Power transmission line multi-source data intelligent fusion health diagnosis system

Abstract

This invention relates to the field of intelligent diagnostic technology for multi-source data of transmission lines, and more particularly to an intelligent fusion health diagnostic system for multi-source data of transmission lines. The system includes equipment and control devices. An acquisition module obtains raw monitoring data, load current data, and active power data of the transmission line. A compensation module generates a feature matrix using the load current and substitutes it into a thermal impedance model to calculate dynamic disturbances to correct the initial timestamp, outputting standardized time-series data. An indexing module constructs an adaptive elastic time window using the active power fluctuation rate, matching circulating current characteristics with temperature values ​​within the time window to output data correlation pairs. A verification module sets a tolerance threshold based on the dynamic thermal time constant and performs logical verification on the data correlation pairs to output fused data. A diagnostic module calls a graded early warning decision model based on the temperature rise rate characteristics to quantify the overheating risk probability and outputs the health status result. This invention solves the problems of correlation errors and model distortion caused by asynchronous timestamps of multi-source data.

Description

Technical Field

[0001] This invention relates to the field of intelligent diagnostic technology for multi-source data of transmission lines, and in particular to an intelligent fusion health diagnostic system for multi-source data of transmission lines. Background Technology

[0002] Intelligent fusion health diagnosis of transmission lines using multi-source data is a technical method for automatically assessing the operating status of transmission lines by integrating data from multiple monitoring devices such as circulating current sensors and temperature sensors. This technology relies on data fusion algorithms to match and align time-series data from different sources on a unified time reference to construct a comprehensive data model that accurately reflects the correlation between current and temperature. The fused data supports the extraction of key features such as temperature rise rate and enables fault risk identification based on a hierarchical early warning model, thereby supporting condition-based maintenance decisions.

[0003] Existing real-time fusion health diagnosis technologies for multi-source heterogeneous monitoring data in new energy transmission lines suffer from the following technical challenges: Firstly, multi-source monitoring devices, such as circulating current sensors and temperature sensors, have independent clock sources. In the harsh marine environment of offshore wind power transmission lines, clock drift is unavoidable, leading to microsecond-level differences in the collected timestamps. This timestamp asynchrony causes fundamental correlation errors during data fusion, resulting in the system incorrectly matching current and temperature data from different time points. These correlation errors distort the true correspondence between current and temperature, causing inaccurate calculations of key features such as the temperature rise rate. For example, when the circulating current data timestamp is several microseconds later than the actual temperature data timestamp, high-current periods may be incorrectly correlated with low-temperature periods, thus underestimating the risk of joint temperature rise, generating a distorted health status model, and delaying fault warnings. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a multi-source data intelligent fusion health diagnosis system for transmission lines. This invention solves the technical problem that the basic correlation error during data fusion is caused by the microsecond-level synchronization of timestamps of multi-source monitoring data, resulting in a distorted health status model.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows: The intelligent fusion health diagnosis system for multi-source data of transmission lines provided by this invention includes equipment and a control device. The control device establishes a communication connection with the equipment and includes: The acquisition module is used to acquire the original monitoring data, load current data and active power data corresponding to the transmission line using the equipment. The original monitoring data is accompanied by an initial timestamp. The compensation module is used to generate a feature matrix using the load current data, substitute the feature matrix into the thermal impedance model to output conductor heat generation and temperature gradient distribution data, perform integral calculations using the temperature gradient distribution data and the preset fiber thermo-optic coefficient to derive the corresponding dynamic disturbance amount inside the communication cable, and use the dynamic disturbance amount to add or subtract corrections to the initial timestamp to output standardized time series data. The index module is used to calculate the first derivative corresponding to the active power data to determine the instantaneous fluctuation rate of active power, use the instantaneous fluctuation rate of active power as the independent variable to construct a decay function output time window, and match and combine the circulation feature vector included in the standardized time series data with the temperature value within the time window to output data association pairs. The verification module is used to extract the time of occurrence of the peak current and the time of temperature change in the data association pair, subtract the time of occurrence of the peak current from the time of temperature change to determine the time difference, calculate the dynamic thermal time constant using the metal material parameters, use the dynamic thermal time constant as the tolerance threshold, and retain the data whose time difference is within the tolerance threshold as the fused data. The diagnostic module is used to extract temperature rise rate features from the fused data, input the temperature rise rate features into a pre-trained graded early warning decision model to output overheating risk probability, and use the overheating risk probability to match the corresponding health level to output the line health status model result.

[0006] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the equipment includes a circulating current monitoring device and a temperature monitoring device; The control device includes a clock module; The clock module sends a synchronization message with a base timestamp to the circulation monitoring device and the temperature monitoring device. The circulation monitoring device and the temperature monitoring device receive the synchronization message and collect corresponding environmental parameters starting from the base timestamp, then output the raw monitoring data.

[0007] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the compensation module includes an extraction sub-module; The extraction submodule is used to extract the effective value of the current and harmonic components of the cable conductor corresponding to the transmission line within a set period from the load current data, and to splice the effective value of the current and the harmonic components to output the feature matrix.

[0008] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the compensation module includes a derivation submodule; The derivation submodule is used to substitute the feature matrix into the thermal impedance model to output the corresponding conductor heat generation, use the thermodynamic conduction formula to calculate the temperature gradient distribution data corresponding to the heat generation of the conductor transferred to the communication cable layer, integrate the temperature gradient distribution data and the preset optical fiber thermo-optic coefficient to output the refractive index deformation value, and use the refractive index deformation value as the dynamic disturbance quantity.

[0009] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the index module includes a time window sub-module; The time window module is used to extract the maximum and minimum active power values ​​within a set time period from the active power data, subtract the minimum active power value from the maximum active power value and divide by the corresponding duration of the set time period to output the first derivative, determine the instantaneous fluctuation rate of active power from the first derivative, input the instantaneous fluctuation rate of active power into the decay function, the decay function is a preset negative exponential decay function formula, output the corresponding window time length, and determine the window time length as the time window.

[0010] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the index module includes a binding sub-module; The binding submodule is used to extract the initial timestamp, which has been corrected by addition and subtraction, carried in the standardized time series data as the primary key, construct a time hash index table using the initial timestamp, and retrieve the circulation feature vector and temperature value corresponding to the same initial timestamp using the time hash index table within the time length defined by the time window. The retrieved circulation feature vector and temperature value are then packaged, stored, and output as the data association pair.

[0011] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the verification module includes a capture submodule; The capture submodule is used to receive the data association pair, perform wavelet transform decomposition on the circulating current feature vector included in the data association pair to output current components corresponding to different frequency bands, filter out the current peak data corresponding to the current exceeding the preset amplitude threshold from the current components, and record the current peak data at the corresponding time node in the standardized time series data as the time when the current peak occurs.

[0012] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the verification module includes a calculation submodule; The calculation submodule is used to connect to a preset material database, retrieve the specific heat capacity value and mass value of the metal material corresponding to the transmission line from the preset material database as the metal material parameter, multiply the specific heat capacity value by the mass value and divide by a preset heat dissipation coefficient to output the dynamic thermal time constant.

