Online monitoring method for electrical equipment

Abstract

The invention relates to the field of on-line monitoring of electrical equipment, in particular to an on-line monitoring method of the electrical equipment, which comprises the following steps: acquiring multi-dimensional monitoring parameters of the electrical equipment, including an electrical parameter, a temperature parameter and an insulation parameter; performing feature extraction on the collected multi-dimensional monitoring parameters to obtain feature parameters corresponding to the monitoring parameters, and constructing a feature vector set; based on historical normal operation data and real-time acquisition data of the electrical equipment, establishing a dynamic threshold model, and calculating a threshold corresponding to the characteristic parameter through the dynamic threshold model; feature parameters in the feature vector set are compared with corresponding thresholds output by the dynamic threshold model, and whether the electrical equipment is abnormal or not is judged; when the abnormity exists, outputting an abnormity monitoring result; and if not, returning to S1. According to the invention, by collecting multi-dimensional parameters, calibrating, correcting, de-noising and extracting features, and combining historical and real-time data to build a model and verify, data preprocessing, threshold adaptation and anomaly judgment are realized, and the monitoring reliability is improved.

Description

Technical Field

[0001] This invention relates to the field of online monitoring of electrical equipment, and more particularly to an online monitoring method for electrical equipment. Background Technology

[0002] As power systems develop towards higher voltage and larger capacity, the operational reliability of electrical equipment is crucial to system stability. Online monitoring technology, as a key means of ensuring equipment safety, has gradually evolved. Early technologies were mostly limited to monitoring single electrical or temperature parameters, making it difficult to comprehensively reflect the operating status of equipment. Although subsequent technologies expanded to multi-dimensional parameter acquisition, introduced devices such as Hall sensors and fiber optic grating sensors to improve data acquisition capabilities, and attempted to optimize monitoring effects through basic data processing, they still remain at a rudimentary stage of acquisition and simple comparison. A complete technical system has not yet been formed in terms of the accuracy control of parameter acquisition, effective suppression of data noise, and dynamic adaptability of anomaly judgment thresholds, making it difficult to meet the precise monitoring needs of equipment under complex operating conditions.

[0003] Existing online monitoring technologies have several shortcomings: First, some solutions lack a systematic calibration mechanism for data acquisition devices. Sensors or testing instruments are prone to data distortion due to zero-point offset and sensitivity decay. Furthermore, interference noise in the raw data is not adequately processed, directly affecting the accuracy of subsequent feature extraction. Second, anomaly detection often uses a fixed threshold mode, which cannot dynamically adjust the threshold range according to changes in equipment load type and fluctuations in operating conditions. This can easily lead to missed anomalies due to excessively high thresholds or false alarms due to excessively low thresholds. Third, most technologies only judge anomalies by comparing a single parameter with a threshold, without considering the inherent relationship between electrical, temperature, and insulation parameters. This results in one-sided anomaly detection, making it difficult to accurately identify potential equipment failure risks and potentially misleading maintenance decisions. Summary of the Invention

[0004] The purpose of this invention is to overcome the above-mentioned problems and provide an online monitoring method for electrical equipment. To achieve the above objective, this invention adopts the following technical solution:

[0005] A method for online monitoring of electrical equipment includes the following steps;

[0006] Step S1: Collect multi-dimensional monitoring parameters of electrical equipment, including electrical parameters, temperature parameters, and insulation parameters;

[0007] Step S2: Extract features from the collected multi-dimensional monitoring parameters to obtain the feature parameters corresponding to the monitoring parameters, and construct a feature vector set;

[0008] Step S3: Based on the historical normal operation data and real-time acquisition data of electrical equipment, establish a dynamic threshold model, and calculate the threshold corresponding to the feature parameters through the dynamic threshold model;

[0009] Step S4: Compare the feature parameters in the feature vector set with the corresponding thresholds output by the dynamic threshold model to determine whether there is an abnormality in the electrical equipment;

[0010] Step S5: If an anomaly is detected, output the anomaly monitoring result; if no anomaly is detected, return to step S1.

[0011] Further, in step S1, a Hall sensor is used to collect electrical parameters, a fiber optic grating sensor is used to collect temperature parameters, and a dielectric loss tester is used to collect insulation parameters; the collection frequency of the monitoring parameters is determined according to the rated voltage level of the electrical equipment, and the collection frequency and the rated voltage level satisfy the following relationship:

[0012]

[0013] Where f is the sampling frequency, k is the sampling coefficient, and U is the rated voltage level;

[0014] The electrical parameters include voltage, current, and power; the temperature parameters include equipment surface temperature, winding temperature, and ambient temperature; and the insulation parameters include dielectric loss and leakage current. After collecting the multi-dimensional monitoring parameters, wavelet threshold denoising processing is performed on them. The calculation formula for the denoised monitoring parameters is as follows:

[0015]

[0016] Among them, y t For the noise-reduced monitoring parameters, w j ,k is the wavelet coefficient, ψ(·) is the wavelet basis function, j is the scale index of the wavelet decomposition, J is the number of decomposition levels, K is the number of sampling points, and t is the sampling time.

