Electric equipment line risk online monitoring method based on multi-source data fusion

Abstract

The invention discloses an electric equipment line risk online monitoring method based on multi-source data fusion, and particularly relates to the field of electric line monitoring, and the method comprises the steps: collecting the electric, insulation, physical and environmental data of a line according to a preset frequency, and carrying out the standardization processing to obtain corresponding standardization parameters; dividing the standardized parameters into input and output dimensions, calculating output dimension parameter weights by using an entropy weight method, and constructing a theoretical value model of each output dimension by using multiple linear regression; respectively constructing electrical, insulating and physical risk functions based on the standardized parameters and theoretical values; constructing a collaborative coupling matrix in combination with each risk function value and historical data, and calculating a total risk value through a nonlinear fusion logic construction function; defining a dynamic threshold value based on each risk value, judging an abnormal point when the threshold value is exceeded, and triggering iterative optimization of the model; the fault probability is calculated by combining the total risk value and the abnormal point data, and a three-level early warning mechanism is set to judge the line risk; according to the scheme, multi-dimensional risk collaborative monitoring is realized, and the risk identification accuracy and real-time performance are improved.

Description

Technical Field

[0001] This invention relates to the field of power line monitoring technology, and more specifically, to a method for online monitoring of power equipment and power line risks based on multi-source data fusion. Background Technology

[0002] In the safe and stable operation of power systems, real-time status monitoring of electrical equipment and lines is a core element in ensuring power supply reliability. Currently, line monitoring technology has become an important support for ensuring the stable operation of power systems, and a relatively mature application system has been formed in the sensing, transmission, and basic analysis stages. At the sensing level, the industry generally uses dedicated sensors to collect key line operating data, covering electrical parameters such as voltage, current, and active power, external influencing factors such as ambient temperature and humidity and micro-meteorological conditions, as well as indicators reflecting the physical state of the line, such as joint temperature. Furthermore, different acquisition frequencies can be adapted according to the line voltage level. In terms of data transmission, wireless and wired technologies are often combined to achieve real-time data transmission and meet the transmission needs of different laying scenarios. In the data processing stage, basic standardization methods are often used to eliminate the differences in the dimensions between parameters, and simple historical data comparison and threshold judgment are used to initially identify obvious abnormal conditions of the line, providing basic monitoring information for operation and maintenance personnel and assisting in carrying out daily line status verification and basic risk prediction.

[0003] However, it still has some drawbacks in practical use, such as: 1. The data collection dimension is singular, relying heavily on electrical parameters and failing to integrate key data from other dimensions. This leads to blind spots in identifying hidden risks such as line insulation aging and physical deformation, and cannot fully reflect the true operating status of the line. It is also prone to omissions due to incomplete data. 2. Risk assessment lacks a multi-dimensional collaborative mechanism, often calculating the risk of abnormality of a single parameter separately, ignoring the coupled effects of electrical overload, reduced insulation capacity, and physical deformation. For example, only monitoring current overload without correlating leakage current changes makes it impossible to capture high-risk scenarios with multiple factors superimposed, resulting in insufficient accuracy of assessment results. 3. The theoretical model is fixed and no dynamic iteration mechanism is set. The model parameters are not updated after being determined based on the initial samples. When the line operating environment changes, the deviation between the model and the actual operating conditions increases, the risk calculation error gradually accumulates, and the long-term monitoring reliability decreases. 4. The early warning mechanism is crude, often relying solely on a single threshold to judge risk without combining it with the quantitative classification of fault probability. For example, if an early warning is triggered only by current exceeding a threshold, it is impossible to distinguish between different levels of potential risks and minor fault risks, making it difficult for maintenance personnel to accurately determine the priority of handling, which can easily lead to over-maintenance or delayed response. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides an online monitoring method for electrical equipment line risks based on multi-source data fusion, which addresses the problems mentioned in the background art through the following solutions.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an online monitoring method for electrical equipment line risks based on multi-source data fusion, comprising: S1: Multi-source data acquisition and standardization processing: Collect line electrical data, insulation data, physical data and environmental data according to the preset frequency, and obtain the standardized parameters corresponding to each parameter after standardization processing; S2: Construction of the Dimensional Theoretical Model: Define the standardized parameters corresponding to each parameter as input and output dimensions, use the entropy weight method to calculate the weight of the standardized parameters in each output dimension, and use multiple linear regression to construct the theoretical value model of the standardized parameters in each output dimension. S3: Multidimensional Nonlinear Risk Calculation: Based on standardized parameter and theoretical value models, electrical risk functions, insulation risk functions, and physical risk functions are constructed respectively; S4: Cooperative Coupling Matrix Generation and Multi-Dimensional Risk Fusion: Based on the risk function values ​​of each dimension and historical data, a cooperative coupling matrix is ​​constructed; based on the function values ​​of electrical risk, insulation risk and physical risk and the cooperative coupling matrix, a multi-dimensional risk fusion function is constructed through nonlinear fusion logic to calculate the total risk value; S5: Dynamic Iterative Optimization of Theoretical Model: Based on the risk function values ​​of each dimension and the total risk value, a dynamic threshold is defined. When the risk value continuously exceeds the dynamic threshold, it is judged as an outlier. Outlier data is collected and model iteration is triggered. S6: Fault Probability Calculation and Graded Early Warning: Based on the total risk value and anomaly statistics, calculate the fault probability, set a three-level early warning mechanism based on the fault probability, and judge the line risk.