[0013] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the verification module includes a discrimination submodule; The discrimination submodule is used to calculate the first derivative of the temperature value changing with time within the data association pair, find the time node corresponding to the first derivative exceeding a preset rate of change threshold to determine the temperature change moment, subtract the current peak occurrence moment from the temperature change moment to output the time difference, compare the time difference with the tolerance threshold, delete the corresponding data association pair when the time difference is greater than the tolerance threshold, and output the corresponding data association pair as the fused data when the time difference is less than or equal to the tolerance threshold.

[0014] Furthermore, in the power transmission line data fusion health diagnosis system of the present invention, the diagnosis module includes an output sub-module; The output submodule is used to read the temperature values ​​included in the fused data, calculate the temperature difference between adjacent time nodes and output a temperature gradient matrix, input the temperature gradient matrix into a principal component analysis algorithm model for dimensionality reduction and output principal component vectors to determine the temperature rise rate feature, input the temperature rise rate feature into the hierarchical early warning decision model including a multi-layer decision tree structure for node probability calculation and output the overheating risk probability, and search a pre-set probability interval mapping table to obtain the health level name to which the overheating risk probability belongs and output it as the result of the line health status model.

[0015] Beneficial effects of this invention: The intelligent fusion health diagnosis system for multi-source data of transmission lines provided by this invention has significant technological advancements and beneficial effects. The compensation module generates a feature matrix using load current data and substitutes it into a thermal impedance model. By calculating the conductor's heat generation and the corresponding temperature gradient distribution data within the communication cable layer, it derives the refractive index deformation value caused by the Joule heating effect. This allows for microsecond-level dynamic feedforward compensation correction of the initial timestamp attached to the original monitoring data, eliminating the deviation in standardized time-series data caused by clock drift and thermal delay in the harsh marine environment of offshore wind power transmission lines. The indexing module calculates the first derivative corresponding to the active power data to determine the instantaneous fluctuation rate of active power and constructs an adaptive elastic time window. This enables the system to shrink the matching interval during transient processes of drastic power fluctuations, achieving precise matching and combination of circulating current feature vectors and temperature values ​​within the standardized time-series data. The verification module extracts the peak current occurrence time and temperature abrupt change time of the data association pair, sets a tolerance threshold using the dynamic thermal time constant calculated from the metal material parameters, and performs a physical consistency judgment. This effectively eliminates physical violation data points caused by residual synchronization errors, forming high-confidence fused data. The diagnostic module extracts temperature rise rate features based on fused data and quantifies the probability of overheating risk. This solves the underlying correlation error caused by microsecond-level time stamp asynchrony of multi-source monitoring data, avoids distortion of the health status model, and improves the reliability of fault early warning for new energy transmission lines under complex operating conditions. Attached Figure Description

[0016] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0017] Figure 1 This is a system architecture diagram of the intelligent fusion health diagnosis system for multi-source data of power transmission lines according to the present invention. Detailed Implementation

[0018] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. 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. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0019] Please see Figure 1The present invention provides a multi-source data intelligent fusion health diagnosis system for power transmission lines, comprising equipment and a control device. The control device establishes a communication connection with the equipment and includes: The acquisition module is used to acquire the original monitoring data, load current data and active power data corresponding to the transmission line using the equipment. The original monitoring data is accompanied by an initial timestamp. The compensation module is used to generate a feature matrix using the load current data, substitute the feature matrix into the thermal impedance model to output conductor heat generation and temperature gradient distribution data, perform integral calculations using the temperature gradient distribution data and the preset fiber thermo-optic coefficient to derive the corresponding dynamic disturbance amount inside the communication cable, and use the dynamic disturbance amount to add or subtract corrections to the initial timestamp to output standardized time series data. The index module is used to calculate the first derivative corresponding to the active power data to determine the instantaneous fluctuation rate of active power, use the instantaneous fluctuation rate of active power as the independent variable to construct a decay function output time window, and match and combine the circulation feature vector included in the standardized time series data with the temperature value within the time window to output data association pairs. The verification module is used to extract the time of occurrence of the peak current and the time of temperature change in the data association pair, subtract the time of occurrence of the peak current from the time of temperature change to determine the time difference, calculate the dynamic thermal time constant using the metal material parameters, use the dynamic thermal time constant as the tolerance threshold, and retain the data whose time difference is within the tolerance threshold as the fused data. The diagnostic module is used to extract temperature rise rate features from the fused data, input the temperature rise rate features into a pre-trained graded early warning decision model to output overheating risk probability, and use the overheating risk probability to match the corresponding health level to output the line health status model result.

[0020] The transmission line data fusion health diagnosis system includes equipment and control devices. The control device establishes a communication connection with the equipment and includes an acquisition module, a compensation module, an indexing module, a verification module, and a diagnostic module. The acquisition module uses the equipment to obtain the original monitoring data, load current data, and active power data corresponding to the transmission line. The original monitoring data refers to the physical environment readings directly acquired by the sensor's underlying hardware, accompanied by an initial timestamp. The load current data represents the main current parameters flowing inside the line conductors, and the active power data reflects the actual effective electrical energy parameters transmitted from the wind farm to the grid. The compensation module uses the load current data to generate a feature matrix including multi-dimensional current characteristics. The compensation module uses this feature matrix to input into a thermal impedance model to output the corresponding dynamic disturbance quantity inside the communication cable. The thermal impedance model reflects the heat transfer resistance characteristics inside the cable structure, and the dynamic disturbance quantity represents the nonlinear microsecond-level deviation caused by temperature changes in optical cable communication delay. The compensation module uses the dynamic disturbance quantity to add or subtract corrections to the initial timestamp to output standardized time-series data. The standardized time-series data, after compensation calculations, eliminates time drift errors caused by the physical environment. The indexing module calculates the first derivative of the active power data to determine the instantaneous fluctuation rate of active power. Using this fluctuation rate as an independent variable, it constructs a decay function and outputs a time window. Within this window, the indexing module matches and combines the circulating current feature vectors and temperature values ​​included in the standardized time-series data to output data association pairs. The verification module extracts the peak current occurrence time and temperature abrupt change time within the data association pairs. It subtracts the peak current occurrence time from the temperature abrupt change time to determine the time difference. The verification module calculates the dynamic thermal time constant using metal material parameters and uses this dynamic thermal time constant as a tolerance threshold, retaining data with time differences within this threshold as fused data. The diagnostic module extracts temperature rise rate features from the fused data. It inputs these features into a pre-trained hierarchical early warning decision model to output an overheating risk probability. In the real-time monitoring and diagnostic scenario of offshore wind power transmission lines, the diagnostic module uses the overheating risk probability to match the corresponding health level and outputs the line health status model result.

[0021] To meet the physical acquisition architecture requirements of the underlying monitoring data source, the power transmission line data fusion health diagnosis system divides the equipment into lower-level structures. The equipment includes circulating current monitoring devices and temperature monitoring devices. The circulating current monitoring device is installed at the grounding location of the cable's metal sheath to capture grounding circulating current signals, while the temperature monitoring device is attached to the surface of the cable joint to obtain parameters related to surface heat changes. The control device includes a clock module, which acts as the source of a unified time reference for the entire network, initiating synchronization. The clock module sends synchronization messages with a basic timestamp to the circulating current monitoring device and the temperature monitoring device. The underlying hardware nodes parse the synchronization messages, align them with the local oscillator, and complete the conversion from analog signals in the physical world to digital signals with time stamps. The circulating current monitoring device and the temperature monitoring device receive the synchronization messages and, starting from the basic timestamp, collect corresponding environmental parameters and output raw monitoring data.

