in response time between the transient signal of partial discharge and the detection information of SF6 decomposition products, fast variable time windows and slow variable time windows are established respectively, and the temporal correlation between the two is established. A main criterion is constructed based on the fast variable characteristic parameters of partial discharge, an auxiliary criterion is constructed based on the slow variable characteristic parameters of decomposition products, and a rejection criterion for excluding non-fault factors is constructed. The anomaly degree of partial discharge is calculated based on the fast variable characteristic parameters of partial discharge, the anomaly degree of decomposition products is calculated based on the slow variable characteristic parameters of decomposition products, an interference penalty item is calculated based on the rejection criterion, and a comprehensive risk score is calculated based on the anomaly degree of partial discharge, the anomaly degree of decomposition products, and the interference penalty item.

[0043] The output module, based on the comprehensive risk score and the combined results of the primary criteria, secondary criteria, and rejection criteria, is used to classify and confirm the status of GIS equipment and output the classification and confirmation results.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] (1) Defining the partial discharge signal and the decomposition information as fast variables and slow variables respectively, and fully considering the difference in their response time, is more in line with the fault evolution mechanism of GIS;

[0046] (2) Adopt the process of "fast variable anomaly detection - continuous review - slow variable corroboration - rejection item elimination - hierarchical output" to improve the accuracy of fault confirmation;

[0047] (3) It can retain transient suspected and persistent suspicious states when the decomposition products have not changed significantly, thus taking into account both early warning and subsequent confirmation;

[0048] (4) Introduce rejection criteria such as normal operation gas production, maintenance residue and external interference to reduce false alarm rate;

[0049] (5) Output multi-level status results to facilitate maintenance personnel to formulate retesting, tracking, repair or emergency response plans according to risk level. Attached Figure Description

[0050] Figure 1 This is a flowchart of the overall process of the method of the present invention.

[0051] Figure 2 This is a schematic diagram of the hierarchical confirmation logic consisting of the main criterion, the auxiliary criterion, and the rejection criterion in this invention.

[0052] Figure 3 This is a schematic diagram illustrating the temporal correlation between the fast variable time window and the slow variable time window in this invention. Detailed Implementation

[0053] The present invention will be further described below with reference to the accompanying drawings and embodiments, but the present invention is not limited to the following embodiments.

[0054] Combination Figures 1-3 This embodiment provides a method for confirming partial discharge faults in GIS that takes into account the cumulative hysteresis of SF6 decomposition products, including the following steps:

[0055] S1. Obtain partial discharge fast variable information

[0056] Partial discharge transient signals are acquired by a partial discharge detection unit installed on the GIS equipment. These signals can be one or more of ultra-high frequency signals (electromagnetic wave signals), ultrasonic signals, and high-frequency current signals. The acquired signals are then filtered, denoised, pulse-identified, and feature-extracted to obtain fast variable characteristic parameters.

[0057] The fast variable characteristic parameters include at least one or more of the following: pulse amplitude An, pulse repetition rate Rn, number of events per unit time Nn, pulse energy En, and multi-channel consistency coefficient Kn. The sampling frequency, detection bandwidth, gain, and trigger threshold of the partial discharge detection unit can be preset according to the voltage level of the GIS equipment, sensor type, installation location, and on-site electromagnetic environment, or adaptively adjusted based on historical monitoring data. The fast variable characteristic parameters can be extracted using a fixed-length window, a sliding window, or an event-triggered window.

[0058] S2. Obtain information on slow variables of SF6 decomposition products.

[0059] The gas sampling and analysis unit acquires SF6 decomposition product detection information from the corresponding gas chamber in the GIS. The decomposition products may include one or more of SO2, SOF2, SO2F2, H2S, and HF. Baseline correction, trend analysis, and feature extraction are performed on the detection results to obtain slow variable characteristic parameters.

[0060] The slow variable characteristic parameters include at least one or more of the following: current concentration Cn, rate of change per unit time Vn, cumulative change over multiple detection cycles Mn, and characteristic component ratio Qn. The baseline value can be determined based on data from the initial commissioning phase of the GIS equipment, normal operation phase, historical health sample data, or data from the stable phase after maintenance. The detection cycle for the slow variable characteristic parameters can be set according to the sampling method, detection sensitivity, target gas chamber volume, and target decomposition product type. For implementations involving parallel detection of multiple decomposition products, the characteristic parameters of each decomposition product can be extracted separately and then combined for calculation.