[0006] The technical effects and advantages of this invention are as follows: 1. Data collection covers multiple dimensions, simultaneously collecting electrical, insulation, physical, and environmental data. After standardization processing to eliminate dimensional differences, it comprehensively captures various characteristics of line operation, avoids missing hidden risks, and provides a complete data foundation for risk assessment. 2. Construct a multi-dimensional collaborative assessment system. By establishing theoretical models and constructing individual risk functions in different dimensions, and then combining them with a collaborative coupling matrix to integrate multiple risk values, the system fully considers the coupling effects of different risk dimensions, such as the superimposed effects of electrical overload and reduced insulation capacity, thereby improving the comprehensiveness and accuracy of risk assessment.

[0007] 3. It has the ability to dynamically iterate the theoretical model. Based on the abnormal data of risk values ​​exceeding the dynamic threshold, it triggers the update of model parameters. By adding normal samples, it refits the weights and regression model to ensure that the model always adapts to changes in line operating conditions, reduces long-term monitoring errors, and ensures monitoring reliability.

[0008] 4. Achieve fault probability quantification and graded early warning. Calculate fault probability by combining total risk value and anomaly data, divide into three levels of early warning according to probability, and output risk sources and handling suggestions. This allows maintenance personnel to clearly understand the risk level and response direction, accurately match handling resources, avoid over-maintenance or delayed response, and improve maintenance efficiency. Attached Figure Description

[0009] Figure 1 This is a schematic diagram of the overall structure of the present invention.

[0010] Figure 2 This is a schematic diagram of steps S1-S2 of the present invention.

[0011] Figure 3 This is a schematic diagram of steps S3-S4-S5 of the present invention.

[0012] Figure 4 This is a schematic diagram of step S6 of the present invention.

[0013] Figure 5 This is a scatter plot showing the relationship between the electrical risk value and the total risk value in this invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] refer to Figures 1-5 The online monitoring method for electrical equipment and line risks based on multi-source data fusion, as shown, includes: S1: Multi-source data acquisition and standardization processing: The system collects four core data categories in real time, reflecting the electrical characteristics, insulation performance, physical condition, and environmental impact of the power line, comprehensively covering key operational status information. Simultaneously, a standardized algorithm is employed to eliminate dimensional differences between different data points, mapping the raw data to a unified value range. The specific steps are as follows: S101: Multi-source data acquisition: According to the preset frequencies: 5Hz for low-voltage lines and 2Hz for medium-voltage lines, four types of data are collected from the lines. The specific collection methods and parameter definitions are as follows: Electrical data acquisition: Line voltage: The real-time operating voltage of the line is collected using a voltage sensor and denoted as U, with the unit being V; Line current: The real-time loop current of the line is collected using a current sensor and denoted as I, in A. Active power: The active power consumed by the line in real time is collected by a power sensor and is denoted as P, with the unit being kW; Power factor: Calculated based on the collected line voltage U, line current I, and active power P, denoted as: If the circuit is a three-phase circuit, then If it is a single-phase line, then .

[0016] Insulation data acquisition: Insulation resistance to ground: The insulation performance index between the line and the ground is collected using an insulation resistance sensor, i.e., the real-time insulation resistance between the line and ground, denoted as . ,unit: ; Leakage current: A leakage current sensor is used to collect the minute current flowing through abnormal paths in the line, i.e., the real-time leakage current of the line, denoted as . , Unit: mA.

[0017] Physical data acquisition: Joint contact temperature: The real-time temperature at the line joint caused by contact resistance is collected using a type K thermocouple, i.e., the line joint contact temperature, denoted as . Unit: ℃; Line vibration amplitude: The vibration amplitude of the line body caused by external interference or internal faults is collected by vibration sensors, i.e., the real-time vibration amplitude of the line, denoted as . Unit: g; Cable Deformation: This refers to the real-time cable deformation caused by external forces or thermal expansion and contraction of the flexible cable, measured using displacement sensors. It is only applicable to flexible cables and is denoted as [missing information]. , Unit: mm.

[0018] Environmental data collection: Ambient temperature: The real-time air temperature of the environment where the wiring is laid is collected using a temperature and humidity sensor and recorded as follows: Unit: ℃; Ambient humidity: Real-time relative humidity of the air in the environment where the line is laid is collected using temperature and humidity sensors and recorded as follows. Unit: %RH; Ambient dust concentration: The mass concentration of suspended dust in the ambient air near the installation site is collected using a dust sensor; this is the real-time ambient dust concentration, denoted as [missing information]. ,unit: .

[0019] All collected data is transmitted to the edge computing node via a hybrid transmission method of industrial Ethernet and LoRaWAN. The transmission process must meet the requirements of latency ≤100ms and packet loss rate ≤0.3% to ensure data real-time performance and integrity.

[0020] S102: Data Standardization: The min-max standardization method is used to map the collected raw data to the interval [0, 1], eliminating the interference of different data due to differences in units on subsequent calculations. A standardization parameter is defined, denoted as... The unique corresponding original parameter x has the following specific mathematical function: ; Where: x represents the original data, i.e., the various parameters defined in S101, including: U, I, P, , , , , , , , , ; The standardized parameter corresponding to the original parameter x; , These represent the statistical minimum and maximum values ​​of the original parameter in 1000 sets of samples from normal line operation, ensuring that the standardization results closely reflect the characteristics of normal line operation; output the standardized data after standardization calculation: , , , , , , , , , , , .