[0017] Further, in step S2, the effective values ​​of electrical parameters, the contents of two characteristic subharmonics, the temperature rise rate of temperature parameters, and the dielectric loss and leakage current values ​​of insulation parameters are extracted; the constructed feature vector set is represented as follows:

[0018] X = [x1, x2, x3, x4, x5, x6];

[0019] Where X is the feature vector set, x1 is the effective value of the electrical parameter, x2 is the first characteristic harmonic content, x3 is the second characteristic harmonic content, x4 is the temperature rise rate of the temperature parameter, x5 is the dielectric loss value of the insulation parameter, and x6 is the leakage current value of the insulation parameter.

[0020] The formula for calculating the effective value of the electrical parameters is as follows:

[0021]

[0022] Where x1 is the effective value of the electrical parameter, N is the number of sampling points in one acquisition cycle, and i t The instantaneous value of the electrical parameter at the t-th sampling point;

[0023] The formula for calculating the temperature rise rate of the temperature parameter is:

[0024]

[0025] Where x4 is the temperature rise rate of the temperature parameter, T t The temperature value at the current moment, T t-Δt The temperature value is the temperature value at the previous acquisition time, and Δt is the acquisition interval.

[0026] Furthermore, based on the load type and operating conditions of electrical equipment, harmonic data of electrical equipment under historical normal operating conditions are collected, the proportion of each harmonic content to the effective value of electrical parameters is calculated, and the harmonics with the highest and second highest proportions are identified as the first and second characteristic harmonics.

[0027] The formula for calculating the first characteristic subharmonic content x2 is:

[0028]

[0029] Where x2 is the first characteristic subharmonic content, i 2,t Let be the instantaneous value of the first characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter;

[0030] The formula for calculating the second characteristic subharmonic content x3 is:

[0031]

[0032] Where x3 is the second characteristic subharmonic content, i 3,t Let be the instantaneous value of the second characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter.

[0033] Furthermore, in step S3, the average value of the feature parameters in the historical normal operation data is calculated as the historical benchmark value, and the average value of the real-time feature parameter data after feature extraction in the previous several collection cycles is calculated as the real-time benchmark value.

[0034] The formula for calculating the baseline value of the dynamic threshold model is as follows:

[0035] B = α × B0 + (1 - α) × B1;

[0036] Where B is the benchmark value, α is the weighting coefficient, B0 is the historical benchmark value, and B1 is the real-time benchmark value;

[0037] The threshold calculation formula for the i-th feature parameter is:

[0038] T i =B×(1+β) i );

[0039] Among them, T i β is the threshold for the i-th feature parameter. i Let be the threshold coefficient of the i-th feature parameter, where i is the index of the feature parameter;

[0040] The standard deviation of each characteristic parameter is calculated based on historical normal operation data, and the threshold coefficient is adjusted. The calculation formula is as follows:

[0041]

[0042] Where, β i γ is the threshold coefficient for the i-th feature parameter, γ is the adjustment coefficient, and σ is the threshold coefficient for the i-th feature parameter. i B is the standard deviation, and B is the benchmark value.

[0043] Further, in step S4, the formula for calculating the deviation between the i-th feature parameter and its corresponding threshold is:

[0044] Δx i =|x i -T i |;

[0045] Where, Δx i x is the deviation value. i For the i-th feature parameter, T i The threshold is denoted by i, and i is the index of the feature parameter.

[0046] The formula for calculating the anomaly index is:

[0047]

[0048] Where D is the anomaly index, ω i The weights of the i-th feature parameter are 1;

[0049] When D is greater than the preset anomaly threshold and there are at least two Δx i When the deviation exceeds a preset threshold, stratified verification is initiated; the effective value of the electrical parameter is cross-verified with the content of the first and second characteristic harmonics, and the ratio of the first characteristic harmonic content to the effective value and the ratio of the second characteristic harmonic content to the effective value are calculated.

[0050] When both ratios exceed the statistical range of the corresponding ratios in historical normal operation data, the electrical parameters are judged to be abnormal; the temperature rise rate of the temperature parameter is correlated with the dielectric loss value and leakage current value of the insulation parameter. When the temperature rise rate exceeds the threshold and the dielectric loss value or leakage current value exceeds the threshold, the equipment is judged to be abnormal.