[0022] The compensation module includes an extraction submodule to extract multi-dimensional electrical parameters. This submodule extracts the RMS current value and harmonic components of the corresponding cable conductors within a set period from the load current data. The RMS current value reflects the dominant factor in the steady-state heating of the transmission system, while the harmonic components represent the high-frequency oscillations and additional heating interference generated by the converter equipment. The extraction submodule performs a data splicing operation to transform the original single-dimensional scalar data into a structured matrix format. This structured matrix format provides boundary input conditions for calculating the evolution of complex spatial thermal fields. The extraction submodule then splices the RMS current value and harmonic components to output a feature matrix.

[0023] The compensation module includes a derivation submodule to establish the electrothermal multiphysics coupling mapping relationship. This submodule substitutes the characteristic matrix into the thermal impedance model to output the corresponding conductor heat generation. It then uses thermodynamic conduction formulas to calculate the temperature gradient distribution data corresponding to the heat transfer from the conductor to the communication cable layer. This temperature gradient distribution data depicts the spatial attenuation surface morphology of heat propagation from the inner conductor to the outer optical cable layer. The derivation submodule integrates the temperature gradient distribution data and the preset fiber thermo-optic coefficient to output the refractive index deformation value. The preset fiber thermo-optic coefficient represents a constant physical property change in quartz material due to heating. In high-load offshore wind power transmission scenarios, the current heating causes a slowdown in the speed of light in the communication link, resulting in a timestamp shift effect, which is transformed into a specific numerical variable. The derivation submodule uses the refractive index deformation value as a dynamic disturbance.

[0024] The index module is configured with a time window submodule that executes business-oriented dynamic relational spatial calculation logic. The time window module extracts the maximum and minimum active power values ​​within a set time period from the active power data. It then subtracts the minimum active power value from the maximum and divides the result by the corresponding duration of the set time period, outputting the first derivative. This first derivative is determined as the instantaneous fluctuation rate of active power. The time window module inputs this instantaneous fluctuation rate into a decay function, which uses a preset negative exponential decay function formula. In business scenarios where frequent grid dispatch adjustments cause severe power fluctuations, the time window module calculates the window time length corresponding to the minimum value. In scenarios with stable power output, the time window module calculates the window time length corresponding to the larger value. This window time length is then determined as the time window.

[0025] The indexing module includes a binding submodule that executes the underlying heterogeneous data synchronization and association logic. The binding submodule extracts the initial timestamp, after addition and subtraction correction, from the standardized time-series data as the primary key. It then constructs a time hash index table using this corrected initial timestamp. This time hash index table facilitates the evolution of heterogeneous monitoring data from a discrete, disordered state to a bound state at the same physical occurrence time. Within the time window defined by the time period, the binding submodule uses the time hash index table to retrieve the circulation feature vector and temperature value corresponding to the same initial timestamp. Since the circulation feature vector and temperature value are originally discretely distributed across different database nodes, the binding submodule encapsulates and stores the retrieved circulation feature vector and temperature value, outputting a data association pair.

[0026] The verification module incorporates a capture submodule for high-frequency feature signal layer separation and analysis. The capture submodule receives data association pairs and performs wavelet transform decomposition on the circulating current feature vectors included in the data association pairs, outputting current components corresponding to different frequency bands. Wavelet transform decomposition simultaneously locates abrupt signal components in both the time and frequency domains, thus positioning continuous waveform slices as discrete time anchors. The capture submodule filters out current peak data exceeding a preset amplitude threshold from the current components. In business application scenarios where wind turbine switching operating states generates transient impacts, the current peak data undergoes significant abrupt changes. The capture submodule records the corresponding time node of the current peak data within the standardized time-series data as the time of current peak occurrence.

[0027] The verification module loads the calculation submodule to introduce objective constraint parameters of the line's physical structure. The calculation submodule connects to a preset material database, which stores basic engineering design parameters for mainstream submarine cable joint models. The calculation submodule retrieves the specific heat capacity and mass values ​​of the corresponding metal materials for the transmission line from the database as metal material parameters. The calculation submodule multiplies the specific heat capacity by the mass value and divides it by a preset heat dissipation coefficient to output a dynamic thermal time constant. The heat dissipation coefficient reflects the cooling capacity of the external water-cooled insulation environment of the joint. The calculation submodule uses the dynamic thermal time constant to quantify the inherent property of the thermal response delay of a specific physical material to severe current impacts, thereby changing the judgment criteria for subsequent verification data.

[0028] The verification module includes a logical falsification mechanism to distinguish the execution time differences of the sub-modules. The sub-module calculates the first derivative of the temperature value within the data association pair over time. It identifies time points where the first derivative exceeds a preset rate of change threshold as moments of temperature abrupt change. The sub-module subtracts the current peak occurrence time from the temperature abrupt change time and outputs the time difference. The sub-module compares the time difference with a tolerance threshold. If the time difference is greater than the tolerance threshold, the sub-module executes an objective common-sense judgment procedure to eliminate false data associations caused by residual errors and deletes the corresponding data association pair. If the time difference is less than or equal to the tolerance threshold, the sub-module outputs the corresponding data association pair as fused data.

[0029] The diagnostic module comprises an output submodule that handles model dimensionality reduction and business decision generation. The output submodule reads temperature values ​​from the fused data, calculates the temperature difference between adjacent time points to output a temperature gradient matrix, and inputs this matrix into a principal component analysis (PCA) algorithm model for dimensionality reduction, outputting principal component vectors as the temperature rise rate feature. The PCA algorithm model uses orthogonal transformations to compress multidimensional redundant heat accumulation variables into clearly directional core feature parameters. The output submodule inputs the temperature rise rate feature into a hierarchical early warning decision model with a multi-layered decision tree structure to calculate node probabilities and output the overheating risk probability. The multi-layered decision tree structure uses information entropy gain to progressively remove non-fault factors. The output submodule then searches a pre-defined probability interval mapping table to obtain the corresponding health level name for the overheating risk probability and outputs it as the result of the line health status model.

[0030] The data fusion health diagnostic system for transmission lines is applied to the monitoring of offshore wind power transmission lines. In the harsh marine environment, independent clock sources often experience microsecond-level deviations, leading to fundamental correlation errors. The system includes equipment and a control unit, with the control unit establishing a communication connection with the equipment. The equipment includes circulating current monitoring devices and temperature monitoring devices. The circulating current monitoring device is installed at the grounding point of the cable's metal sheath to capture grounding circulating current signals, while the temperature monitoring device is attached to the surface of the cable joint to obtain parameters related to changes in surface heat. The control unit includes a clock module, which acts as the unified time reference source for the entire network, initiating synchronization. The clock module sends synchronization messages with a basic timestamp to the circulating current monitoring device and the temperature monitoring device. The circulating current monitoring device and the temperature monitoring device receive the synchronization messages and collect corresponding environmental parameters starting from the basic timestamp, outputting raw monitoring data. The control unit also includes an acquisition module, which uses the equipment to acquire the corresponding raw monitoring data, load current data, and active power data of the transmission line. The raw monitoring data includes an initial timestamp.