[0021] S2: Construction of a multi-dimensional theoretical model: Using the standardized data output from S1 as input, and combining it with the line fault mechanism, the four types of data—electrical, insulation, physical, and environmental—are clearly defined as input and output relationships. Environmental data is used as a risk influencing factor, while electrical, insulation, and physical data are used as risk representations of the line itself. Through a progressive logic of risk dimension definition, parameter weight quantification, and linear regression modeling, theoretical value models for each dimension's parameters are constructed. The reasonable value range of parameters under normal operating conditions is quantified, providing a core benchmark for subsequent risk identification through the deviation between actual and theoretical values. The specific steps are as follows: S201: Risk Dimension Definition and Parameter Weight Calculation: S2011: Risk Dimension Definition: Based on the functional attributes of data and the correlation with failures, the standardized data in S1 is reconstructed in terms of dimensions. The specific logic is as follows: Input dimension (environment dimension): Retain the environmental class standardized parameters from S1: environmental temperature standardized parameter. Standardized parameters of ambient humidity Standardized parameters for environmental dust concentration This dimension of data represents the external factors influencing line risk and serves as input to the model. Output Dimensions (Core Risk Dimensions): The three types of data in S1 that directly reflect the line's own state are defined as three core risk dimensions, i.e., risk representation dimensions at the model output. The standardized parameters included in each dimension and their correspondence with S1 are as follows: Electrical risk dimension: corresponds to S1 electrical category data, including standardized parameters of line voltage. Standardized parameters of line current Active power standardized parameters Power factor normalization parameter This directly reflects the stability of the electrical operation of the line; Insulation risk dimension: Corresponds to S1 insulation data, including standardized parameters of insulation resistance to ground. Leakage current standardized parameters This directly reflects the quality of the line insulation performance; Physical risk dimension: Corresponds to S1 physical data, including standardized parameters of joint contact temperature. Standardized parameters of line vibration amplitude Standardized parameters of cable deformation It directly reflects the integrity of the physical structure of the line; S2012: Entropy Weight Method for Quantifying Intra-Dimensional Parameter Weights: To differentiate the contributions of different parameters within the same risk dimension to the dimensional risk, the entropy weight method is used to calculate the weights of standardized parameters within each core risk dimension, defining the weight parameters. The physical meaning is a certain standardized parameter. The relative importance within the corresponding risk dimension, with a value range of [0, 1]; It should be further noted that all the standardized parameters mentioned are standardized data output by S1; It should be further noted that the weight quantification process is based on at least 1000 sets of sample data from normal line operation, and the specific steps are as follows: Parameter weighting calculation: For the j-th standardized parameter, calculate the proportion of its value in the i-th sample group to the total value of the parameter in all samples. Its specific mathematical function is: ,in Let j be the value of the standardized parameter in the i-th sample group. =0 =0; Information entropy calculation: based on parameter weighting Calculate the information entropy of the j-th parameter. The smaller the entropy value, the greater the difference in parameter values, and the stronger the ability to distinguish risks. Its specific mathematical function is: ,like =0, then define To avoid calculation errors; Weight calculation: Information entropy is converted into weights, and its specific mathematical function is as follows: Where m is the total number of standardized parameters within the corresponding core risk dimension, with m=4 for the electrical dimension, m=2 for the insulation dimension, and m=3 for the physical dimension, ensuring that the sum of the weights of all parameters within the same dimension is 1.

[0022] S202: Construction of a Multidimensional Theoretical Model The general expression of the model is as follows: The model is constructed using the multiple linear regression method. The core logic is to quantify the influence of environmental factors on the state of the railway line itself. , where: output quantity This is the theoretical value of a standardized parameter within the electrical, insulation, and physical risk dimensions. The physical meaning of this parameter is: under current environmental conditions, the parameter... The reasonable value under normal conditions uniquely corresponds to the standardized parameter. ; Input quantities: Three standardized parameters in the environmental dimension (Ambient temperature) (Ambient humidity) (Environmental dust concentration); Model parameters: , , These are the regression coefficients corresponding to the input quantities, reflecting the degree of influence of each environmental factor on the output quantities. The intercept is the model intercept. Regression coefficients for different standardized parameters are distinguished from the intercept by specific subscripts, i.e. ; All regression coefficients in the model , , With intercept All of them are obtained by using 1,000 sets of normal operation samples of the line in S2012. The goal is to minimize the sum of squared residuals between the actual and theoretical values ​​of the parameters. The least squares method is used to fit the solution. After the fitting is completed, the accuracy requirements must be met: under normal operating conditions, the average relative deviation between the actual and theoretical values ​​of all standardized parameters in the core risk dimension is ≤5%, and the maximum relative deviation is ≤10%, to ensure that the model can accurately represent the value rules of the parameters under normal operating conditions. Based on this model, using the three standardized parameters of the environmental dimension as unified inputs, theoretical value models for all standardized parameters within the core risk dimension can be constructed respectively, with the specific correspondence as follows: All were measured and standardized. , , For input; For the electrical risk dimension: calculations were performed using a general theoretical model formula. , , , Theoretical value , , , ; For the insulation risk dimension: calculations were performed using a general theoretical model formula. , Theoretical value , ; For the physical risk dimension: calculations were performed using a general theoretical model formula. , , Theoretical value , , .