[0061] The characteristic component ratio Qg includes one or more of SO2F2 / SOF2, SO2 / SOF2, or other decomposition product combination ratios used to characterize the fault evolution state, and the characteristic component ratio Qg is normalized to obtain Qn.

[0062] S3. Establish time windows for fast variables and slow variables.

[0063] Based on the response differences between transient anomalies in partial discharge and changes in decomposition products, a fast variable time window for fast variable analysis and a slow variable time window for slow variable analysis are established.

[0064] Let t0 be the time when the partial discharge anomaly is first detected. Then: the instantaneous window W1 is defined as W1=[t0-Δt1, t0+Δt1]; the short-term window W2 is defined as W2=[t0, t0+T2]; the medium-term window W3 is defined as W3=[t0+τ, t0+τ+T3]; and the long-term window W4 is defined as W4=[t0+τ, t0+τ+T4]. Where Δt1 is the single abnormal pulse identification tolerance time, T2 is the continuous verification time, T3 is the decomposition trend identification time, T4 is the decomposition cumulative confirmation time (where T4 is greater than T3), and τ is the decomposition response lag time.

[0065] The response lag time τ of the decomposed product is determined according to the following empirical expression: τ = k1·V / Q + k2·P + k3; where V is the equivalent volume of the target gas chamber, Q is the sampling flow rate or equivalent gas exchange flow rate, P is the gas chamber operating pressure or diffusion influence parameter, and k1, k2, and k3 are empirical coefficients.

[0066] The parameters Δt1, T2, T3, and T4 can be determined based on the partial discharge detection sampling period, the decomposition product sampling period, the target gas chamber structural parameters, historical monitoring data statistics, or experimental calibration results, and can be preset or adaptively adjusted according to the equipment operating conditions. The empirical coefficients k1, k2, and k3 can be determined based on different GIS models, gas chamber structures, sampling path lengths, decomposition product diffusion test results, or historical sample regression analysis results.

[0067] The τ is preferably a positive value; when the τ calculated according to the above empirical expression is less than the preset minimum lag time, the preset minimum lag time can be taken to avoid the slow variable time window starting too early.

[0068] In another implementation, τ can also be determined using a fixed empirical value, a lookup table method, or a historical statistical method.

[0069] S4. Construct the primary criteria, secondary criteria, and rejection criteria.

[0070] The primary criterion is the fast variable of partial discharge, the secondary criterion is the slow variable of SF6 decomposition products, and the rejection criterion is the interference elimination information.

[0071] The main criteria may include: whether the fast variable exceeds the first abnormality threshold; whether the partial discharge abnormality recurs within multiple consecutive fast variable time windows; whether the partial discharge abnormality shows an increasing trend; and whether corresponding abnormalities occur under two or more partial discharge detection methods.

[0072] Auxiliary criteria may include: whether the concentration of the target decomposition product exceeds the second abnormal threshold; whether the rate of change of the target decomposition product exceeds the third abnormal threshold; whether the cumulative increase of the target decomposition product exceeds the confirmation threshold; whether the ratio of characteristic components meets the preset fault conditions; and whether the decomposition product continues to increase over multiple consecutive detection cycles.

[0073] The veto criteria may include: the GIS equipment has normal opening and closing or switching operation records within a preset time; the GIS equipment has residual decomposition materials after maintenance; the partial discharge abnormality is only a single event and does not recur within the subsequent fast variable time window; external electromagnetic interference causes transient pulse false triggering; mechanical vibration or environmental noise causes ultrasonic abnormalities.

[0074] The data for the veto criteria can be obtained through at least one of the following methods: GIS station control system, SCADA system, monitoring backend, circuit breaker auxiliary contacts, disconnector switch position contacts, operating mechanism status input, maintenance log system, manual input module, environmental monitoring module, vibration sensor or noise detection unit.

[0075] The first, second, and third anomaly thresholds, and the confirmation threshold can be preset and determined based on equipment type, operating voltage level, historical health samples, historical fault samples, on-site test results, or operation and maintenance experience, and can be adaptively corrected based on long-term operating baselines. The main criterion, auxiliary criterion, and rejection criterion can be implemented using a combination of Boolean logic, weighted logic, or rule tables.

