Power equipment defect identification system and method based on multi-sensor data fusion

Abstract

The invention discloses a power equipment defect identification system and method based on multi-sensor data fusion, and relates to the field of multi-sensor information fusion. A partial discharge sensor, a dual-light temperature measurement sensor, a gas sensor and a grounding circulation sensor are deployed in a monitoring area as multi-type sensors to obtain power data; after preprocessing, data characteristics of partial discharge, dual-light temperature measurement, gas and grounding circulation are extracted. The correlation degree is obtained by calculating the correlation coefficient, covariance and mutual information among the characteristics and carrying out weighted average, and the characteristics associated with the equipment defects are screened out. A prediction model is established based on equipment defect associated features, a regression coefficient is solved by adopting a least square method, the model is corrected regularly according to historical data, and early warning of different grades is given out according to defect degree grades. According to the method, multi-source data are fused, key features are screened, the model is dynamically updated, accurate identification and graded early warning of the defects of the power equipment are achieved, and the monitoring reliability and real-time performance are improved.

Description

Technical Field

[0001] The present invention relates to the field of multi-sensor information fusion, and in particular to a system and method for identifying defects in electric power equipment based on multi-sensor data fusion. Background Art

[0002] The field of power equipment defect identification has gradually transitioned from traditional manual inspections to intelligent identification systems. Early systems relied primarily on manual visual inspections or simple instrumentation, which presented challenges such as low efficiency, limited coverage, and high risks associated with operating in high-risk environments. With technological advancements, intelligent identification technology based on multi-source sensors has become mainstream. These technologies utilize high-precision cameras, infrared thermal imagers, partial discharge sensors, and other equipment mounted on drones and robots, enabling all-weather, multi-dimensional data collection.

[0003] However, current intelligent recognition technology mainly performs threshold detection and early warning on a single physical quantity, and is unable to achieve joint prediction of different physical quantities. In addition, there are many types of power equipment and their operating conditions vary greatly. There is an urgent need for a universal prediction framework to integrate the data of power equipment. Summary of the Invention

[0004] The purpose of the present invention is to provide a system and method for identifying defects in power equipment based on multi-sensor data fusion, so as to solve the problems raised in the prior art.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] Step S1: deploy multiple types of sensors in the monitoring area, wherein the multiple types of sensors include a partial discharge sensor, a dual-light temperature sensor, a gas sensor, and a ground loop current sensor; and obtain power data through the multiple types of sensors;

[0007] The monitoring area is the area where defects in power equipment are identified; the partial discharge sensor includes a high-frequency current sensor and an ultrasonic sensor; the dual-light temperature sensor includes an infrared thermal imager and a visible light camera; the gas sensor includes a multi-gas detector for detecting sulfur hexafluoride and its decomposition products; the ground loop current sensor includes a through-type current transformer; and the power data includes data directly obtained by the multiple types of sensors.

[0008] Step S2: extracting features from the power data after preprocessing, calculating correlations between the features, and screening out features associated with equipment defects;

[0009] Step S2-1, the preprocessing includes: calculating the mean μ and standard deviation σ of the power data, treating data with values outside the range of [μ-3σ, μ+3σ] as outliers and removing them; filling missing data with linear interpolation; and normalizing the power data;

[0010] Step S2-2, extracting features of the pre-processed power data, where the pre-processed features include partial discharge data features, dual-light temperature measurement data features, gas data features, and ground loop current data features;

[0011] The partial discharge data features include: the average apparent discharge amount, the number of discharges, the average discharge pulse width, and the discharge energy; the dual-light temperature measurement data features include: the maximum surface temperature of the equipment, the temperature gradient, the temperature change rate, and the area of the heating area; the gas data features include: the average concentration of each gas, the gas concentration change rate, and the gas concentration ratio; the ground loop current data features include: the average ground loop current amplitude, the ground loop current waveform distortion rate, and the ground loop current change rate;