[0051] Further, in step S5, the anomaly monitoring results include the anomaly characteristic parameter type, deviation value, and fault risk level; the fault risk level is determined based on the sum of the deviation values ​​of the anomaly characteristic parameters, and the sum calculation formula is:

[0052]

[0053] Where S is the sum of deviation values, i is the index of the characteristic parameter, and Δx i This represents the deviation between the i-th feature parameter and its corresponding threshold.

[0054] Further, in step S1, a standard signal matching the rated parameters of the electrical equipment is output from a standard signal source to calibrate the Hall sensor, fiber optic grating sensor, and dielectric loss tester. After calibration, the collected multi-dimensional monitoring parameters are corrected based on the calibration parameters. The corrected data is then used for wavelet threshold denoising. The correction formula is as follows:

[0055]

[0056] Among them, z t The data is the corrected monitoring data, where t is the sampling time and y′ is the data from the sampling point. t , b is the raw monitoring data collected before calibration, a is the zero offset of the sensor or instrument, and a is the sensitivity correction coefficient of the sensor or instrument.

[0057] Furthermore, in step S3, when preprocessing historical normal operation data, the 3σ criterion is used to remove data that exceeds the range of [μ-3σ, μ+3σ], where μ is the mean of the characteristic parameters in the historical normal operation data and σ is the standard deviation of the characteristic parameters in the historical normal operation data. The data after removing anomalies is smoothed by moving average.

[0058] The smoothing formula is:

[0059]

[0060] Where, q s This is the smoothed historical normal operation data, where s is the data sequence number, M is the moving average window size, k is the data index, and p... k This refers to historical normal operating data after removing outlier data.

[0061] The advantages of this invention are:

[0062] 1. This invention collects multi-dimensional monitoring parameters of electrical equipment, including electrical, temperature, and insulation parameters. It uses a standard signal source to calibrate the corresponding acquisition sensors and instruments, corrects the collected data, and performs wavelet threshold denoising. At the same time, it extracts key features of each parameter to construct a feature vector set, thereby achieving accurate acquisition and effective preprocessing of monitoring data. This avoids the limitations of single-dimensional parameter monitoring and monitoring errors caused by noise interference in the original data, providing high-quality data support for subsequent equipment anomaly judgment and improving the effectiveness and reliability of monitoring data.

[0063] 2. This invention establishes a dynamic threshold model by combining historical normal operation data of electrical equipment with real-time acquired data. The model's baseline value is calculated by weighting historical and real-time baseline values, and the threshold coefficients of each characteristic parameter are adjusted based on the standard deviation of historical data, thus achieving dynamic adaptation of the threshold. This method overcomes the shortcomings of fixed thresholds, which are difficult to adapt to changes in equipment load type and operating conditions, ensuring the rationality of thresholds under different operating conditions and effectively reducing the risk of abnormal misjudgments or missed judgments caused by fixed thresholds.

[0064] 3. This invention calculates the deviation values ​​and anomaly index of each characteristic parameter from its corresponding threshold. When preset conditions are met, stratified verification is initiated, cross-validating the correlation between electrical parameters and the correlation between temperature and insulation parameters. Simultaneously, it outputs the type of abnormal characteristic parameter, the deviation value, and the fault risk level, achieving accurate determination, location, and risk assessment of equipment anomalies. This process avoids the one-sidedness of judging from a single parameter, further improving the reliability and practicality of online monitoring and providing a clear basis for equipment maintenance. Attached Figure Description

[0065] The accompanying drawings, which form part of this application, are used to provide a further understanding of the application and to make other features, objects, and advantages of the application more apparent. The illustrative embodiments and descriptions of this application are used to explain the application and do not constitute an undue limitation of the application.

[0066] In the attached diagram:

[0067] Figure 1 This is a flowchart of an online monitoring method for electrical equipment in Example 1. Detailed Implementation

[0068] The present invention will now be described in detail and specifically through specific embodiments to enable a better understanding of the invention. However, the following embodiments do not limit the scope of protection of the present invention.