[0076] S5. Calculate the partial discharge anomaly degree, decomposition anomaly degree, and interference penalty term.

[0077] The partial discharge anomaly degree (FPD) is defined as: FPD = w1·An + w2·Rn + w3·Nn + w4·En + w5·Kn; where An, Rn, Nn, En, and Kn are the normalized values ​​of pulse amplitude, pulse repetition rate, number of events per unit time, pulse energy, and multi-channel consistency characteristics, respectively; w1, w2, w3, w4, and w5 are the corresponding weighting coefficients, and w1+w2+w3+w4+w5=1. When only a single partial discharge signal channel is used, Kn can be omitted.

[0078] The preferred values ​​for An, Rn, Nn, En, and Kn are [0,1]. The values ​​for w1, w2, w3, w4, and w5 can be determined using an improved CRITIC method based on the differences in the ability of different partial discharge characteristics to characterize the anomalies. When Kn is omitted, the remaining weighting coefficients are preferably renormalized so that the sum of all weighting coefficients involved in the calculation is still 1.

[0079] The anomaly degree FGas of the decomposition product is defined as: FGas = v1·Cn + v2·Vn + v3·Mn + v4·Qn; where Cn, Vn, Mn, and Qn are the normalized values ​​of the concentration, rate of change, cumulative increment, and ratio of characteristic components of the decomposition product, respectively, and v1, v2, v3, and v4 are the corresponding weighting coefficients, and v1+v2+v3+v4=1.

[0080] The preferred values ​​for Cn, Vn, Mn, and Qn are [0,1]. v1, v2, v3, and v4 can be determined using an improved CRITIC method based on the contribution of different decomposition product characteristics to fault confirmation. When a certain decomposition product characteristic is temporarily unavailable, FGas can be calculated by renormalizing the remaining weight coefficients used in the calculation.

[0081] The interference penalty term FInt is defined as: FInt = p1·O + p2·M + p3·S + p4·U; where O represents the normal opening and closing operation mark, M represents the background mark of residual decomposition products after maintenance, S represents the mark of partial discharge abnormality occurring only once and not recurring, U represents the mark of external electromagnetic interference or mechanical noise, and p1, p2, p3, and p4 are penalty coefficients. Preferably, O, M, S, and U are 0 or 1; they can also be graded as 0, 0.5, and 1.

[0082] The values ​​p1, p2, p3, and p4 can be determined using an improved CRITIC method based on the degree of influence of different interference factors on the risk of false alarms. Preferably, p1 + p2 + p3 + p4 = 1, so that the interference penalty term FInt remains within a preset range; or, the calculated FInt is subjected to amplitude limiting processing, so that its value range is limited to [0,1]. The values ​​O, M, S, and U can be simultaneously true or partially true; when multiple interference factors exist simultaneously, FInt can be calculated based on their superimposed effects.

[0083] The steps for calculating the weighting coefficients of the partial discharge anomaly degree FPD, the decomposition anomaly degree FGas, and the interference penalty term FInt using the improved CRITIC method include:

[0084] S51. Normalize the original feature data;

[0085] Normalization method 1: For the j-th feature, Xj = (X-Xmin) / (Xmax-Xmin);

[0086] Normalization method 2: Any feature quantity X is normalized using a segmented threshold method: when X≤X0, the normalized value Xn=0; when X0<X<X1, Xn=(X-X0) / (X1-X0); when X≥X1, Xn=1; where X0 is the initial abnormal threshold and X1 is the obvious abnormal threshold.

[0087] Xmin and Xmax can be determined based on historical normal sample ranges, test calibration ranges, equipment factory parameters, or long-term operating baselines; X0 and X1 can be determined based on initial abnormality thresholds and obvious abnormality thresholds. Preferably, when linear normalization is used, if the calculated Xn is less than 0, it is set to 0; if the calculated Xn is greater than 1, it is set to 1.

[0088] S52. Calculate the volatility measure V_coff for each characteristic quantity. The volatility measure for the j-th characteristic quantity is: , where σj is

[0089] The normalized standard deviation of the j-th feature is μj, which is its mean.