[0012] Step S2-3, calculating the mean value of the features of the same type in the extracted preprocessed features, and calculating the correlation coefficient and covariance between any two different features based on the value of the features of the same type in the preprocessed features and the mean value; calculating the joint probability density function and the marginal probability density function of each of the different features, and calculating the mutual information between any two different features based on the joint probability density function and the marginal probability density function;

[0013] Step S2-4: Calculate the degree of association between different pairs of features using a weighted average method:

[0014]

[0015] where a ij is the degree of association, i and j represent different features, w1, w2, w3 are weight coefficients, ρ ij is the correlation coefficient between i and j, is the covariance between i and j, MI ij is the mutual information between i and j;

[0016] A correlation threshold is set, and the correlation degree with an absolute value greater than the correlation threshold is screened out. The two features corresponding to the correlation degree are used as the features of equipment defect correlation.

[0017] Step S3: Establish a prediction model based on the selected features, and regularly update the prediction model.

[0018] Quantify the degree of defects in power equipment and establish a prediction model based on the degree of defects in power equipment and the characteristics associated with the defects in the equipment:

[0019] D=β0+β1F1+β2F2+...+β m F m +∈

[0020] Where D represents the quantified degree of power equipment defects, β0 is the intercept term, β1, β2...β m is the regression coefficient, ∈ is the error term, F1, F2...F m is the value of the characteristic associated with the equipment defect when the current power equipment is defective; the regression coefficient is solved by the least square method;

[0021] A model update cycle is set, and after a model update cycle, historical data within the model update cycle is integrated, and the regression coefficient of the prediction model is corrected based on the integrated data.

[0022] Step S4: pre-process the real-time collected power data and input it into the prediction model, and perform an early warning response according to the prediction model.

[0023] The real-time collected power data is pre-processed and features are extracted and input into the prediction model to obtain the value of the degree of defect of the power equipment; the degree of defect is graded, and different levels of warnings are issued according to the classification and the value of the degree of defect of the power equipment.

[0024] Compared with the prior art, the present invention has the following beneficial effects:

[0025] 1. This invention deploys multiple sensors, including partial discharge, dual-light temperature measurement, gas, and ground loop current, to comprehensively collect power data, overcoming the limitations of single-sensor monitoring. Preprocessing removes outliers, fills missing values, and performs normalization to extract multidimensional features. Correlation coefficients, covariance, and mutual information are used to weight correlation and calculate correlations, screening key features and laying the foundation for accurate defect identification.

[0026] 2. This invention quantifies the degree of defects in power equipment and establishes a prediction model using the least squares method. This model integrates historical data and regularly adjusts the regression coefficients to adapt to changing equipment operating conditions. Real-time data is processed and input into the model, which then issues graded warnings based on defect severity, enabling dynamic and intelligent monitoring and improving the accuracy and timeliness of defect identification.

[0027] 3. This invention builds a complete system encompassing monitoring deployment, data processing, model building, and real-time application, with each module operating collaboratively. Combining multi-sensor fusion acquisition with multi-dimensional feature analysis, supplemented by an updateable prediction model and a hierarchical early warning mechanism, it forms a highly versatile power equipment defect identification framework, effectively solving the monitoring challenges faced by a wide range of equipment types and widely varying operating conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 Schematic diagram of the flow of the method for identifying defects in power equipment based on multi-sensor data fusion according to the present invention;

[0029] Figure 2Schematic diagram of the structure of the power equipment defect identification system based on multi-sensor data fusion of the present invention; DETAILED DESCRIPTION

[0030] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0031] Example 1: Figure 1 As shown, the present invention provides a technical solution, a method for identifying defects in power equipment based on multi-sensor data fusion, and the method for identifying defects in power equipment includes the following steps:

[0032] Step S1: deploy multiple types of sensors in the monitoring area, wherein the multiple types of sensors include a partial discharge sensor, a dual-light temperature sensor, a gas sensor, and a ground loop current sensor; and obtain power data through the multiple types of sensors;

[0033] The monitoring area is the area where defects in power equipment are identified; the partial discharge sensor includes a high-frequency current sensor and an ultrasonic sensor; the dual-light temperature sensor includes an infrared thermal imager and a visible light camera; the gas sensor includes a multi-gas detector for detecting sulfur hexafluoride and its decomposition products; the ground loop current sensor includes a through-type current transformer; and the power data includes data directly obtained by the multiple types of sensors.