[0069] Example 1

[0070] like Figure 1 As shown, an online monitoring method for electrical equipment includes the following steps;

[0071] Step S1: Collect multi-dimensional monitoring parameters of electrical equipment, including electrical parameters, temperature parameters, and insulation parameters;

[0072] Step S2: Extract features from the collected multi-dimensional monitoring parameters to obtain the feature parameters corresponding to the monitoring parameters, and construct a feature vector set;

[0073] Step S3: Based on the historical normal operation data and real-time acquisition data of electrical equipment, establish a dynamic threshold model, and calculate the threshold corresponding to the feature parameters through the dynamic threshold model;

[0074] Step S4: Compare the feature parameters in the feature vector set with the corresponding thresholds output by the dynamic threshold model to determine whether there is an abnormality in the electrical equipment;

[0075] Step S5: If an anomaly is detected, output the anomaly monitoring result; if no anomaly is detected, return to step S1.

[0076] In a specific embodiment, online monitoring is implemented using a 10kV high-voltage switchgear as the monitoring object. Step S1 uses an A1324 Hall sensor, an FBG-S100 fiber optic grating sensor, and a JSY-800 dielectric loss tester to collect electrical parameters (voltage 220V, current 5A, power 1.1kW), temperature parameters (equipment surface 30℃, winding 35℃, ambient 25℃), and insulation parameters (dielectric loss 0.005, leakage current 10μA). Step S2 extracts key features of each parameter and constructs a feature vector set. Step S3 takes the historical normal operation data average of 5.1A as B0, the current three-cycle data average of 5.2A as B1, α=0.6, and calculates B=0.6×5.1+(1-0.6)×5.2=5.14A, establishing a dynamic threshold model. Step S4 compares the feature parameters with the threshold. Step S5 outputs an abnormal result; otherwise, returns to S1. This process enables multi-dimensional cyclic monitoring, avoiding the limitations of a single parameter and improving the comprehensiveness of monitoring.

[0077] Further, in step S1, a Hall sensor is used to collect electrical parameters, a fiber optic grating sensor is used to collect temperature parameters, and a dielectric loss tester is used to collect insulation parameters; the collection frequency of the monitoring parameters is determined according to the rated voltage level of the electrical equipment, and the collection frequency and the rated voltage level satisfy the following relationship:

[0078]

[0079] Where f is the sampling frequency, k is the sampling coefficient, and U is the rated voltage level;

[0080] The electrical parameters include voltage, current, and power; the temperature parameters include equipment surface temperature, winding temperature, and ambient temperature; and the insulation parameters include dielectric loss and leakage current. After collecting the multi-dimensional monitoring parameters, wavelet threshold denoising processing is performed on them. The calculation formula for the denoised monitoring parameters is as follows:

[0081]

[0082] Among them, y t For the noise-reduced monitoring parameters, w j ,k is the wavelet coefficient, ψ(·) is the wavelet basis function, j is the scale index of the wavelet decomposition, J is the number of decomposition levels, K is the number of sampling points, and t is the sampling time.

[0083] In a specific embodiment, step S1 uses an A1324 Hall sensor to collect electrical parameters, an FBG-S100 fiber optic grating sensor to collect temperature parameters, and a JSY-800 individual loss tester to collect insulation parameters. The equipment's rated voltage U = 10kV, and the acquisition coefficient k = 0.2 is used, according to the formula... Calculated collection rate The collected electrical parameters include voltage 220V, current 5A, and power 1.1kW; temperature parameters include surface temperature 30℃, winding temperature 35℃, and ambient temperature 25℃; insulation parameters include dielectric loss value 0.005 and leakage current 10μA. Wavelet threshold denoising was performed on the data, with J=3, K=100, t=1, and w... 1,1 =0.8, wavelet basis function ψ is taken as db4, calculated as follows This part has a value of 0.8 × 0.3 = 0.24, and after summing, y1 ≈ 220V. Noise reduction decreases data interference, providing a more accurate foundation for subsequent processing.

[0084] Further, in step S2, the effective values ​​of electrical parameters, the contents of two characteristic subharmonics, the temperature rise rate of temperature parameters, and the dielectric loss and leakage current values ​​of insulation parameters are extracted; the constructed feature vector set is represented as follows:

[0085] X = [x1, x2, x3, x4, x5, x6];

[0086] Where X is the feature vector set, x1 is the effective value of the electrical parameter, x2 is the first characteristic harmonic content, x3 is the second characteristic harmonic content, x4 is the temperature rise rate of the temperature parameter, x5 is the dielectric loss value of the insulation parameter, and x6 is the leakage current value of the insulation parameter.

[0087] The formula for calculating the effective value of the electrical parameters is as follows:

[0088]

[0089] Where x1 is the effective value of the electrical parameter, N is the number of sampling points in one acquisition cycle, and i t The instantaneous value of the electrical parameter at the t-th sampling point;

[0090] The formula for calculating the temperature rise rate of the temperature parameter is:

[0091]

[0092] Where x4 is the temperature rise rate of the temperature parameter, T t The temperature value at the current moment, T t-Δt The temperature value is the temperature value at the previous acquisition time, and Δt is the acquisition interval.