[0090] S53. Calculate the Pearson correlation coefficient between each feature quantity and construct the correlation matrix R; where r_jk is the correlation coefficient between the j-th feature quantity and the k-th feature quantity;

[0091] S54. Calculate the average incoherence among the features as the conflict measure C_coff, and the conflict measure for the j-th feature is... , where n is the total number of features;

[0092] The larger this value, the weaker the average correlation between indicator j and other indicators, and the stronger the information conflict.

[0093] S55. Calculate the independence measure D_coff for each characteristic quantity. Using one indicator as the dependent variable and the other indicators as independent variables, a multiple linear regression was performed to obtain the multiple correlation coefficient. ;

[0094] in The variance of an indicator can be explained by other indicators, so Dj represents the proportion of unique information that it cannot explain, i.e., independence.

[0095] S56. Take the geometric mean to obtain the comprehensive information content T_coff The weights w, v, and p are obtained by normalization.

[0096] S6. Calculate the comprehensive risk score.

[0097] The comprehensive risk score R is defined as: R = α(t)·FPD + β(t)·FGas - γ·FInt; where α(t), β(t), and γ are weighting coefficients. Preferably, α(t) and β(t) are dynamic weights and satisfy: α(t) = α0·e^(-t / τ); β(t) = β0·(1-e^(-t / τ)); preferably, γ is a fixed weighting coefficient, obtained by historical sample calibration or empirical setting; where t is the time difference from the first detection time t0 of the partial discharge anomaly, τ is the response lag time of the decomposition products, and α0 and β0 are the initial weights.

[0098] In the formula, e^(-t / τ) is used to construct a weighting function that decreases over time, ensuring that fast partial discharge variables have a larger proportion in the early stages of anomalies and gradually decrease over time. In the formula, 1-e^(-t / τ) is used to construct a weighting function that increases over time. This dynamic weighting method reflects that fast partial discharge variables dominate in the early stages, while the weight of slow decomposition variable variables gradually increases over time. When t is small, α(t) is large and β(t) is small, making the overall risk score dominated by fast partial discharge variables. As t increases, α(t) gradually decreases while β(t) gradually increases, gradually increasing the weight of slow decomposition variable variables in the overall risk score, thus reflecting the cumulative lag characteristic of SF6 decomposition products relative to the transient signal of partial discharge. This dynamic weighting setting allows the present invention to balance the early detection of partial discharge anomalies with the subsequent confirmation of decomposition product evidence.

[0099] The preferred values ​​for FPD, FGas, and FInt are [0,1]. α0, β0, and γ can be determined based on historical fault samples, historical normal samples, experimental calibration results, expert experience, or model training results. α0 and β0 satisfy the total positive weight constraint to maintain the stability of the comprehensive risk score scale; in a preferred embodiment, α0 = β0 = w, where w is the preset total positive weight, thus α(t) + β(t) = w. Preferably, γ ranges from 0 to 1; when FInt has been normalized, γ is used to characterize the deduction intensity of interference terms on the comprehensive risk score. In another embodiment, the calculated R can be amplitude-limited or re-normalized and then compared with the grading threshold; preferably, when R is less than a preset lower limit, the lower limit is taken, and when R is greater than a preset upper limit, the upper limit is taken.

[0100] S7. Output the hierarchical confirmation results

[0101] Based on the comprehensive risk score R and the combined results of the primary criterion, secondary criterion, and veto criterion, the GIS status is divided into four levels: Level 1, transient suspected; Level 2, persistent suspected; Level 3, partial discharge fault confirmed; and Level 4, serious fault warning.

[0102] Specifically: when the primary criterion meets the initial screening condition for anomalies and R < R1, output transient suspected; when the primary criterion continuously meets the condition within multiple consecutive fast variable time windows and R1 ≤ R < R2, output persistent suspected; when the primary criterion and the auxiliary criterion simultaneously meet the fault confirmation condition, the rejection criterion is not valid, and R2 ≤ R < R3, output partial discharge fault confirmation; when the primary criterion indicates an abnormal increase in partial discharge and the auxiliary criterion indicates an abnormal increase in decomposition products, the rejection criterion is not valid, and R ≥ R3, output a serious fault warning; where R1 < R2 < R3.

[0103] When a fast variable anomaly occurs but a slow variable does not meet the fault confirmation criteria, the transient suspected or persistent suspicious state is maintained; when a slow variable meets the fault confirmation criteria within the corresponding slow variable time window, the state is upgraded to partial discharge fault confirmation or serious fault warning.