[0031] To address the service pain point of timestamp delay caused by the slowing of the speed of light in communication links due to current heating, the control device is equipped with a compensation module that performs physical-level error correction logic. This compensation module includes an extraction submodule to extract multi-dimensional electrical parameters. The extraction submodule extracts the effective current value and harmonic components of the corresponding cable conductors within a set period from the load current data. The effective current value reflects the dominant factor in the steady-state heating of the transmission system, while the harmonic components represent the high-frequency oscillations and additional heating interference generated by the converter equipment. The extraction submodule concatenates the effective current value and harmonic components to output a feature matrix. The compensation module also includes a derivation submodule, which establishes a multi-physics coupling mapping relationship between electrothermal fields. This derivation submodule substitutes the feature matrix into a thermal impedance model to output the corresponding conductor heat generation. The derivation submodule uses thermodynamic conduction formulas to calculate the temperature gradient distribution data corresponding to the heat transfer from the conductor to the communication cable layer. This temperature gradient distribution data depicts the spatial attenuation surface morphology of heat propagation from the inner conductor to the outer optical cable layer. The derivation submodule integrates the temperature gradient distribution data and the preset fiber thermo-optic coefficient to output the refractive index deformation value. The derivation submodule uses the refractive index deformation value as a dynamic disturbance, and the compensation module uses the dynamic disturbance to add or subtract the initial timestamp to correct and output standardized time series data.

[0032] To address business scenarios where frequent grid dispatch adjustments lead to significant power fluctuations, the control device incorporates an index module to establish a synchronous association mapping of underlying heterogeneous data. This index module includes a time window submodule that executes business-oriented dynamic association space calculation logic. The time window submodule extracts the maximum and minimum active power values ​​within a set time period from the active power data. It then subtracts the minimum active power value from the maximum and divides the result by the corresponding duration of the set time period, outputting a first derivative. This first derivative is then used as the instantaneous active power volatility. The time window module inputs this instantaneous volatility into a decay function, which is a preset negative exponential decay function formula. It outputs the corresponding window time length, which is then defined as the time window. Internally, the index module includes a binding submodule that executes the underlying heterogeneous data synchronous association logic. This binding submodule extracts the initial timestamp, after adjustments, from the standardized time-series data as the primary key. It then uses this adjusted initial timestamp to construct a time hash index table. Within the time window, the binding submodule uses the time hash index table to retrieve the circulation feature vector and temperature value corresponding to the same initial timestamp. The binding submodule then encapsulates and stores the retrieved circulation feature vector and temperature value to output the data association pair.

[0033] The control device incorporates a verification module that introduces a logical falsification mechanism based on the time differences in the execution of objective constraint parameters of the line's physical structure. The verification module includes a capture submodule for high-frequency characteristic signal layer separation and analysis. The capture submodule receives data association pairs and performs wavelet transform decomposition on the circulating current feature vectors within these pairs, outputting current components corresponding to different frequency bands. It then filters out current peak data exceeding a preset amplitude threshold from these current components and records the corresponding time node within the standardized time-series data as the current peak occurrence time. The verification module also integrates a calculation submodule connected to a preset material database. This database stores basic engineering design parameters for mainstream submarine cable joint models. The calculation submodule retrieves the specific heat capacity and mass values ​​of the corresponding metal materials for the transmission line from the preset material database as metal material parameters. The calculation submodule multiplies the specific heat capacity value by the mass value and divides it by a preset heat dissipation coefficient to output a dynamic thermal time constant. The verification module also includes a discrimination submodule, which calculates the first derivative of the temperature value within the data association pair over time. The discrimination submodule identifies the time node where the first derivative exceeds a preset rate of change threshold as the moment of temperature change. The discrimination submodule subtracts the moment of temperature change from the moment of current peak occurrence to output the time difference. The discrimination submodule compares the time difference with a tolerance threshold. The dynamic thermal time constant is used as the tolerance threshold to quantify the inherent property of the thermal response delay of a specific physical material to a severe current impact. When the time difference is greater than the tolerance threshold, the discrimination submodule executes an objective common sense judgment procedure to eliminate false data associations caused by residual errors and deletes the corresponding data association pair. When the time difference is less than or equal to the tolerance threshold, the discrimination submodule outputs the corresponding data association pair as fused data.

[0034] The control device's diagnostic module undertakes model dimensionality reduction and business decision generation tasks. This module includes an output submodule. The output submodule reads temperature values ​​from the fused data and calculates the temperature difference between adjacent time points, outputting a temperature gradient matrix. The output submodule then inputs the temperature gradient matrix into a principal component analysis (PCA) algorithm model for dimensionality reduction, outputting principal component vectors as the temperature rise rate feature. The PCA algorithm model uses orthogonal transformation to compress multidimensional redundant heat accumulation variables into clearly directional core feature parameters. The output submodule then inputs the temperature rise rate feature into a hierarchical early warning decision model with a multi-layered decision tree structure for node probability calculation, outputting the overheating risk probability. The multi-layered decision tree structure uses information entropy gain to progressively remove non-fault factors. Finally, the output submodule searches a pre-defined probability interval mapping table to obtain the corresponding health level name for the overheating risk probability and outputs it as the line health status model result.

[0035] The raw monitoring data represents the conversion results of the underlying analog electrical signals read by the sensing nodes deployed on the submarine fiber optic composite cable before digital smoothing filtering. The raw monitoring data is packaged with an initial timestamp and physical environment readings. Specifically, the physical environment readings include the absolute temperature value at the cable joint sheath and the microampere-level circulating current amplitude induced at the metal sheath grounding point. Upon receiving a synchronization command from the control device, the independent crystal oscillator within the underlying hardware node latches the current counter state. The underlying hardware node then appends the latched value as the initial timestamp to the header of the physical environment reading data frame to form a complete data packet.

[0036] Load current data depicts the set of charge transfer parameters actually flowing through the main core of a three-phase high-voltage AC transmission line. The acquisition module extracts waveform sampling points from multiple consecutive power frequency cycles and performs discrete Fourier transform operations. From these sampling points, the acquisition module extracts the effective value of the 50 Hz fundamental current, reflecting the steady-state energy transmission. It also extracts the amplitude of high-frequency harmonic components excited by the switching action of the high-power converter in offshore wind power. The extraction submodule concatenates the effective value of the 50 Hz fundamental current and the amplitude of the high-frequency harmonic components into a two-dimensional array structure, prioritizing column order. This two-dimensional array structure is defined as a feature matrix, thus elevating the scattered electrical parameters to a level suitable for the partial differential equations of the spatial thermal field.