[0093] In a specific embodiment, step S2 extracts feature parameters to construct a feature vector set. When calculating the effective value x1 of the electrical parameter, N = 50, and the instantaneous value i at the sampling point is... t The values ​​are 5A, 5.1A…5.2A, according to the formula. Calculated When calculating the temperature rise rate x4, T t =38℃, T t-Δt =35℃, Δt=5min, according to the formula The calculated value is x4 = 0.6℃ / min.

[0094] The dielectric loss value x5 = 0.006 and the leakage current value x6 = 12 μA were extracted, and the first and second characteristic subharmonic contents x2 = 0.08 and x3 = 0.05 were determined, constructing X = [5.1, 0.08, 0.05, 0.6, 0.006, 12]. Key feature extraction makes the data more representative and improves the efficiency of subsequent judgments.

[0095] Furthermore, based on the load type and operating conditions of electrical equipment, harmonic data of electrical equipment under historical normal operating conditions are collected, the proportion of each harmonic content to the effective value of electrical parameters is calculated, and the harmonics with the highest and second highest proportions are identified as the first and second characteristic harmonics.

[0096] The formula for calculating the first characteristic subharmonic content x2 is:

[0097]

[0098] Where x2 is the first characteristic subharmonic content, i 2,t Let be the instantaneous value of the first characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter;

[0099] The formula for calculating the second characteristic subharmonic content x3 is:

[0100]

[0101] Where x3 is the second characteristic subharmonic content, i 3,t Let be the instantaneous value of the second characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter.

[0102] In a specific embodiment, based on the inductive load and full-load condition of the high-voltage switchgear, historical normal operation harmonic data is collected, and the proportion of each harmonic to x1 is calculated. The 3rd harmonic accounts for 8% and the 5th harmonic accounts for 5%, which are defined as the first and second characteristic harmonics. When calculating x2, N = 50, and the instantaneous values ​​of i2,t are 0.4A, 0.41A...0.42A, according to the formula... Calculated x2 = 0.41 / 5.1 ≈ 0.08. When calculating x3, the instantaneous values ​​of i3,t are 0.25A…0.26A, which is calculated as follows: x3 = 0.255 / 5.1 ≈ 0.05. Harmonic selection tailored to the operating conditions allows characteristic parameters to better match the actual equipment, improving the accuracy of anomaly identification.

[0103] Furthermore, in step S3, the average value of the feature parameters in the historical normal operation data is calculated as the historical benchmark value, and the average value of the real-time feature parameter data after feature extraction in the previous several collection cycles is calculated as the real-time benchmark value.

[0104] The formula for calculating the baseline value of the dynamic threshold model is as follows:

[0105] B = α × B0 + (1 - α) × B1;

[0106] Where B is the benchmark value, α is the weighting coefficient, B0 is the historical benchmark value, and B1 is the real-time benchmark value;

[0107] The threshold calculation formula for the i-th feature parameter is:

[0108] T i =B×(1+β) i );

[0109] Among them, T i β is the threshold for the i-th feature parameter. i Let be the threshold coefficient of the i-th feature parameter, where i is the index of the feature parameter;

[0110] The standard deviation of each characteristic parameter is calculated based on historical normal operation data, and the threshold coefficient is adjusted. The calculation formula is as follows:

[0111]

[0112] Where, β i γ is the threshold coefficient for the i-th feature parameter, γ is the adjustment coefficient, and σ is the threshold coefficient for the i-th feature parameter.i B is the standard deviation, and B is the benchmark value.

[0113] In a specific embodiment, step S3 calculates the historical baseline value B0 = 5.1A, the real-time baseline value B1 = 5.2A, takes α = 0.6, and calculates B = 5.14A according to B = α × B0 + (1 - α) × B1. The standard deviation of historical data x1 σ1 = 0.1A, takes γ = 2, and calculates according to... The calculated β1 = 2 × (0.1 / 5.14) ≈ 0.039, then according to T... i =B×(1+β) i The calculated T1 = 5.14 × (1 + 0.039) ≈ 5.34A. For x4, B = 0.5℃ / min, σ4 = 0.05℃ / min, the calculated β4 = 0.2, T4 = 0.6℃ / min. Dynamic thresholds adapt to changes in operating conditions, avoiding misjudgments or missed judgments caused by fixed thresholds.