[0023] S3: Multidimensional Nonlinear Risk Calculation Using the standardized data output from S1 and the theoretical model constructed from S2 as input, differentiated nonlinear risk functions are designed to address the fault characteristics of the three core risk dimensions: electrical, insulation, and physical. Risk values ​​for each dimension are calculated, with values ​​ranging from [0, 1]. Higher values ​​indicate higher risk. This approach accurately captures the nonlinear evolution patterns of different types of faults. The specific steps are as follows: S301: Constructing Electrical Risk Dimension Functions : It indicates the quantitative risk level of a line fault caused by abnormal electrical parameters, with two types of abnormal characteristics: core related current and power deviation and parameter change rate. The nonlinear risk function employs a nonlinear combination of Euclidean distance and an exponential function, and its specific mathematical function is as follows:

[0024] in: For the current standardization deviation, Power standardization deviation; The maximum real-time rate of change of the normalized parameters of current and power is given by the following mathematical function: t represents the current time, and t-1 represents the previous sampling time. The sampling period corresponds to the S1 sampling frequency: 5Hz for low-voltage lines and 2Hz for medium-voltage lines, used to capture sudden anomalies such as a sudden increase in current before a short circuit and a sudden change in power during overload. The safety threshold represents the upper limit of the safe rate of change of current and power, with a value of 0.8 A / s. It is set based on the overload protection standard for low-voltage lines, and can be adjusted to 0.5 A / s for medium-voltage lines. The function of the exponential function is as follows: when the rate of change is ≤ the safety threshold, the function value is ≤ 0.5, and the risk increases moderately; when the rate of change is ≥ 2 times the safety threshold, the function value approaches 1, and the risk increases sharply, which is consistent with the nonlinear characteristics of sudden faults. The range of values ​​for this function When the current and power have no deviation and the rate of change is 0, When all deviations are 1 and the rate of change is much greater than the threshold, .

[0025] S302: Constructing the Insulation Risk Dimension Function : It indicates the quantitative risk level of short circuits and leakage faults caused by the deterioration of insulation performance of the line, and is mainly associated with two types of characteristics: cumulative insulation deterioration and sudden leakage current.

[0026] This function employs a nonlinear combination of hyperbolic tangent and logistic regression, and its specific mathematical function is as follows:

[0027] in, This is the cumulative deviation parameter, representing the cumulative value of the standardized deviation of insulation resistance over 5 minutes (300 seconds). The standard deviation of insulation resistance is used to integrate the components to capture the cumulative effect of insulation material aging and avoid misjudgment due to short-term fluctuations. The hyperbolic tangent coefficient represents the rate of influence of cumulative deviation on risk. It is set based on statistics from 1000 insulation degradation samples and takes a value of 2. (Hyperbolic tangent function) The function value increases slowly when the cumulative deviation is small; when the cumulative deviation is ≥1.5, the function value approaches 1, reflecting the characteristic of risk saturation after accumulation to the critical point, thus avoiding the risk from increasing indefinitely. For leakage current deviation, The leakage current risk threshold corresponds to an actual leakage current ≥3mA, set based on low-voltage line leakage protection standards; logical staging function. The function is designed so that when the deviation is less than the threshold, the function value is approximately 0.5, which has little impact on the risk; when the deviation is greater than or equal to the threshold, the function value rapidly approaches 1, and the risk doubles, which is consistent with the sudden surge in risk after the leakage current exceeds the standard. is the step coefficient, representing the intensity coefficient of the step effect of leakage current deviation, with a value of 10; It needs to be normalized to [0, 1], and after normalization, it is still denoted as [0, 1]. .

[0028] S303: Constructing the Physical Risk Dimension Function : This function quantifies the risk level of line faults such as poor contact and line breakage caused by abnormal physical structure, and includes two types of characteristics: critical temperature of core associated joints and vibration-deformation coupling. The function employs a nonlinear combination of an S-shaped function and geometric mean, and its specific mathematical function is as follows:

[0029] in, For joint temperature deviation, The temperature risk threshold corresponds to an actual joint temperature ≥120℃, and is set based on the critical oxidation temperature of copper-aluminum joints. The critical coefficient represents the intensity coefficient of the critical effect of temperature deviation. It is set to 15 to ensure that the function value increases steeply near the threshold and avoid misjudgment of risk when the temperature is close to the critical point. For the standardization deviation of vibration amplitude, Standardized deviation for cable deformation To minimize the product term, which would be zero when both vibration and deformation are zero, thus rendering the coupled parameters meaningless. When the temperature is constant and vibration and deformation are both zero, No risk; when the temperature is supercritical and both vibration and deformation are 1, Extremely high risk.