[0034] Step S2: extracting features from the power data after preprocessing, calculating correlations between the features, and screening out features associated with equipment defects;

[0035] Step S2-1, the preprocessing includes: calculating the mean μ and standard deviation σ of the power data, treating data with values outside the range of [μ-3σ, μ+3σ] as outliers and removing them; filling missing data with linear interpolation; and normalizing the power data;

[0036] Step S2-2, extracting features of the pre-processed power data, where the pre-processed features include partial discharge data features, dual-light temperature measurement data features, gas data features, and ground loop current data features;

[0037] The partial discharge data features include: the average apparent discharge amount, the number of discharges, the average discharge pulse width, and the discharge energy; the dual-light temperature measurement data features include: the maximum surface temperature of the equipment, the temperature gradient, the temperature change rate, and the area of the heating area; the gas data features include: the average concentration of each gas, the gas concentration change rate, and the gas concentration ratio; the ground loop current data features include: the average ground loop current amplitude, the ground loop current waveform distortion rate, and the ground loop current change rate;

[0038] Step S2-3, calculating the mean value of the features of the same type in the extracted preprocessed features, and calculating the correlation coefficient and covariance between any two different features based on the value of the features of the same type in the preprocessed features and the mean value; calculating the joint probability density function and the marginal probability density function of each of the different features, and calculating the mutual information between any two different features based on the joint probability density function and the marginal probability density function;

[0039] Step S2-4: Calculate the degree of association between different pairs of features using a weighted average method:

[0040]

[0041] where a ij is the degree of association, i and j represent different features, w1, w2, w3 are weight coefficients, ρ ij is the correlation coefficient between i and j, is the covariance between i and j, MI ij is the mutual information between i and j;

[0042] A correlation threshold is set, and the correlation degree with an absolute value greater than the correlation threshold is screened out. The two features corresponding to the correlation degree are used as the features of equipment defect correlation.

[0043] Step S3: Establish a prediction model based on the selected features, and regularly update the prediction model.

[0044] Quantify the degree of defects in power equipment and establish a prediction model based on the degree of defects in power equipment and the characteristics associated with the defects in the equipment:

[0045] D=β0+β1F1+β2F2+...+β m F m +∈

[0046] Where D represents the quantified degree of power equipment defects, β0 is the intercept term, β1, β2...β m is the regression coefficient, ∈ is the error term, F1, F2...F mis the value of the characteristic associated with the equipment defect when the current power equipment is defective; the regression coefficient is solved by the least square method;

[0047] A model update cycle is set, and after a model update cycle, historical data within the model update cycle is integrated, and the regression coefficient of the prediction model is corrected based on the integrated data.

[0048] Step S4: pre-process the real-time collected power data and input it into the prediction model, and perform an early warning response according to the prediction model.

[0049] The real-time collected power data is pre-processed and features are extracted and input into the prediction model to obtain the value of the degree of defect of the power equipment; the degree of defect is graded, and different levels of warnings are issued according to the classification and the value of the degree of defect of the power equipment.

[0050] For example:

[0051] Take the 110kV transformer of a substation as the monitoring area and the collected apparent discharge as an example:

[0052] There are 10 groups of apparent discharge data, the unit is pC

[0053] [10,15,20,25,30,80,22,18,20,21]

[0054] If the data between 10 and 15 are missing, linear interpolation is used to fill in the missing data:

[0055] x i-1 =10

[0056] x i+1 =15

[0057]

[0058] Now that there is no missing data, calculate the mean:

[0059]

[0060] Calculate the standard deviation:

[0061]

[0062] The range is: [26.1-3×18.7,26.1+3×18.7]=[-30,82.2]. All 10 sets of data are within the range and are retained.