[0104] R1, R2, and R3 can be preset based on historical fault samples, historical normal samples, field test data, risk tolerance requirements, or operation and maintenance strategies, and can be adaptively corrected according to equipment type, operating conditions, and historical baselines. Preferably, R1, R2, and R3 use the same dimensions and the same range as the comprehensive risk score R.

[0105] The method of the present invention further includes: setting a preset protection period To for normal opening and closing or normal interruption operations. When the GIS equipment has a normal operation record in To, the anomaly degree of the decomposed material is corrected: FGas' = λ·FGas; where 0<λ<1; or the comprehensive risk score is corrected: R' = R-δ; where δ is the operation interference correction amount.

[0106] The preset protection period To can be determined based on the type of GIS equipment, the recovery pattern of decomposition products after normal operation, historical sample statistics, or experimental calibration results. λ and δ can be determined empirically, through historical sample statistics, or experimental calibration, based on the degree of influence of normal operation on the background rise of decomposition products. In different implementations, it is possible to choose to correct the anomaly degree of decomposition products, or to correct the comprehensive risk score, or to use a combination of both correction methods.

[0107] This embodiment also provides a GIS partial discharge fault confirmation system that considers the lag in the accumulation of SF6 decomposition products, including:

[0108] Partial discharge signal acquisition module, used to acquire transient partial discharge signals from GIS equipment;

[0109] The SF6 decomposition product sampling and analysis module is used to obtain SF6 decomposition product detection information in the gas chamber of GIS equipment;

[0110] The data processing module is used to extract fast variable characteristic parameters of partial discharge and slow variable characteristic parameters of decomposition products. Based on the difference in response time between the transient signal of partial discharge and the detection information of SF6 decomposition products, fast variable time windows and slow variable time windows are established respectively, and the temporal correlation between the two is established. A main criterion is constructed based on the fast variable characteristic parameters of partial discharge, an auxiliary criterion is constructed based on the slow variable characteristic parameters of decomposition products, and a rejection criterion for excluding non-fault factors is constructed. The anomaly degree of partial discharge is calculated based on the fast variable characteristic parameters of partial discharge, the anomaly degree of decomposition products is calculated based on the slow variable characteristic parameters of decomposition products, an interference penalty item is calculated based on the rejection criterion, and a comprehensive risk score is calculated based on the anomaly degree of partial discharge, the anomaly degree of decomposition products, and the interference penalty item.

[0111] The output module, based on the comprehensive risk score and the combined results of the primary criteria, secondary criteria, and rejection criteria, is used to classify and confirm the status of GIS equipment and output the classification and confirmation results.

[0112] Example 1

[0113] An example considering gas generation interference during normal opening and closing: In response to the possibility of background elevation of decomposition products after normal opening and closing operations of GIS circuit breakers or related gas chambers, the system connects to the equipment operation record interface to form the normal operation mark O in the rejection criteria.

[0114] When the system detects an abnormal partial discharge variable accompanied by an increase in decomposition products, it first checks whether the GIS equipment has a normal operation record within the preset protection period To. If a normal operation record exists, it sets O=1 and increases the interference penalty term FInt according to the formula FInt= p1·O + p2·M + p3·S + p4·U; or it corrects the anomaly degree of decomposition products according to the correction rule, i.e., sets FGas' = λ·FGas, where 0<λ<1; or it corrects the comprehensive risk score, i.e., sets R' = R-δ.

[0115] In this scenario, the system focuses on observing whether the rapid variable anomalies within multiple W2 values ​​persist. If no further significant partial discharge pulse anomalies occur, and the corresponding decomposition products gradually decrease within W3 or W4, the current status is only output as a lower-level status, and an insulation fault is not confirmed. If rapid variable anomalies continue to occur, and Cn, Vn, or Mn continue to increase, a partial discharge fault confirmation or a severe fault warning can still be output.

[0116] Example 2

[0117] Fault confirmation based on high-frequency current and SF6 decomposition products: A high-frequency current sensor is installed in the GIS grounding branch or a suitable location to acquire high-frequency current pulses as a fast variable signal of partial discharge. After feature extraction of the high-frequency current pulses, one or more normalized values ​​of An, Rn, Nn, and En are obtained, and the partial discharge anomaly degree (FPD) is calculated accordingly.