[0030] S4: Cooperative Coupling Matrix Generation and Multi-Dimensional Risk Fusion Taking the electrical risk function value, insulation risk function value, and physical risk function value output by S3 as input, a synergistic coupling matrix is ​​generated through a dual-drive approach of mechanism and data to quantify the synergistic amplification effect between different risk dimensions. Then, a fusion formula is constructed by combining the dynamic risk contribution value to integrate the risk values ​​of the three dimensions into the total line risk value. This allows for accurate quantification of overall risks along the route, avoiding misjudgment of risks from a single dimension or omission of coordinated risks. The specific steps are as follows: S401: Co-coupling matrix generation : Matrix element definition: Define the cooperative coupling matrix C as a 3×3 matrix, corresponding to the electrical dimension E, the insulation dimension Ins, and the physical dimension Ph. Matrix elements... , , represents the coefficient of synergistic contribution to the total line failure probability when the risks of the i-th dimension and the j-th dimension coexist, with a value range of [0, 1]: like A value close to 1 indicates a very strong synergistic effect between the two risk dimensions. like A value close to 0.5 indicates that the risks in the two dimensions are simply superimposed, with no additional synergistic effect. like A value close to 0 indicates that the two dimensions of risk are unrelated. Matrix generation: matrix elements Generated through reproducible data statistics and mechanism correction algorithms, the specific steps are as follows: S4011: High-risk threshold calibration: Based on normal line operation data (sample size ≥ 1000 groups), calculate the distribution characteristics of risk values ​​for each dimension output by S3, and take the 99th percentile of the normal data as the high-risk threshold for each dimension. : High-risk threshold in electrical dimension When the electrical risk value exceeds this value, it is judged as high electrical risk; High risk threshold for insulation dimension When the insulation risk value exceeds this value, it is judged as a high-risk insulation condition. High risk threshold in physical dimension The physical meaning is that when the physical risk value exceeds this value, it is judged as high physical risk.

[0031] Threshold calibration logic: ensures that there is only a 1% probability of misjudging as high risk under normal operating conditions, providing a unified benchmark for subsequent data statistics.

[0032] S4012: Historical Data Statistics: Collect historical operation data of the lines (sample size ≥ 1000 sets, covering all scenarios including peak load, seasonal changes, and equipment aging). Each sample set must include risk values ​​for each dimension and whether a fault occurred during that period. A fault is defined as an event requiring line downtime for maintenance, denoted as F=1; no fault is denoted as F=0. Extract the following core statistics from the data: Risk value of the i-th dimension The number of samples, the number of high-risk samples in the i-th dimension; The i-th and j-th dimensions simultaneously satisfy the high-risk condition. and The number of samples, and the number of high-risk samples in both dimensions (ij). : The number of samples with high risk and failure in the i-th dimension, and the number of samples with high risk and failure in the i-th dimension; : The number of samples with high collaborative risk and failure, and the number of samples with high collaborative risk failure; S4013: Coupling Coefficient Calculation: For any two dimensions i and j, their coupling coefficient The specific mathematical function for calculating the conditional failure probability ratio is as follows:

[0033] Among them, molecules , represents the conditional failure probability when dimensions ij are simultaneously at high risk. Physically, it represents the probability of a line failure occurring in a scenario where both dimensions are simultaneously at high risk. Its calculation formula is: ,like Then take When there are no cooperating samples, it is tentatively defined as no synergistic effect; denominator , representing the maximum joint failure probability when i or j is individually at high risk, is based on the probability addition formula. Assuming that the two dimensions are calculated independently, its physical meaning is the maximum possible failure probability in a single-dimensional high-risk scenario. As a benchmark for comparing synergistic effects, its calculation formula is: ,in, , representing the probability of a conditional failure that is individually high-risk in the i-th dimension, if Then take ; Normalization coefficient , represents the correction value for constraining the coefficient to the interval [0, 1], and its calculation formula is:

[0034] Calculate the conditional failure probability ratios for all dimension combinations (E-Ins, E-Ph, Ins-Ph), and use the maximum value as the denominator to ensure the final... To avoid coefficients exceeding a reasonable range in certain scenarios; S4014: Matrix Filling and Symmetry Handling: Matrix elements are divided into diagonal elements (self-coupling) ) and off-diagonal elements (mutual coupling) The filling logic is as follows: diagonal elements : Characterizes the self-reinforcing effect when a single dimension of risk continues to rise, and is calculated as follows: ,in, refer to , ; refer to 0.5 is a fixed correction factor to ensure .

[0035] off-diagonal elements The result calculated in step A3 is directly adopted, and since the fault mechanisms of i cooperating with j and j cooperating with i are the same, the result is taken as... This simplifies calculations while ensuring logical consistency.

[0036] S402: Construct a multi-dimensional risk fusion function and calculate the total risk value. : The total risk value represents the quantification of the overall operational risk of the line, ranging from [0, 1]. A higher value indicates a higher probability of line failure, requiring priority for early warning and handling. The multi-dimensional risk fusion function uses a non-linear fusion logic that multiplies the dynamic risk contribution by the collaborative coupling matrix. Its specific mathematical function is as follows:

[0037] The dynamic risk contribution represents the relative importance of the risk value of the i-th dimension in the current total risk, which changes dynamically with the real-time risk value: if the risk value of a certain dimension is significantly higher than that of other dimensions, the contribution of that dimension will automatically increase without the need for preset weights. Normalization coefficient 1 / 3: Physically, it scales the result of the double summation to the range of [0, 1]. Since the maximum value of the double summation of the three dimensions is approximately 3, dividing by 3 ensures that the total risk value does not exceed the range.