[0063] Normalize the 10 sets of data according to the minimum-maximum rule. For example, select the apparent discharge data with a value of 20pC:

[0064] It is known that the smallest apparent discharge in the 10 sets of data is 10pC and the largest apparent discharge is 80pC. The normalization process is:

[0065]

[0066] Extract the characteristics of power data. Take the average apparent discharge value q and the maximum surface temperature T of the equipment as an example. There are 10 sets of data of the average apparent discharge value q and the maximum surface temperature T of the equipment at the same time. They are presented as sample - average apparent discharge value - maximum surface temperature of the equipment:

[0067] 1-10-40, 2-12-42, 3-15-45, 4-8-38, 5-11-41, 6-13-43, 7-14-44, 8-9-39, 9-10.5-40.5, 10-12.5-42.5

[0068] First, calculate the mean of the apparent discharge values and the mean of the maximum surface temperature of the equipment:

[0069]

[0070] According to the formula:

[0071]

[0072] where ρ XY is the correlation coefficient, X and Y are two different types of features, and are the means of X and Y respectively, n is the number of data points, and i is the number of sampling times;

[0073] Calculate the correlation coefficient between the average apparent discharge value and the maximum surface temperature of the equipment:

[0074]

[0075] According to the formula:

[0076]

[0077] Among them C ov is the covariance, X and Y are two different types of features, and are the means of X and Y respectively, n is the number of data points, and i is the number of sampling times;

[0078] Similarly, the covariance between the mean value of apparent discharge and the maximum surface temperature of the equipment is calculated:

[0079] C ov ≈6

[0080] From the correlation coefficient and covariance, it can be seen that the mean value of apparent discharge and the maximum surface temperature of the equipment are highly linearly correlated.

[0081] Similar to the above solution, according to the formula:

[0082]

[0083] Where MI(X,Y) is the value of mutual information, p(x,y) is the joint probability density function of X and Y, and p(x) and p(y) are the marginal probability density functions of X and Y respectively;

[0084] Nonlinear relationships between features can be calculated.

[0085] Calculations yield characteristics associated with equipment defects. For example, the mean apparent discharge, number of discharges, and maximum equipment surface temperature are calculated as F1, F2, and F3, respectively. The degree of defects in the power equipment is then quantified and assigned a value, D. For example, during an assessment, a transformer was found to have mild partial discharge, which did not significantly impact normal operation. Experts comprehensively scored the defect and determined it to be a minor defect, assigning the dependent variable D a value of 1. Another assessment detected abnormal gas concentration and elevated temperature inside the transformer, indicating a moderate defect, with D assigned a value of 2. Severe defects were assigned a value of 3. After multiple assessments, a series of dependent variable values corresponding to the degree of equipment defects were obtained.

[0086] Now, based on the data collected based on the characteristics of the three equipment defects, the corresponding power equipment defect degree is found according to the formula:

[0087] D=β0+β1F1+β2F2+...+β m F m +∈

[0088] The intercept term is set to 1, the error term is set to 0, and the prediction model is established to obtain:

[0089] D=β0+β1F1+β2F2+β3F3

[0090] The least squares method is used to solve the regression coefficient. For example, the prediction model is obtained at this time:

[0091] D=-2.1+0.1F1+0.2F2+0.05F3

[0092] The degree of the defect is graded, and different levels of warnings are issued according to the grade and the value of the degree of the defect of the power equipment.

[0093] For example, a D value less than 1 is defined as a slight defect, a D value greater than or equal to 1 and less than 2 is defined as a moderate defect, and a D value greater than or equal to 2 is defined as a severe defect.

[0094] The real-time collected power data is pre-processed to extract features associated with equipment defects, which are input into the prediction model, and an early warning response is performed based on the prediction model.