[0118] If the high-frequency current anomaly is frequently triggered in multiple consecutive W2 cells, and Nn and En show an increasing trend, the system determines it to be a persistent suspicious condition. If at least one of Cn, Vn, and Mn in the corresponding gas cell (H2S, SO2, or other target decomposition products) continuously increases in W3 or W4 and reaches the preset confirmation condition, the system determines it to be a confirmed partial discharge fault. If the high-frequency current anomaly significantly increases and the decomposition products accumulate rapidly, and the corresponding FGas and comprehensive risk score R both further increase, a serious fault warning is output.

[0119] Example 3

[0120] Graded confirmation based on comprehensive risk scoring: Let t0 be the time when a partial discharge anomaly is first detected. Based on the equivalent volume V of the target gas chamber, the sampling flow rate Q, and the operating pressure or diffusion influence parameter P, the decomposition product response lag time τ is calculated using the formula τ = k1·V / Q + k2·P + k3, and W1, W2, W3, and W4 are established accordingly. Within the short-term window W2, the normalized values ​​An, Rn, Nn, En, and Kn of the fast variable characteristics are extracted, and the partial discharge anomaly degree FPD is calculated using the formula FPD = w1·An + w2·Rn + w3·Nn + w4·En + w5·Kn. Within the medium-term window W3 and the long-term window W4, the normalized values ​​Cn, Vn, Mn, and Qn of the decomposition product concentration, rate of change, cumulative increment, and component ratio are extracted, and the decomposition product anomaly degree FGas is calculated using the formula FGas = v1·Cn + v2·Vn + v3·Mn + v4·Qn.

[0121] Combining normal operation records, maintenance records, and environmental interference information, the values ​​of O, M, S, and U are determined, and the interference penalty term FInt is calculated according to the formula FInt = p1·O + p2·M + p3·S + p4·U. Subsequently, the comprehensive risk score R is calculated according to the formula R = α(t)·FPD + β(t)·FGas - γ·FInt, where α(t) = α0·e^(-t / τ) and β(t) = β0·(1-e^(-t / τ)). In the early stage of anomaly, α(t) is relatively large and β(t) is relatively small, making the comprehensive risk score dominated by the fast variable of partial discharge; as time goes by, α(t) gradually decreases while β(t) gradually increases, so that the weight of the slow variable of decomposition products in the comprehensive risk score gradually increases.

[0122] When the calculated R satisfies R1≤R<R2 and the main criterion is continuously satisfied, the system outputs a persistent suspicion; when R2≤R<R3 and the auxiliary criterion satisfies the fault confirmation condition and the rejection criterion is not established, the system outputs a partial discharge fault confirmation; when R≥R3 and the main criterion indicates an abnormal increase in partial discharge and the auxiliary criterion indicates an abnormal increase in decomposition products, the system outputs a serious fault warning.

[0123] In this specification, "fast variables" refer to anomalous features that respond to the transient process of partial discharge and are observable over a short timescale; "slow variables" refer to anomalous features that respond to the consequences of partial discharge, primarily characterized by the generation and accumulation of decomposition products, and are observable over a longer timescale. The "time window" is not limited to a fixed-length window; it can also be a sliding window, an adaptive window, or a segmented statistical window. The "threshold" is not limited to a fixed constant threshold; it can also be adaptively set based on equipment type, historical baseline, operating conditions, or environmental conditions.

[0124] Any equivalent substitutions or conventional transformations made by those skilled in the art to the partial discharge signal type, decomposition product type, criterion form, window length, model form, and output level name without departing from the spirit and substance of this invention shall fall within the protection scope of this invention.