[0114] Further, in step S4, the formula for calculating the deviation between the i-th feature parameter and its corresponding threshold is:

[0115] Δx i =|x i -Y i |;

[0116] Where, Δx i x is the deviation value. i For the i-th feature parameter, T i The threshold is denoted by i, and i is the index of the feature parameter.

[0117] The formula for calculating the anomaly index is:

[0118]

[0119] Where D is the anomaly index, ω i The weights of the i-th feature parameter are 1;

[0120] When D is greater than the preset anomaly threshold and there are at least two Δx i When the deviation exceeds a preset threshold, stratified verification is initiated; the effective value of the electrical parameter is cross-verified with the content of the first and second characteristic harmonics, and the ratio of the first characteristic harmonic content to the effective value and the ratio of the second characteristic harmonic content to the effective value are calculated.

[0121] When both ratios exceed the statistical range of the corresponding ratios in historical normal operation data, the electrical parameters are judged to be abnormal; the temperature rise rate of the temperature parameter is correlated with the dielectric loss value and leakage current value of the insulation parameter. When the temperature rise rate exceeds the threshold and the dielectric loss value or leakage current value exceeds the threshold, the equipment is judged to be abnormal.

[0122] In a specific embodiment, step S4 calculates the deviation value and the anomaly index. x1 = 5.5A, T1 = 5.34A, according to Δx i =|x i -T i |Calculated Δx1=0.16A; x4=0.7℃ / min, T4=0.6℃ / min, Δx4=0.1℃ / min. Take ω1=0.2, ω2=0.2, ω3=0.15, ω4=0.2, ω5=0.15, ω6=0.1, Δx2=0.02, Δx3=0.01, Δx5=0.001, Δx6=2μA, press The calculated D≈0.18, exceeding the preset threshold of 0.15, and Δx1 and D4 exceeding the deviation threshold, thus initiating stratified verification. The ratios of the 3rd and 5th harmonics to x1 exceeded historical ranges, and the temperature rise rate and dielectric loss exceeded thresholds, indicating anomalies. Stratified verification avoids the bias of judging by a single parameter and improves accuracy.

[0123] Further, in step S5, the anomaly monitoring results include the anomaly characteristic parameter type, deviation value, and fault risk level; the fault risk level is determined based on the sum of the deviation values ​​of the anomaly characteristic parameters, and the sum calculation formula is:

[0124]

[0125] Where S is the sum of deviation values, i is the index of the characteristic parameter, and Δx i This represents the deviation between the i-th feature parameter and its corresponding threshold.

[0126] In a specific embodiment, step S5 outputs the abnormal monitoring results. The abnormal characteristic parameters are the effective value of current, temperature rise rate, and dielectric loss value, corresponding to Δx1 = 0.16A, Δx4 = 0.1℃ / min, and Δx5 = 0.001, with additional parameters Δx2 = 0.02, Δx3 = 0.01, and Δx6 = 2μA. The calculated value of S is 2.291. The default risk levels are S<1 (low risk), 1-3 (medium risk), and 3 (high risk). This is classified as medium risk. The output clearly identifies anomalies and risk levels, allowing maintenance personnel to accurately formulate maintenance plans, reducing blind maintenance and improving equipment maintenance efficiency.

[0127] Further, in step S1, a standard signal matching the rated parameters of the electrical equipment is output from a standard signal source to calibrate the Hall sensor, fiber optic grating sensor, and dielectric loss tester. After calibration, the collected multi-dimensional monitoring parameters are corrected based on the calibration parameters. The corrected data is then used for wavelet threshold denoising. The correction formula is as follows:

[0128]

[0129] Among them, zt The data is the corrected monitoring data, where t is the sampling time and y′ is the data from the sampling point. t , b is the raw monitoring data collected before calibration, a is the zero offset of the sensor or instrument, and a is the sensitivity correction coefficient of the sensor or instrument.

[0130] In a specific embodiment, step S1 uses an SG3525 standard signal source to calibrate the acquisition device, outputting a 220V standard voltage signal matching the device's rated parameters. When calibrating the A1324 Hall sensor, its zero-point offset b = 0.5V and sensitivity correction coefficient a = 1.02 were measured. The original voltage data y′t = 221V acquired before calibration was... Calculated The value is close to the standard value of 214V. After calibration, the data is corrected and then denoised to eliminate the effects of sensor zero-point offset and sensitivity attenuation, making the collected data more accurate and providing reliable data support for subsequent feature extraction and anomaly detection.

[0131] Furthermore, in step S3, when preprocessing historical normal operation data, the 3σ criterion is used to remove data that exceeds the range of [μ-3σ, μ+3σ], where μ is the mean of the characteristic parameters in the historical normal operation data and σ is the standard deviation of the characteristic parameters in the historical normal operation data. The data after removing anomalies is smoothed by moving average.