[0095] Example 2, as Figure 2 As shown, the present invention provides a power equipment defect identification system based on multi-sensor data fusion, the intelligent power equipment defect identification system includes a monitoring deployment module, a data processing module, a model building module, and a real-time application module; the monitoring deployment module is used to deploy multiple types of sensors in the monitoring area and obtain power data; the data processing module pre-processes the power data, extracts features, and calculates correlation, and filters out features associated with equipment defects; the model building module establishes a prediction model based on the features associated with equipment defects and regularly corrects the regression coefficient based on historical data; the real-time application module is used to process the real-time collected power data and input it into the prediction model, and trigger an early warning response based on the output value of the power defect degree;

[0096] The output end of the monitoring deployment module is electrically connected to the input end of the data processing module; the output end of the data processing module is electrically connected to the input end of the model construction module; the output end of the model construction module is electrically connected to the input end of the implementation application module.

[0097] The monitoring deployment module includes a sensor deployment unit and a data acquisition unit; the sensor deployment unit is used to deploy multiple types of sensors in the monitoring area; and the data acquisition unit is used to acquire power data through the multiple types of sensors.

[0098] The data processing module includes a data preprocessing unit and a feature association processing unit; the data preprocessing unit is used to eliminate outliers in the power data, fill missing values, and normalize the power data; the feature association processing unit is used to extract the features of the power data and screen out features associated with equipment defects through correlation coefficient, covariance, mutual information, and weighted average method.

[0099] The model construction module includes a model construction unit and a model update unit; the model construction unit is used to establish a prediction model based on the characteristics associated with equipment defects and the quantified degree of defects in the power equipment; the model update unit is used to set a model update cycle and correct the regression coefficient of the prediction model according to the set model update cycle.

[0100] The real-time application module includes an input processing unit and an early warning response unit; the input processing unit is used to preprocess and extract features of the real-time collected power data; the early warning response unit is used to issue graded early warnings based on the output value of the degree of defect of the power equipment.

[0101] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for identifying defects in power equipment based on multi-sensor data fusion, characterized by: Step S1: deploy multiple types of sensors in the monitoring area, wherein the multiple types of sensors include a partial discharge sensor, a dual-light temperature sensor, a gas sensor, and a ground loop current sensor; and obtain power data through the multiple types of sensors; Step S2: extracting features from the power data after preprocessing, calculating correlations between the features, and screening out features associated with equipment defects; Step S3: establishing a prediction model based on the selected features, and regularly updating the prediction model; Step S4: pre-process the real-time collected power data and input it into the prediction model, and perform an early warning response according to the prediction model.

2. The method for identifying defects in power equipment based on multi-sensor data fusion according to claim 1, characterized in that: The step S1 further includes: the monitoring area is an area for identifying defects in power equipment; the partial discharge sensor includes a high-frequency current sensor and an ultrasonic sensor; the dual-light temperature sensor includes an infrared thermal imager and a visible light camera; the gas sensor includes a multi-gas detector for detecting sulfur hexafluoride and its decomposition products; the ground loop current sensor includes a through-type current transformer; and the power data includes data directly obtained by the multiple types of sensors.

3. The method for identifying defects in power equipment based on multi-sensor data fusion according to claim 1, characterized in that: The step S2 further includes: Step S2-1, the preprocessing includes: calculating the mean μ and standard deviation σ of the power data, treating data with values outside the range of [μ-3σ, μ+3σ] as outliers and removing them; filling missing data with linear interpolation; and normalizing the power data; Step S2-2, extracting features of the pre-processed power data, where the pre-processed features include partial discharge data features, dual-light temperature measurement data features, gas data features, and ground loop current data features; The partial discharge data features include: the average apparent discharge amount, the number of discharges, the average discharge pulse width, and the discharge energy; the dual-light temperature measurement data features include: the maximum surface temperature of the equipment, the temperature gradient, the temperature change rate, and the area of the heating area; the gas data features include: the average concentration of each gas, the gas concentration change rate, and the gas concentration ratio; the ground loop current data features include: the average ground loop current amplitude, the ground loop current waveform distortion rate, and the ground loop current change rate; Step S2-3, calculating the mean value of the features of the same type in the extracted preprocessed features, and calculating the correlation coefficient and covariance between any two different features based on the value of the features of the same type in the preprocessed features and the mean value; calculating the joint probability density function and the marginal probability density function of each of the different features, and calculating the mutual information between any two different features based on the joint probability density function and the marginal probability density function; Step S2-4: Calculate the degree of association between different pairs of features using a weighted average method: where a ij is the degree of association, i and j represent different features, w1, w2, w3 are weight coefficients, ρ ij is the correlation coefficient between i and j, is the covariance between i and j, MI ij is the mutual information between i and j; A correlation threshold is set, and the correlation degree with an absolute value greater than the correlation threshold is screened out. The two features corresponding to the correlation degree are used as the features of equipment defect correlation.