Claims

1. A method for confirming partial discharge faults in GIS considering the cumulative hysteresis of SF6 decomposition products, characterized in that, include: Step S1: Obtain the transient signal of partial discharge from the GIS equipment and extract the characteristic parameters of the rapid variable of partial discharge; Step S2: Obtain the SF6 decomposition product detection information in the corresponding gas chamber of the GIS equipment, and extract the slow variable characteristic parameters of the decomposition products; Step S3: Based on the difference in response time between the partial discharge transient signal and the SF6 decomposition product detection information, establish a fast variable time window and a slow variable time window respectively, and establish the temporal correlation between the two. Step S4: Construct a main criterion based on the fast variable characteristic parameters of partial discharge, construct an auxiliary criterion based on the slow variable characteristic parameters of decomposition products, and construct a rejection criterion for excluding non-fault factors; Step S5: Calculate the partial discharge anomaly degree based on the partial discharge fast variable characteristic parameters, calculate the decomposition anomaly degree based on the decomposition slow variable characteristic parameters, and calculate the interference penalty term based on the rejection criterion. Step S6: Calculate the comprehensive risk score based on the partial discharge anomaly degree, decomposition anomaly degree, and interference penalty term; Step S7: Based on the comprehensive risk score and the combined results of the main criteria, auxiliary criteria and rejection criteria, the status of GIS equipment is classified and confirmed, and the classification and confirmation results are output.

2. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, The transient signal of partial discharge includes one or more of ultra-high frequency signals, ultrasonic signals and high frequency current signals, and the fast variable characteristic parameters of partial discharge include one or more of pulse amplitude, pulse repetition rate, number of events per unit time, pulse energy and multi-channel consistency characteristics. The SF6 decomposition products include one or more of SO2, SOF2, SO2F2, H2S, and HF. The slow variable characteristic parameters of the decomposition products include one or more of the following: current concentration value, rate of change per unit time, cumulative change over multiple detection cycles, and characteristic component ratio.

3. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, In step S3, the fast variable time window includes an instantaneous window W1 for identifying single partial discharge anomalies and a short window W2 for identifying recurring partial discharge anomalies. The slow variable time window includes a medium-time window W3 for identifying the trend of decomposition product changes and a long-time window W4 for identifying the degree of decomposition product accumulation. Let the time when the partial discharge anomaly is first detected be t0. The instantaneous window W1 is defined as W1=[t0-Δt1, t0+Δt1], the short window W2 is defined as W2=[t0, t0+T2], the medium-time window W3 is defined as W3=[t0+τ, t0+τ+T3], and the long-time window W4 is defined as W4=[t0+τ, t0+τ+T4]. Wherein, Δt1 is the single abnormal pulse identification tolerance time, T2 is the continuous verification time, T3 is the decomposition product trend identification time, T4 is the decomposition product accumulation confirmation time, and T4 is greater than T3. τ is the decomposition product response lag time.

4. The GIS partial discharge fault confirmation method according to claim 3, characterized in that, The decomposition product response lag time τ = k1·V / Q + k2·P + k3; where V is the equivalent volume of the target gas chamber, Q is the sampling flow rate or equivalent gas exchange flow rate, P is the gas chamber operating pressure or diffusion influence parameter, and k1, k2, and k3 are empirical coefficients.

5. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, The partial discharge anomaly degree FPD mentioned in step S5 is: FPD = w1·An + w2·Rn + w3·Nn + w4·En + w5·Kn; where An, Rn, Nn, En, and Kn are the normalized values ​​of pulse amplitude, pulse repetition rate, number of events per unit time, pulse energy, and multi-channel consistency characteristics, respectively; w1, w2, w3, w4, and w5 are the corresponding weighting coefficients, and w1+w2+w3+w4+w5=1. The anomaly degree of the decomposition product FGas = v1·Cn + v2·Vn + v3·Mn + v4·Qn; where Cn, Vn, Mn, and Qn are the normalized values ​​of the concentration, rate of change, cumulative increment, and ratio of characteristic components of the decomposition product, respectively, and v1, v2, v3, and v4 are the corresponding weighting coefficients, and v1+v2+v3+v4=1; The interference penalty term FInt = p1·O + p2·M + p3·S + p4·U; where O represents the normal opening and closing or normal interruption operation mark, M represents the background mark of residual decomposition products after maintenance, S represents the mark of partial discharge abnormality that only occurred once and did not recur, U represents the mark of external electromagnetic interference or mechanical noise, and p1, p2, p3, and p4 are penalty coefficients.