[0132] The smoothing formula is:

[0133]

[0134] Where, q s This is the smoothed historical normal operation data, where s is the data sequence number, M is the moving average window size, k is the data index, and p... k This refers to historical normal operating data after removing outlier data.

[0135] In a specific embodiment, step S3 preprocesses historical normal operation data. For the historical data of x1, the mean μ = 5.1A and standard deviation σ = 0.1A are calculated. Data with a mean of 4.7A and a standard deviation of 5.5A are removed according to the 3σ standard. A moving average window M = 5 is taken, and the data after removing outliers are p1 = 5.0A, p2 = 5.1A, p3 = 5.2A, p4 = 5.1A, and p5 = 5.0A. When s = 5, according to... Calculated Preprocessed data is more stable, reducing the impact of historical data fluctuations on baseline value calculation and improving the reliability of the dynamic threshold model.

[0136] The specific embodiments of the present invention have been described in detail above, but they are merely examples, and the present invention is not equivalent to the specific embodiments described above. For those skilled in the art, any equivalent modifications and substitutions to the present invention are also within the scope of the present invention. Therefore, all equivalent transformations and modifications made without departing from the spirit and scope of the present invention should be covered within the scope of the present invention.

Claims

1. A method for online monitoring of electrical equipment, characterized in that, Includes the following steps; Step S1: Collect multi-dimensional monitoring parameters of electrical equipment, including electrical parameters, temperature parameters, and insulation parameters; Step S2: Extract features from the collected multi-dimensional monitoring parameters to obtain the feature parameters corresponding to the monitoring parameters, and construct a feature vector set; Step S3: Based on the historical normal operation data and real-time acquisition data of electrical equipment, establish a dynamic threshold model, and calculate the threshold corresponding to the feature parameters through the dynamic threshold model; Step S4: Compare the feature parameters in the feature vector set with the corresponding thresholds output by the dynamic threshold model to determine whether there is an abnormality in the electrical equipment; Step S5: If an anomaly is detected, output the anomaly monitoring results; If no abnormality is found, return to step S1.

2. The online monitoring method for electrical equipment according to claim 1, characterized in that, In step S1, a Hall sensor is used to collect electrical parameters, a fiber optic grating sensor is used to collect temperature parameters, and a dielectric loss tester is used to collect insulation parameters. The collection frequency of the monitoring parameters is determined according to the rated voltage level of the electrical equipment, and the collection frequency and the rated voltage level satisfy the following relationship: Where f is the sampling frequency, k is the sampling coefficient, and U is the rated voltage level; The electrical parameters include voltage, current, and power; the temperature parameters include equipment surface temperature, winding temperature, and ambient temperature; and the insulation parameters include dielectric loss and leakage current. After collecting the multi-dimensional monitoring parameters, wavelet threshold denoising processing is performed on them. The calculation formula for the denoised monitoring parameters is as follows: Among them, y t For the noise-reduced monitoring parameters, w j ,k is the wavelet coefficient, ψ(·) is the wavelet basis function, j is the scale index of the wavelet decomposition, J is the number of decomposition levels, K is the number of sampling points, and t is the sampling time.

3. The online monitoring method for electrical equipment according to claim 2, characterized in that, In step S2, the effective values ​​of electrical parameters, the contents of two characteristic subharmonics, the temperature rise rate of temperature parameters, and the dielectric loss and leakage current values ​​of insulation parameters are extracted; the constructed feature vector set is represented as follows: X = [x1, x2, x3, x4, x5, x6]; Where X is the feature vector set, x1 is the effective value of the electrical parameter, x2 is the first characteristic harmonic content, x3 is the second characteristic harmonic content, x4 is the temperature rise rate of the temperature parameter, x5 is the dielectric loss value of the insulation parameter, and x6 is the leakage current value of the insulation parameter. The formula for calculating the effective value of the electrical parameters is as follows: Where x1 is the effective value of the electrical parameter, N is the number of sampling points in one acquisition cycle, and i t The instantaneous value of the electrical parameter at the t-th sampling point; The formula for calculating the temperature rise rate of the temperature parameter is: Where x4 is the rate of temperature rise, T t T represents the temperature at the current moment. t-Δt The temperature value is the temperature value at the previous acquisition time, and Δt is the acquisition interval.