[0037] Active power data maps to the instantaneous megawatt-level electrical energy parameters of all units in a remote offshore wind farm that are connected to the grid and actually used for work. The grid dispatch center frequently issues active power control commands to the wind farm based on wind condition changes, causing sharp increases or significant transient decreases in grid-connected electrical energy parameters. The control device records the maximum and minimum active power values ​​within each second of the time slice. The time window module performs differential calculations to calculate the difference between the maximum and minimum active power values. The time window module divides this difference by a fixed time slice length to obtain the first derivative, which is then used as the instantaneous active power volatility to quantify the degree of abrupt changes in the wind farm's energy output.

[0038] The thermal impedance model effectively simulates the physical hindrance effect of the cross-linked polyethylene insulation layer and water-blocking layer inside the submarine optical cable on the outward diffusion of Joule heating from the internal core. The derivation submodule expands the conductor's heat generation along the radial coordinate to solve the partial differential equation of heat conduction, obtaining temperature gradient distribution data. Based on this temperature gradient distribution data and a preset optical fiber thermo-optic coefficient, the derivation submodule performs path-by-path integration calculations. The derivation submodule then visualizes the results of the integration calculations as dynamic disturbances, which characterize the deviation in the arrival time of the synchronization optical pulse caused by the slowing of the speed of light inside the communication optical fiber due to high temperatures.

[0039] A pre-defined negative exponential decay function formula is used to construct a nonlinear mapping relationship to balance the fidelity of transient feature capture with the computational power consumption of system addressing. The exponent base is generated by engineering experiments, and the exponent power is taken as the negative form of the instantaneous fluctuation rate of active power. The time window submodule takes the instantaneous fluctuation rate of active power as input, calculates it using the pre-defined negative exponential decay function formula, and outputs the corresponding window time length. The time window module assigns the window time length to the time window parameter. When the wind farm load fluctuates violently and causes frequent high-frequency transient features, the time window module compresses the time window to an extremely narrow range. When the wind farm is in a steady-state, low-wind-power operation, the time window module relaxes the time window restriction to the millisecond level.

[0040] Data association pairs represent data aggregates that are forcibly aligned to the physical occurrence time within the framework of a time hash index table. The binding submodule forcibly associates circulation feature vectors and surface temperature values ​​belonging to the same spatial physical node location as key-value pairs within a tolerance span specified by the time window. The discrimination submodule uses a physics-based algorithm to filter out erroneously associated data pairs. The discrimination submodule compares the time difference verification results and retains data association pairs that conform to objective thermal laws. The discrimination submodule then renames the retained data association pairs and inputs them into the upper-level diagnostic model for business evaluation.

[0041] The dynamic thermal time constant characterizes the physical sluggish inertia phenomenon where the surface temperature of a specific metallic structure begins to rise significantly after being subjected to a sudden impact of a large current. The calculation submodule retrieves the specific heat capacity values ​​of the copper-aluminum alloy joint and the mass value of the metal body, clearly marked in the engineering design drawings. It also calibrates a preset heat dissipation coefficient based on the surrounding hydrological flow velocity parameters. The calculation submodule calculates the dynamic thermal time constant by dividing the product of the specific heat capacity value and the mass value by the preset heat dissipation coefficient. The verification module solidifies the dynamic thermal time constant into a tolerance threshold, serving as a criterion for verifying whether the sequence of current and temperature abrupt changes conforms to the laws of thermodynamic conduction in nature.

[0042] The temperature rise rate characteristic represents the rapid upward trend of self-heating caused purely by abnormal contact resistance or insulation degradation, after eliminating the interference of ambient temperature alternation between day and night. The output submodule extracts the temperature gradient matrix of adjacent nodes and inputs it into the principal component analysis (PCA) algorithm model. The PCA algorithm model uses eigenvalue decomposition to reduce the dimensionality of multi-dimensional redundant heating parameters, mapping them to an orthogonal subspace to extract principal component vectors that serve as the temperature rise rate characteristic. The hierarchical early warning decision model internally nests a multi-layered decision tree structure to analyze the temperature rise rate characteristic. The hierarchical early warning decision model traverses the node conditional branches, calculates the overheating risk probability in percentage form, and pushes it to the control center.

[0043] The thermal impedance model is constructed by equivalently transforming the physical spatial structure and material properties of the submarine optical-electric composite cable. The derivation submodule abstracts the cross-linked polyethylene insulation layer, water-blocking layer, and armor layer inside the optical-electric composite cable into discrete distributed nodes with specific thermal resistance and thermal capacity values. The derivation submodule uses the thermal network topology connection relationship between the nodes to form a system of partial differential equations. In terms of data processing, the derivation submodule receives the feature matrix output by the extraction submodule. Based on Joule's law, the derivation submodule transforms the current parameters in the feature matrix into the boundary conditions of the internal heat source power of the central conductor. The derivation submodule solves the system of partial differential equations for heat conduction to calculate the temperature gradient distribution data of heat conduction radially from the inside to the outside to the fiber layer. The derivation submodule inputs the temperature gradient distribution data into the integral formula of the fiber thermo-optic coefficient to derive the output refractive index deformation value.

[0044] Principal Component Analysis (PCA) algorithm models utilize orthogonal transformation theory from linear algebra to construct dimensionality-reduction mapping matrices. The aim of PCA model construction is to eliminate environmental noise interference while preserving the maximum variance information within multidimensional temperature data. In terms of data processing, the output submodule combines temperature differences from multiple sensor nodes at different time series and inputs them into the PCA algorithm model to form an initial feature matrix. The PCA algorithm model performs centering and covariance matrix calculations on the initial feature matrix, and then solves for the eigenvectors corresponding to the covariance matrix through eigenvalue decomposition. Finally, the PCA algorithm model projects the initial feature matrix into the eigenvector space corresponding to the largest eigenvalue and outputs a 1-dimensional principal component vector as the temperature rise rate feature.

[0045] The tiered early warning decision model is generated through ensemble learning using a multi-layered decision tree structure based on the random forest algorithm framework. The model is pre-loaded with massive amounts of historical normal operation benchmark data and overheating breakdown fault feature samples for node splitting condition calibration. In terms of data processing, the output submodule inputs the extracted temperature rise rate features into the tiered early warning decision model. Each independent decision tree node within the model performs threshold segmentation and path guidance on the temperature rise rate features based on the principle of maximizing information entropy gain. The model then aggregates the classification results from the leaf nodes of all decision trees and performs weighted voting calculations. Finally, the tiered early warning decision model outputs a continuous probability value between 0 and 100% as the overheating risk probability.

[0046] The capture submodule uses a preset amplitude threshold to filter background electrical noise to extract effective harmonic abrupt change signals. The preset amplitude threshold is objectively calibrated based on the background electrical interference level of the high-power offshore wind power converter at normal switching frequencies. The capture submodule sets the base value of the preset amplitude threshold to 50 amperes. When the absolute amplitude of the high-frequency current component output by wavelet transform decomposition exceeds the 50-ampere standard limit, the capture submodule determines that the corresponding high-frequency current component belongs to a high-frequency current harmonic transient peak caused by converter operation. The capture submodule extracts the timestamps corresponding to the high-frequency current components with numerical characteristics exceeding 50 amperes and records them as the time of current peak occurrence.