6. The GIS partial discharge fault confirmation method according to claim 5, characterized in that, The weighting coefficients of the partial discharge anomaly degree FPD, the decomposition anomaly degree FGas, and the interference penalty term FInt are calculated using the improved CRITIC method, specifically including: S51. Normalize the feature data corresponding to all weight coefficients; S52. Calculate the volatility measure V_coff for each characteristic quantity. The volatility measure for the j-th characteristic quantity is: , where σj is the normalized standard deviation of the j-th feature and μj is its mean; S53. Calculate the Pearson correlation coefficient between each characteristic quantity and construct the correlation matrix R; S54. Calculate the average incoherence among the features as the conflict measure C_coff, and the conflict measure for the j-th feature is... , where n is the total number of features, r_jk is an element in the correlation matrix R, representing the Pearson correlation coefficient between the j-th feature and the k-th feature; S55. Calculate the independence measure D_coff for each characteristic quantity. ,in This represents the proportion of variation that an indicator can explain using other indicators. It is obtained by performing multiple linear regression with one indicator as the dependent variable and the other indicators as independent variables. ; S56. Take the geometric mean to obtain the comprehensive information content T_coff Then, the weights corresponding to each feature are obtained by normalization.

7. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, In step S6, the comprehensive risk score R = α(t)·FPD + β(t)·FGas - γ·FInt; where FPD, FGas, and FInt are the partial discharge anomaly degree, the decomposition anomaly degree, and the interference penalty term, respectively; α(t), β(t), and γ are weighting coefficients, and the weighting coefficients α(t) and β(t) are dynamic weighting coefficients that satisfy: α(t) = α0·e^(-t / τ); β(t) = β0·(1-e^(-t / τ)); where t is the time difference from the first detection time t0 of the partial discharge anomaly, τ is the decomposition response lag time, and α0 and β0 are the set initial weighting coefficients.

8. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, Also includes: For normal opening and closing or normal disconnection operations, a preset protection period To is set. When the GIS equipment has a normal operation record in To, the anomaly degree of the decomposition product is corrected: FGas' = λ·FGas; where 0<λ<1, FGas is the anomaly degree of the decomposition product; or the comprehensive risk score is corrected: R' = R-δ; where δ is the operation interference correction amount, and R is the comprehensive risk score.

9. The GIS partial discharge fault confirmation method according to claim 1, characterized in that, Step S7 specifically includes: when the primary criterion meets the initial screening condition for anomalies and the comprehensive risk score R is less than the first grading threshold R1, output transient suspected condition; when the primary criterion continuously meets the condition within multiple consecutive fast variable time windows, and the first grading threshold R1 is not greater than the comprehensive risk score R and R is less than the second grading threshold R2, output persistent suspected condition; when the primary criterion and auxiliary criterion simultaneously meet the fault confirmation condition, the rejection criterion is not established, and the second grading threshold R2 is not greater than the comprehensive risk score R and R is less than the third grading threshold R3, output partial discharge fault confirmation; when the primary criterion indicates an abnormal increase in partial discharge, the auxiliary criterion indicates an abnormal increase in decomposition products, the rejection criterion is not established, and the comprehensive risk score R is not less than the third grading threshold R3, output a serious fault warning; wherein, R1 <R2<R3。 10. A GIS partial discharge fault confirmation system considering the cumulative hysteresis of SF6 decomposition products, characterized in that, include: Partial discharge signal acquisition module, used to acquire transient partial discharge signals from GIS equipment; The SF6 decomposition product sampling and analysis module is used to obtain SF6 decomposition product detection information in the gas chamber of GIS equipment; The data processing module is used to extract fast variable characteristic parameters of partial discharge and slow variable characteristic parameters of decomposition products. Based on the difference in response time between the transient signal of partial discharge and the detection information of SF6 decomposition products, fast variable time windows and slow variable time windows are established respectively, and the temporal correlation between the two is established. A main criterion is constructed based on the fast variable characteristic parameters of partial discharge, an auxiliary criterion is constructed based on the slow variable characteristic parameters of decomposition products, and a rejection criterion for excluding non-fault factors is constructed. The anomaly degree of partial discharge is calculated based on the fast variable characteristic parameters of partial discharge, the anomaly degree of decomposition products is calculated based on the slow variable characteristic parameters of decomposition products, an interference penalty item is calculated based on the rejection criterion, and a comprehensive risk score is calculated based on the anomaly degree of partial discharge, the anomaly degree of decomposition products, and the interference penalty item. The output module, based on the comprehensive risk score and the combined results of the primary criteria, secondary criteria, and rejection criteria, is used to classify and confirm the status of GIS equipment and output the classification and confirmation results.

Images