4. The method for identifying defects in power equipment based on multi-sensor data fusion according to claim 1, characterized in that: The step S3 further includes: Quantify the degree of defects in power equipment and establish a prediction model based on the degree of defects in power equipment and the characteristics associated with the defects in the equipment: D=β0+β1F1+β2F2+...+β m F m +∈ Where D represents the quantified degree of power equipment defects, β0 is the intercept term, β1, β2...β m is the regression coefficient, ∈ is the error term, F1, F2...F m is the value of the characteristic associated with the equipment defect when the current power equipment is defective; the regression coefficient is solved by the least square method; A model update cycle is set, and after a model update cycle, historical data within the model update cycle is integrated, and the regression coefficient of the prediction model is corrected based on the integrated data.

5. The method for identifying defects in power equipment based on multi-sensor data fusion according to claim 1, characterized in that: The step S4 further includes: pre-processing the real-time collected power data and extracting features to input into the prediction model to obtain the value of the degree of defect of the power equipment; grading the degree of defect, and issuing different levels of warnings according to the grading and the value of the degree of defect of the power equipment.

6. A power equipment defect identification system based on multi-sensor data fusion, which is applied to the power equipment defect identification method based on multi-sensor data fusion according to any one of claims 1 to 5, characterized in that: The system includes a monitoring deployment module, a data processing module, a model building module, and a real-time application module. The monitoring deployment module is used to deploy multiple types of sensors in the monitoring area and acquire power data. The data processing module preprocesses the power data, extracts features, and calculates correlations to screen out features associated with equipment defects. The model building module establishes a prediction model based on the features associated with equipment defects and regularly adjusts the regression coefficient based on historical data. The real-time application module processes the real-time collected power data and inputs it into the prediction model, triggering an early warning response based on the output power defect level. The output end of the monitoring deployment module is electrically connected to the input end of the data processing module; the output end of the data processing module is electrically connected to the input end of the model construction module; the output end of the model construction module is electrically connected to the input end of the implementation application module.

7. The power equipment defect identification system based on multi-sensor data fusion according to claim 6 is characterized in that: The monitoring deployment module includes a sensor deployment unit and a data acquisition unit; the sensor deployment unit is used to deploy multiple types of sensors in the monitoring area; and the data acquisition unit is used to acquire power data through the multiple types of sensors.

8. The power equipment defect identification system based on multi-sensor data fusion according to claim 6 is characterized by: The data processing module includes a data preprocessing unit and a feature association processing unit; the data preprocessing unit is used to remove abnormal values of power data, fill missing values and normalize the power data; The feature association processing unit is used to extract features of the power data and screen out features associated with equipment defects through correlation coefficient, covariance, mutual information and weighted average method.

9. The power equipment defect identification system based on multi-sensor data fusion according to claim 6 is characterized in that: The model construction module includes a model construction unit and a model update unit; the model construction unit is used to establish a prediction model based on the characteristics associated with equipment defects and the quantified degree of defects in the power equipment; the model update unit is used to set a model update cycle and correct the regression coefficient of the prediction model according to the set model update cycle.

10. The power equipment defect identification system based on multi-sensor data fusion according to claim 6, characterized in that: The real-time application module includes an input processing unit and an early warning response unit; the input processing unit is used to pre-process and extract features of the real-time collected power data; The early warning response unit is used to issue a graded early warning according to the output value of the degree of defect of the power equipment.

Images