[0047] The discrimination submodule uses a preset rate-of-change threshold to locate the initial physical inflection point where the surface temperature of the line joint experiences a significant increase due to a large current surge. The preset rate-of-change threshold is determined by inversely based on the thermal conductivity limit of the outer sheath insulation material of the submarine optical-electric composite cable under standard operating conditions. The specific parameter of the preset rate-of-change threshold is set to 0.02 degrees Celsius per second. The discrimination submodule then searches for the absolute time point corresponding to three consecutive sampling periods where the first derivative of the temperature sequence exceeds 0.02 degrees Celsius per second. The discrimination submodule identifies this absolute time point as the moment of temperature abrupt change to support subsequent thermoelectric inertial time difference comparison logic.

[0048] The verification module uses a tolerance threshold to define a reasonable time delay range between current transients and temperature transients, conforming to the laws of physical thermal inertia. The tolerance threshold is directly mapped from the dynamic thermal time constant calculated by the calculation submodule. In specific engineering implementation environments, the verification module sets the basic value range of the tolerance threshold to between 2.5 seconds and 4.8 seconds. When the time difference calculated by the discrimination submodule falls within the range of 2.5 seconds to 4.8 seconds, the discrimination submodule determines that the underlying data association pairs conform to natural physical causality. The discrimination submodule retains the data association pairs within the tolerance threshold range and outputs them as fused data, which is then fed into the upper-level diagnostic logic unit.

[0049] The extraction submodule obtains the RMS current value and harmonic components of the corresponding cable conductors of the transmission line within a set period. The extraction submodule then concatenates the RMS current value and harmonic components into a column vector and outputs a feature matrix. The expression for the feature matrix is ​​defined as follows: Extracting the feature matrix expression defined in the submodule, Represents the characteristic matrix, This indicates the effective value of the current in the cable conductor of the transmission line within a set period. to This indicates the harmonic components corresponding to the cable conductors of the transmission line within a set period. This represents the highest order harmonic constant. This represents the matrix transpose operation symbol. After the feature matrix is ​​constructed by the extraction submodule, it is input into the thermal calculation link.

[0050] The derivation submodule substitutes the characteristic matrix into the thermal impedance model and outputs the corresponding conductor heat generation. The formula for calculating the conductor heat generation is set as follows: In deriving the formula for calculating the heat generation of a conductor as defined in the submodule, Indicates the heat generated by the conductor. This indicates the effective value of the current in the cable conductor of the transmission line within a set period. This indicates the DC resistance value of the cable conductor. This indicates the cable conductor corresponding to the transmission line within a set period. Second harmonic components Indicates the conductor of the cable corresponds to the first The value of secondary AC resistance. This represents the highest order harmonic constant. This represents the index parameter for the cumulative summation of harmonic orders. The derivation submodule uses thermodynamic conduction formulas to calculate the temperature gradient distribution data corresponding to the heat transfer from the conductor to the communication cable layer. The formula for calculating the temperature gradient distribution data is expressed as follows: In deriving the formula for calculating the temperature gradient distribution data defined in the submodule, Represents temperature gradient distribution data. Indicates the heat generated by the conductor. This indicates the equivalent thermal resistance value between the communication cable layer and the cable conductor. This indicates the time length corresponding to the set period. This represents the physical thermal time constant corresponding to the cable conductor. This represents the natural constant. The derivation submodule integrates the temperature gradient distribution data and the preset fiber thermo-optic coefficient to output the refractive index deformation value, and uses the refractive index deformation value as a dynamic perturbation quantity. The derivation formula for the dynamic perturbation quantity is as follows: In the derivation formula of the dynamic disturbance quantity defined in the derivation submodule, This represents the dynamic disturbance. This indicates the physical length of the communication optical fiber embedded in the submarine optical-electric composite cable. The physical constant representing the speed of light in a vacuum. This indicates the preset optical fiber thermo-optic coefficient. This represents the temperature gradient distribution data. The derivation submodule uses dynamic perturbation to perform numerical increment / decrement compensation operations on the initial timestamp, outputting standardized time-series data.

[0051] The time window module extracts the maximum and minimum active power values ​​within a set time period from the active power data. The module subtracts the minimum active power value from the maximum and divides the result by the corresponding duration of the set time period, outputting the first derivative. This first derivative is then used as the instantaneous fluctuation rate of active power. The formula for calculating the instantaneous fluctuation rate of active power is: In the formula for calculating the instantaneous fluctuation rate of active power defined in the time window module, This represents the instantaneous fluctuation rate of active power. This indicates the maximum active power value. This represents the minimum active power value. This indicates the duration of the set time period. The time window submodule inputs the instantaneous fluctuation rate of active power into the preset negative exponential decay function formula, outputs the corresponding window time length, and defines the window time length as the time window. The preset negative exponential decay function formula is set as follows: In the time window calculation formula defined in the time window submodule, Indicates a time window. This represents a pre-defined constant for the length of a reference time window. Represents the natural constant. This represents a constant representing the volatility sensitivity coefficient. This represents the instantaneous fluctuation rate of active power.

[0052] The calculation submodule retrieves the specific heat capacity and mass values ​​of the corresponding metal material for the transmission line from the preset material database as metal material parameters. The calculation submodule multiplies the specific heat capacity value by the mass value and divides it by a preset heat dissipation coefficient to output the dynamic thermal time constant. The formula for calculating the dynamic thermal time constant is: In the dynamic thermal time constant calculation formula defined in the calculation submodule, Represents the dynamic thermal time constant. This indicates the specific heat capacity value included in the metal material parameters. This indicates the mass value included in the metal material parameters. This represents the preset heat dissipation coefficient. The discrimination submodule uses the dynamic thermal time constant to set a tolerance threshold and performs physical violation data filtering. The output submodule reads the temperature values ​​included in the fused data and calculates the temperature difference between adjacent time points, outputting a temperature gradient matrix. The formula for constructing the temperature gradient matrix is: In the formula for constructing the temperature gradient matrix defined in the output submodule, Represents the temperature gradient matrix. to This represents the continuous temperature values ​​corresponding to each independent time point within the fused data, arranged in chronological order. The output submodule inputs the temperature gradient matrix into the principal component analysis algorithm model for dimensionality reduction and outputs the principal component vectors, which are then used to determine the temperature rise rate feature.

[0053] The formula for calculating the temperature rise rate characteristic is: In the temperature rise rate characteristic calculation formula defined in the output submodule, Indicates the characteristic of the rate of temperature rise. This represents the orthogonal eigenvector matrix corresponding to the largest eigenvalue extracted through eigenvalue decomposition within the principal component analysis algorithm model. This represents the matrix transpose operator. This represents the temperature gradient matrix. The output submodule takes the temperature rise rate characteristics as input to a hierarchical early warning decision model with a multi-layered decision tree structure, calculates node probabilities, and outputs the overheating risk probability. The formula for calculating the overheating risk probability is: In the overheating risk probability calculation formula defined in the output submodule, Indicates the probability of overheating risk. This represents the total number of multi-level decision tree structures included within the hierarchical early warning decision-making model. Indicates the first Each tree is identified by an independent numerical label. Indicates the first The tree structure independently predicts the local overheating risk based on the input parameters. This indicates the characteristic of the rate of temperature rise.