[0038] It should be further explained that the experimental scenario designed in this experiment is to simulate the full operating conditions of electrical equipment circuits from normal operation to high-risk state, covering four typical scenarios: Low-risk scenarios: light line load, good insulation performance, stable physical structure, and minimal environmental interference; Single-dimensional risk increase scenario: Simulate the independent risks of increased electrical load, insulation aging, and slight deformation of physical structure respectively; Dual-dimensional risk superposition scenario: simulating the synergistic effect of two types of risks; Multi-dimensional risk superposition scenario: from two-dimensional superposition to three-dimensional high risk, until approaching the failure state; Based on the above experimental scenario, the risk values ​​were calculated using the experimental data and are presented in the following table:

[0039] S5 Theoretical Model Dynamic Iterative Optimization: To address the theoretical model deviations caused by aging losses, long-term environmental changes, and equipment parameter drift during line operation, the risk values ​​of each dimension output by S3 and the total risk value output by S4 are used as triggers. Through a closed-loop logic of anomaly point identification and model iterative updates, the multi-dimensional theoretical model constructed by S2 is dynamically optimized to ensure that the theoretical values ​​always match the real-time normal state of the line, avoiding risk misjudgments caused by static models and guaranteeing long-term monitoring accuracy. The specific steps are as follows: S501: Anomaly detection, iteration trigger condition: S5011: Dynamic Threshold Definition: Defines a dynamic threshold. This represents the anomaly judgment benchmark for a certain risk value at time t, which is dynamically updated according to the normal operating status of the line to avoid the problem that a fixed threshold cannot adapt to long-term changes. The calculation formula is: ,in: The moving average of the 100 normal risk values ​​before time t represents the average level of risk values ​​during the recent normal operation of the line. The moving standard deviation of the 100 normal risk values ​​before time t represents the dispersion of risk values ​​during the recent normal operation of the line; the coefficient 2.5 is set based on the normal distribution characteristics.

[0040] S5012: Risk Value Correlation and Threshold Correspondence: Dynamic thresholds must correspond one-to-one with the risk values ​​in S3 and S4 to avoid dimensional confusion. The specific correspondence is as follows: Electrical dimension dynamic threshold Corresponding electrical risk value Based on the first 100 normal groups calculate , ; Insulation dimension dynamic threshold Corresponding insulation risk value Based on the first 100 normal groups calculate , ; Physical dimension dynamic threshold Corresponding physical risk value Based on the first 100 normal groups calculate , ; Total risk dynamic threshold Corresponding total risk value Based on the first 100 normal groups calculate , ; .

[0041] S5013: Outlier Detection Rule: An outlier is identified and model iteration is triggered if any of the following conditions are met: A single-dimensional risk value exceeds the corresponding dynamic threshold: or or ; Total risk value exceeds total dynamic threshold: .

[0042] To avoid accidental triggering of iterations due to momentary interference, an abnormal point must occur in two consecutive sampling windows before a formal triggering occurs: Low voltage lines: sampling frequency 5Hz, sampling period Two consecutive windows, or 0.4 seconds; Medium voltage lines: sampling frequency 2Hz, sampling period Two consecutive windows, or 1 second.

[0043] S502: Model Iterative Update: S5021: Iteration Sample Preparation: After the iteration is triggered, newly added normal samples after the outliers need to be collected and filtered as the basic data for model updates. The requirements are as follows: Sample size: ≥500 groups, to ensure sufficient sample size to support regression fitting accuracy; Sample selection criteria: The risk value of new samples must meet the following requirements. , , Furthermore, there are no fault records on the line, ensuring that the sample is in normal operating condition; Sample content: Includes the original collected data and standardized data of S1, with the same sample structure as the modeling sample of S2.

[0044] S5022: Iterative optimization steps: Based on the newly added normal samples, repeat the entire modeling process of S2 to achieve dynamic updates of model parameters; S5023: Iteration Stop and Model Replacement: The iteration stops and the updated theoretical model is officially implemented when the following conditions are met: No outliers were found in three consecutive sampling windows. , , Continues to be valid; The updated model accuracy meets the requirement of average relative deviation ≤ 5%.

[0045] After the iteration stops, the updated theoretical model parameters will automatically replace the original model parameters. The subsequent risk calculation in S3 and risk fusion in S4 will be performed based on the new model, forming a closed loop of anomaly triggering, iterative optimization, and accuracy assurance.

[0046] S6: Fault Probability Calculation and Graded Early Warning: Total line risk value output by S4 The S5-determined anomaly statistics are used as input. The failure probability is calculated by coupling the total risk value and the anomaly frequency. Then, a three-level early warning threshold is set based on the failure probability. The output is a complete early warning information including the early warning level, risk source, handling suggestions, and failure location. The specific steps are as follows: S601: Failure Probability Calculation : Fault probability represents the quantified probability that a line will fail in the current hour and within the next hour, with a value range of [0, 1]. The closer the value is to 1, the higher the risk of line failure, requiring immediate intervention. The closer the value is to 0, the more stable the line operation is, requiring no additional intervention. The fault probability should reflect both the current overall risk level and the frequency of recent anomalies; its specific mathematical function is as follows: ,in, The abnormal frequency standardization value represents the normalized value of the number of abnormal points on the line within 10 minutes, quantifying the recent frequency of abnormalities. Its calculation formula is: ,in: The total number of anomalies identified by S5 within 10 minutes; The maximum possible number of anomalies within 10 minutes is calculated based on the S1 sampling frequency. For low-voltage lines, the frequency is 5Hz. 10 minutes = 600 seconds. The number of sampling times is 600 / 0.2 = 3000. Therefore... =3000; Medium voltage line 2Hz, sampling times in 10 minutes = 600 / 0.5 = 1200 times, therefore =1200, ensure ; 0.15 is the weighting coefficient for abnormal frequency, which represents the degree to which abnormal frequency contributes to the failure probability, and avoids abnormal frequency from dominating excessively. An upper limit constraint is used to ensure that the probability of failure does not exceed 1.