4. The online monitoring method for electrical equipment according to claim 3, characterized in that, Based on the load type and operating conditions of electrical equipment, harmonic data of electrical equipment under historical normal operating conditions are collected, the proportion of each harmonic content to the effective value of electrical parameters is calculated, and the harmonics with the highest and second highest proportions are identified as the first and second characteristic harmonics. The formula for calculating the first characteristic subharmonic content x2 is: Where x2 is the first characteristic subharmonic content, i 2,t Let be the instantaneous value of the first characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter; The formula for calculating the second characteristic subharmonic content x3 is: Where x3 is the second characteristic subharmonic content, i 3,t Let be the instantaneous value of the second characteristic harmonic at the t-th sampling point, N be the number of sampling points in one acquisition cycle, and x1 be the effective value of the electrical parameter.

5. The online monitoring method for electrical equipment according to claim 4, characterized in that, In step S3, the average value of the feature parameters in the historical normal operation data is calculated as the historical benchmark value, and the average value of the real-time feature parameter data after feature extraction in the previous several collection cycles is calculated as the real-time benchmark value. The formula for calculating the baseline value of the dynamic threshold model is as follows: B = α × B0 + (1 - α) × B1; Where B is the benchmark value, α is the weighting coefficient, B0 is the historical benchmark value, and B1 is the real-time benchmark value; The threshold calculation formula for the i-th feature parameter is: T i =B×(1+β i ); Among them, T i β is the threshold for the i-th feature parameter. i Let be the threshold coefficient of the i-th feature parameter, where i is the index of the feature parameter; The standard deviation of each characteristic parameter is calculated based on historical normal operation data, and the threshold coefficient is adjusted. The calculation formula is as follows: Where, β i γ is the threshold coefficient for the i-th feature parameter, γ is the adjustment coefficient, and σ is the threshold coefficient for the i-th feature parameter. i B is the standard deviation, and B is the benchmark value.

6. The online monitoring method for electrical equipment according to claim 5, characterized in that, In step S4, the formula for calculating the deviation between the i-th feature parameter and its corresponding threshold is: Δx i =|x i -T i |; Where, Δx i x is the deviation value. i For the i-th feature parameter, T i The threshold is denoted by i, and i is the index of the feature parameter. The formula for calculating the anomaly index is: Where D is the anomaly index, ω i The weights of the i-th feature parameter are 1; When D is greater than the preset anomaly threshold and there are at least two Δx i When the deviation exceeds a preset threshold, stratified verification is initiated; the effective value of the electrical parameter is cross-verified with the content of the first and second characteristic harmonics, and the ratio of the first characteristic harmonic content to the effective value and the ratio of the second characteristic harmonic content to the effective value are calculated. When both ratios exceed the statistical range of the corresponding ratios in historical normal operation data, the electrical parameters are judged to be abnormal; the temperature rise rate of the temperature parameter is correlated with the dielectric loss value and leakage current value of the insulation parameter. When the temperature rise rate exceeds the threshold and the dielectric loss value or leakage current value exceeds the threshold, the equipment is judged to be abnormal.

7. The online monitoring method for electrical equipment according to claim 6, characterized in that, In step S5, the anomaly monitoring results include the anomaly characteristic parameter type, deviation value, and fault risk level; the fault risk level is determined based on the sum of the deviation values ​​of the anomaly characteristic parameters, and the sum is calculated using the following formula: Where S is the sum of deviation values, i is the index of the characteristic parameter, and Δx i This represents the deviation between the i-th feature parameter and its corresponding threshold.

8. The online monitoring method for electrical equipment according to claim 7, characterized in that, In step S1, a standard signal matching the rated parameters of the electrical equipment is output from a standard signal source to calibrate the Hall sensor, fiber optic grating sensor, and dielectric loss tester. After calibration, the acquired multi-dimensional monitoring parameters are corrected based on the calibration parameters. The corrected data is then used for wavelet threshold denoising. The correction formula is as follows: Among them, z t The data is the corrected monitoring data, where t is the sampling time and y′ is the data from the sampling point. t , b is the raw monitoring data collected before calibration, a is the zero offset of the sensor or instrument, and a is the sensitivity correction coefficient of the sensor or instrument.

9. The online monitoring method for electrical equipment according to claim 8, characterized in that, In step S3, when preprocessing historical normal operation data, the 3σ criterion is used to remove data that exceed the range of [μ-3σ, μ+3σ], where μ is the mean of the characteristic parameters in the historical normal operation data and σ is the standard deviation of the characteristic parameters in the historical normal operation data. The data after removing anomalies is smoothed by moving average. The smoothing formula is: Where, q s This is the smoothed historical normal operation data, where s is the data sequence number, M is the moving average window size, k is the data index, and p... k This refers to historical normal operating data after removing outlier data.

Images