[0054] This invention provides a complete numerical verification process covering everything from bottom-level computing power derivation to top-level business diagnosis. The acquisition module obtains the effective current value of the cable conductor corresponding to the transmission line within a set period, which is 1000 amperes. The acquisition module also obtains the first high-frequency harmonic component of the cable conductor, which is 100 amperes. The DC resistance value of the cable conductor is 0.0001 ohms, and the first AC resistance value is 0.0002 ohms. The derivation submodule substitutes the conductor heat generation calculation formula to obtain a conductor heat generation of 102 watts. The equivalent thermal resistance value between the communication cable layer and the cable conductor is set to 0.5 Kelvin per watt. The set period corresponds to a time length of 60 seconds, and the physical thermal time constant of the cable conductor is set to 120 seconds. The derivation submodule substitutes the temperature gradient distribution data calculation formula to obtain a temperature gradient distribution data of 20.04 Kelvin. The submarine optical-electric composite cable embeds a communication optical fiber with a physical length of 10,000 meters. The physical constant of the speed of light in a vacuum is taken as 300,000,000 meters per second. The preset optical fiber thermo-optic coefficient is 1 / 100,000. The derivation submodule calculates and outputs a dynamic disturbance value of 6.68 nanoseconds. The maximum active power recorded is 500 megawatts, and the minimum active power is 400 megawatts. The set time period corresponds to a duration of 10 seconds. The time window submodule calculates and outputs an instantaneous fluctuation rate of 10 megawatts per second for active power. The reference time window length constant is set to 50 milliseconds, the fluctuation sensitivity coefficient constant is taken as 0.1, and the time window submodule calculates a time window of 18.39 milliseconds. The specific heat capacity included in the metal material parameters is 390 joules per kilogram in Kelvin, the mass included in the metal material parameters is 10 kilograms, the preset heat dissipation coefficient is set to 130 watts per Kelvin, and the calculation submodule derives and outputs a dynamic thermal time constant of 30 seconds. The temperature gradient matrix constructed by the output submodule is reduced in dimensionality to obtain a temperature rise rate feature value of 0.5. The sum of the comprehensive prediction values ​​of the 100 judgment trees included in the hierarchical early warning decision model is equal to 85. Substituting the output submodule into the overheating risk probability calculation formula, the final output overheating risk probability is 85%.

[0055] Embodiment 1 of this invention provides system operation logic for scenarios involving rapid increases in wind speed at offshore wind farms. The clock module within the control device sends synchronization messages with a base timestamp to the circulation monitoring device and the temperature monitoring device. The circulation monitoring device and the temperature monitoring device receive the synchronization messages and collect corresponding environmental parameters starting from the base timestamp, outputting raw monitoring data. The acquisition module uses the equipment to acquire raw monitoring data corresponding to the transmission line, load current data, and active power data. The extraction submodule within the compensation module extracts the effective current value and harmonic components of the corresponding cable conductor of the transmission line within a set period from the load current data. The extraction submodule then concatenates the effective current value and harmonic components to output a feature matrix. The derivation submodule substitutes the feature matrix into the thermal impedance model to output the corresponding conductor heat generation. It then uses the thermodynamic conduction formula to calculate the temperature gradient distribution data corresponding to the heat transfer from the conductor to the communication cable layer. Next, it integrates the temperature gradient distribution data and the preset fiber thermo-optic coefficient to output the refractive index deformation value, using this value as a dynamic disturbance. The compensation module uses this dynamic disturbance to correct the initial timestamp and output standardized time-series data. The time window submodule within the index module extracts the maximum and minimum active power values ​​within a set time period from the active power data. It subtracts the minimum active power value from the maximum and divides by the corresponding duration of the set time period to output the first derivative. This first derivative is then used as the instantaneous fluctuation rate of active power, which is input into a preset negative exponential decay function formula to output the corresponding window time length. The time window submodule defines the window time length as the time window. The binding submodule extracts the initial timestamp, which has been adjusted by addition and subtraction, from the standardized time series data as the primary key. The binding submodule uses the adjusted initial timestamp to construct a time hash index table. Within the time length defined by the time window, the binding submodule uses the time hash index table to retrieve the circulation feature vector and temperature value corresponding to the same initial timestamp. The binding submodule encapsulates and stores the retrieved circulation feature vector and temperature value to output data association pairs.

[0056] Embodiment 2 of this invention provides data physical verification logic for business application scenarios involving transient impacts caused by wind turbine switching operation states. The verification module includes a capture submodule that receives data association pairs. The capture submodule performs wavelet transform decomposition on the circulating current feature vectors included in the data association pairs, outputting current components corresponding to different frequency bands. The capture submodule filters out current peak data exceeding a preset amplitude threshold from the current components. The capture submodule records the corresponding time node of the current peak data within standardized time-series data as the time of current peak occurrence. The verification module also includes a calculation submodule connected to a preset material database. The calculation submodule retrieves the specific heat capacity and mass values ​​of the corresponding metal material of the transmission line from the preset material database as metal material parameters. The calculation submodule multiplies the specific heat capacity value by the mass value and divides it by a preset heat dissipation coefficient to output a dynamic thermal time constant. A discrimination submodule calculates the first derivative of the temperature value within the data association pairs as a function of time. The discrimination submodule finds the time node corresponding to the first derivative exceeding a preset rate of change threshold to determine the time of temperature abrupt change. The discrimination submodule subtracts the time of current peak occurrence from the time of temperature abrupt change and outputs the time difference. The discrimination submodule compares the time difference with the tolerance threshold. When the time difference is greater than the tolerance threshold, the discrimination submodule executes an objective common sense judgment program to remove the false data associations caused by residual errors and delete the corresponding data association pairs. When the time difference is less than or equal to the tolerance threshold, the discrimination submodule outputs the corresponding data association pairs as fused data.

[0057] Embodiment 3 of this invention provides a status diagnosis logic for power transmission line joints operating under long-term degradation and overheating risks. The diagnosis module includes an output submodule that reads temperature values ​​from the fused data, calculates the temperature difference between adjacent time points to output a temperature gradient matrix, inputs the temperature gradient matrix into a principal component analysis algorithm model for dimensionality reduction, and outputs principal component vectors as the temperature rise rate feature. The output submodule then inputs the temperature rise rate feature into a hierarchical early warning decision model with a multi-layered decision tree structure to calculate node probabilities and output the overheating risk probability. Finally, the output submodule searches a pre-defined probability interval mapping table to obtain the corresponding health level name for the overheating risk probability and outputs it as the result of the line health status model.

[0058] A transmission line data fusion health diagnosis system underwent a 120-day comparative verification test in a real offshore wind farm environment. The testing team selected two identical 220 kV submarine fiber optic composite cables as test subjects. The first cable deployed a traditional network time protocol synchronization system as a control group. The second cable deployed the transmission line data fusion health diagnosis system as the experimental group. The testing team artificially injected 50 simulated current surge pulses into both cables and recorded the accuracy of the fused data correlation and the accuracy of overheating risk probability prediction for both systems.