[0047] S602: Tiered Early Warning and Handling Recommendations: Based on Fault Probability The value range is determined, and a three-level early warning mechanism is established. The threshold range, risk characteristics, and handling suggestions for each level of early warning are all formulated based on engineering practice to ensure operability. The specific levels of early warning are as follows: Level 1 Warning: Threshold Range: Risk Characterization: A certain parameter of the line has a slight deviation, but has not reached the critical state of fault, and there is no risk of fault in the short term; Prompt: Identify the source of risk; Handling Suggestion: Recheck the equipment status of the corresponding risk dimension within 72 hours. No shutdown is required and it will not affect normal power supply.

[0048] Level 2 Warning: Threshold Range: Risk Characterization: There is a clear risk point in the line, which may cause a failure within 24 hours if not addressed; Warning Content: Locate the specific risk location and parameters; Action Recommendation: Arrange a shutdown inspection within 24 hours, replace the abnormal parts, and prevent the risk from escalating.

[0049] Level 3 Warning: Threshold Range: Risk Characterization: The line is experiencing a synergistic superposition of multiple risks, and is nearing a fault state. Serious faults such as short circuits and line breaks may occur at any time. Warning Content: Urgently locate the fault point and display the fault probability simultaneously. Handling Suggestions: Immediately shut down the power supply, activate the emergency repair plan, replace the faulty section of the line or core components, and after the repair is completed, conduct insulation tests and current-carrying tests. Power supply can only be restored after confirming that there is no risk.

[0050] S603: Warning Information Output Format: To adapt to engineering operation and maintenance scenarios, warning information must be output in multiple formats, including visual interfaces, audible and visual alarms, and SMS push notifications. The content includes: Basic information: warning time, line number, monitoring point location; Core risk factors: failure probability (F), early warning level, and dominant risk dimension; Data support: Comparison of actual and theoretical values ​​of abnormal parameters, and statistics on recent outliers; Handling Guidelines: Provide specific operating procedures and contact information for the responsible person.

[0051] Secondly: The accompanying drawings of the embodiments disclosed in this invention only involve the structures involved in the embodiments of this disclosure. Other structures can refer to the general design. In the absence of conflict, the same embodiment and different embodiments of this invention can be combined with each other. In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An online monitoring method for electric equipment line risk based on multi-source data fusion, characterized in that, Comprise: S1: multi-source data acquisition and standardization processing: according to the preset frequency acquisition line electrical data, insulation data, physical data and environmental data, after standardization processing, the standardization parameter corresponding to each parameter is obtained; S2: dimension theory model construction: the standardization parameter corresponding to each parameter is defined as input dimension and output dimension, the weight of the standardization parameter in each output dimension is calculated by entropy weight method, and the theoretical value model of the standardization parameter of each output dimension is constructed by using multiple linear regression; S3: dimension nonlinear risk calculation: based on the standardization parameter and the theoretical value model, electrical risk function, insulation risk function, physical risk function are constructed respectively; S4: collaborative coupling matrix generation and multi-dimensional risk fusion: based on the risk function value of each dimension and the historical data, the collaborative coupling matrix is constructed; Based on the function value of electrical risk, insulation risk and physical risk and the collaborative coupling matrix, a multi-dimensional risk fusion function is constructed by nonlinear fusion logic to calculate the total risk value; S5: dynamic iteration optimization of theoretical model: based on the risk function value of each dimension and the total risk value, a dynamic threshold is defined, when the risk value continuously exceeds the dynamic threshold, it is judged as an abnormal point, the abnormal point data is counted, and the model iteration is triggered; S6: fault probability calculation and hierarchical early warning: based on the total risk value and the abnormal point statistical data, the fault probability is calculated, the three-level early warning mechanism is set based on the fault probability, and the line risk is judged.

2. The method of claim 1, wherein the method is based on multi-source data fusion of electrical equipment line risk online monitoring. The multi-source data acquisition and standardization processing comprises: collecting four types of data of the line at a preset low-voltage line 5Hz and medium-voltage line 2Hz frequency: Electrical data: line voltage, line current, active power, and power factor calculated based on the line voltage, line current and active power; Insulation data: ground insulation resistance, leakage current; Physical data: joint contact temperature, line vibration amplitude, cable deformation; Environmental data: environmental temperature, environmental humidity, environmental dust concentration; Standardization processing: for all the above parameters, the min-max standardization method is adopted, at least 1000 groups of line normal operation samples are taken as the basis to eliminate the dimensional difference between the parameters, and the values of the parameters are uniformly adjusted to the range of 0-1, and each parameter corresponds to a standardized parameter.

3. The method of claim 2, wherein the method further comprises: The dimension theory model construction comprises: Dimension definition: the standardized parameters are divided into input dimension and output dimension, wherein the input dimension is the standardized result of environmental data, and the output dimension is divided into three types of risk-related standardized parameters, which correspond to the standardized results of electrical data, insulation data and physical data, i.e. electrical risk dimension, insulation risk dimension and physical risk dimension; Weight calculation: the entropy weight method is adopted to calculate the weight of each standardized parameter in the above three types of output dimensions based on a preset number of line normal operation samples; Model construction: the multivariate linear regression method is adopted to construct the theoretical value model for the three types of output dimension standardized parameters with the input dimension standardized parameters as the input, the parameters of the model are determined by fitting the normal operation samples, and finally the theoretical value corresponding to each output dimension standardized parameter is obtained.

4. The method of claim 1, wherein the method further comprises: The electrical risk function includes: Based on the standardized parameters of electrical data and the theoretical values ​​of electrical risk dimension output by the model, the theoretical values ​​corresponding to the standardized parameters of each electrical data are calculated; the deviation between the standardized parameter of each electrical data and its corresponding theoretical value is calculated; the sampling period is determined based on the preset acquisition frequency; the rate of change of the standardized parameters of current and active power within the sampling period is calculated; the deviation and the rate of change are combined nonlinearly with the cooperative deviation term and the burst amplification term to form the electrical risk function; and the electrical risk value is output, with a value range of [0, 1].

5. The method of claim 1, wherein: The insulation risk function includes: The standardized parameters of insulation data and the theoretical values ​​corresponding to the standardized parameters of insulation data output by the theoretical value model of insulation risk dimension are used as inputs. The deviations of the standardized parameters of insulation resistance and leakage current and their corresponding theoretical values ​​are calculated respectively. Based on the synergistic effect of the decrease in insulation resistance and the increase in leakage current on insulation performance, the two deviations are combined nonlinearly to form an insulation risk function, and the insulation risk function value is output with a value range of [0, 1].

6. The method of claim 1, wherein: The physical risk function includes: The theoretical values ​​of each physical standardized parameter output by the physical risk dimension model are taken as input. The deviations of the standardized parameters of joint temperature, vibration amplitude and cable deformation from their corresponding theoretical values ​​are calculated respectively. The three types of parameter anomalies are combined with the synergistic effect on the physical structure of the line. The above three types of deviations are combined nonlinearly to form a physical risk function and output the physical risk function value, which takes the value range of [0, 1].

7. The method of claim 1, wherein: The construction of the cooperative coupling matrix includes: Determine the foundation for construction: use electrical risk function values, insulation risk function values, physical risk function values, and historical line operation data containing fault records and normal records under each risk dimension combination as input; Define the matrix: Construct a 3×3 synergistic coupling matrix. Each element in the matrix corresponds to the synergistic contribution strength coefficient of the two risk dimensions. The coefficient value represents the degree to which anomalies in one risk dimension promote the failure of the other risk dimension. The generation process involves: setting high-risk thresholds for each risk dimension; statistically analyzing the frequency of failures occurring when a single dimension exceeds the threshold, two dimensions exceed the threshold simultaneously, or three dimensions exceed the threshold simultaneously in historical data; calculating the synergistic contribution strength coefficients for various risk combinations based on the frequency; and finally filling the coefficients into the corresponding positions in the matrix to complete the construction.

8. The method of claim 1, wherein the method further comprises: The construction of the multi-dimensional risk fusion function includes: Using electrical risk function values, insulation risk function values, physical risk function values, and a collaborative coupling matrix as inputs, and combining historical risk and fault correlation data of the line to verify the rationality of the function, the dynamic contribution of each risk value in the total risk is first calculated. Then, by multiplying the dynamic contribution of each risk value by the corresponding element in the collaborative coupling matrix and then by another type of risk value, the collaborative coupling relationship between the three types of risk values ​​and different risk dimensions is nonlinearly fused, and finally a multi-dimensional risk fusion function is formed, which outputs the total risk value.

9. The method of claim 1, wherein the method further comprises: The theoretical model is dynamically iteratively optimized, including: Based on the electrical risk function value, the insulation risk function value, the physical risk function value and the total risk value, a dynamic threshold is defined, which is determined based on the sliding mean and the sliding standard deviation of each risk value under the normal operation state of a preset number of lines; when any risk value continuously exceeds the corresponding dynamic threshold, an abnormal point is determined and abnormal point data is counted, at which time the model iteration is triggered, after collecting new normal operation samples of the line, the steps of entropy weight calculation and multivariate linear regression model fitting are repeated, the parameters of the theoretical value model of each output dimension are updated, and when the average relative deviation of the updated model is not more than 5%, the updated theoretical value model is enabled to realize dynamic optimization.

10. The method of claim 1, wherein: The fault probability calculation and hierarchical early warning includes: Based on the total risk value and the abnormal point statistical data, the abnormal point statistical data is standardized and combined with the total risk value of the line, the fault probability is obtained through nonlinear operation, and a three-level early warning mechanism is set based on the fault probability, wherein the fault probability in the interval of 0.3 to 0.5 corresponds to the first level of early warning, the fault probability in the interval of 0.5 to 0.8 corresponds to the second level of early warning, and the fault probability not less than 0.8 corresponds to the third level of early warning, and the early warning information including the early warning level, the risk source, the disposal suggestion and the fault positioning is output for different early warning levels to judge the risk state of the line.

Images