[0059] The control group experienced severe clock drift due to ocean current temperature variations and Joule heating under high load current during the 120-day test period. The data association pairs output by the control group exhibited numerous microsecond-level misalignments, resulting in a fusion data association accuracy of only 72.5%. The control group's tiered early warning decision model, due to input distortion, generated 14 false overheating alarms and 3 missed alarms, achieving an overheating risk probability prediction accuracy of 81.2%. The experimental group, including a delay-oriented compensation module and a coupling logic verification module, dynamically corrected time deviations in real time and filtered out physically contradictory data. The experimental group strictly controlled time synchronization errors to within 1 microsecond. The data association pairs output by the experimental group fully conformed to thermoelectric physical inertial logic, achieving a fusion data association accuracy of 99.8%. The tiered early warning decision model in the experimental group accurately captured temperature rise rate characteristics, accurately predicting overheating risks in all 50 simulated overheating events without any false alarms. The overheating risk probability prediction accuracy of the experimental group reached 99.9%. Experimental data verify that the intelligent fusion health diagnosis system for multi-source data of transmission lines completely eliminates basic correlation errors caused by harsh environments, thereby significantly improving the reliability of fault early warning.

Claims

1. A multi-source data intelligent fusion health diagnosis system for power transmission lines, characterized in that, It includes equipment and a control device, wherein the control device establishes a communication connection with the equipment, and the control device includes: The acquisition module is used to acquire the original monitoring data, load current data and active power data corresponding to the transmission line using the equipment. The original monitoring data is accompanied by an initial timestamp. The compensation module is used to generate a feature matrix using the load current data, substitute the feature matrix into the thermal impedance model to output conductor heat generation and temperature gradient distribution data, perform integral calculations using the temperature gradient distribution data and the preset fiber thermo-optic coefficient to derive the corresponding dynamic disturbance amount inside the communication cable, and use the dynamic disturbance amount to add or subtract corrections to the initial timestamp to output standardized time series data. The index module is used to calculate the first derivative corresponding to the active power data to determine the instantaneous fluctuation rate of active power, use the instantaneous fluctuation rate of active power as the independent variable to construct a decay function output time window, and match and combine the circulation feature vector included in the standardized time series data with the temperature value within the time window to output data association pairs. The verification module is used to extract the time of occurrence of the peak current and the time of temperature change in the data association pair, subtract the time of occurrence of the peak current from the time of temperature change to determine the time difference, calculate the dynamic thermal time constant using the metal material parameters, use the dynamic thermal time constant as the tolerance threshold, and retain the data whose time difference is within the tolerance threshold as the fused data. The diagnostic module is used to extract temperature rise rate features from the fused data, input the temperature rise rate features into a pre-trained graded early warning decision model to output overheating risk probability, and use the overheating risk probability to match the corresponding health level to output the line health status model result.

2. The power transmission line data fusion health diagnosis system according to claim 1, characterized in that, The equipment includes a circulation monitoring device and a temperature monitoring device; The control device includes a clock module; The clock module sends a synchronization message with a base timestamp to the circulation monitoring device and the temperature monitoring device. The circulation monitoring device and the temperature monitoring device receive the synchronization message and collect corresponding environmental parameters starting from the base timestamp, then output the raw monitoring data.

3. The power transmission line data fusion health diagnosis system according to claim 2, characterized in that, The compensation module includes an extraction sub-module; The extraction submodule is used to extract the effective value of the current and harmonic components of the cable conductor corresponding to the transmission line within a set period from the load current data, and to splice the effective value of the current and the harmonic components to output the feature matrix.

4. The power transmission line data fusion health diagnosis system according to claim 3, characterized in that, The compensation module includes a derivation submodule; The derivation submodule is used to substitute the feature matrix into the thermal impedance model to output the corresponding conductor heat generation, use the thermodynamic conduction formula to calculate the temperature gradient distribution data corresponding to the heat generation of the conductor transferred to the communication cable layer, integrate the temperature gradient distribution data and the preset optical fiber thermo-optic coefficient to output the refractive index deformation value, and use the refractive index deformation value as the dynamic disturbance quantity.

5. The power transmission line data fusion health diagnosis system according to claim 4, characterized in that, The index module includes a time window submodule; The time window module is used to extract the maximum and minimum active power values ​​within a set time period from the active power data, subtract the minimum active power value from the maximum active power value and divide by the corresponding duration of the set time period to output the first derivative, determine the instantaneous fluctuation rate of active power from the first derivative, input the instantaneous fluctuation rate of active power into the decay function, the decay function is a preset negative exponential decay function formula, output the corresponding window time length, and determine the window time length as the time window.

6. The power transmission line data fusion health diagnosis system according to claim 5, characterized in that, The index module includes a binding submodule; The binding submodule is used to extract the initial timestamp, which has been corrected by addition and subtraction, carried in the standardized time series data as the primary key, construct a time hash index table using the initial timestamp, and retrieve the circulation feature vector and temperature value corresponding to the same initial timestamp using the time hash index table within the time length defined by the time window. The retrieved circulation feature vector and temperature value are then packaged, stored, and output as the data association pair.

7. The power transmission line data fusion health diagnosis system according to claim 6, characterized in that, The verification module includes a capture submodule; The capture submodule is used to receive the data association pair, perform wavelet transform decomposition on the circulating current feature vector included in the data association pair to output current components corresponding to different frequency bands, filter out the current peak data corresponding to the current exceeding the preset amplitude threshold from the current components, and record the current peak data at the corresponding time node in the standardized time series data as the time when the current peak occurs.

8. The power transmission line data fusion health diagnosis system according to claim 7, characterized in that, The verification module includes an operation submodule; The calculation submodule is used to connect to a preset material database, retrieve the specific heat capacity value and mass value of the metal material corresponding to the transmission line from the preset material database as the metal material parameter, multiply the specific heat capacity value by the mass value and divide by a preset heat dissipation coefficient to output the dynamic thermal time constant.

9. The power transmission line data fusion health diagnosis system according to claim 8, characterized in that, The verification module includes a discrimination submodule; The discrimination submodule is used to calculate the first derivative of the temperature value changing with time within the data association pair, find the time node corresponding to the first derivative exceeding a preset rate of change threshold to determine the temperature change moment, subtract the current peak occurrence moment from the temperature change moment to output the time difference, compare the time difference with the tolerance threshold, delete the corresponding data association pair when the time difference is greater than the tolerance threshold, and output the corresponding data association pair as the fused data when the time difference is less than or equal to the tolerance threshold.

10. The power transmission line data fusion health diagnosis system according to claim 9, characterized in that, The diagnostic module includes an output submodule; The output submodule is used to read the temperature values ​​included in the fused data, calculate the temperature difference between adjacent time nodes and output a temperature gradient matrix, input the temperature gradient matrix into a principal component analysis algorithm model for dimensionality reduction and output principal component vectors to determine the temperature rise rate feature, input the temperature rise rate feature into the hierarchical early warning decision model including a multi-layer decision tree structure for node probability calculation and output the overheating risk probability, and search a pre-set probability interval mapping table to obtain the health level name to which the overheating risk probability belongs and output it as the result of the line health status model.

Images