A method, apparatus, equipment, and storage medium for multi-source heterogeneous data fusion processing based on an energy data platform.

Abstract

This application discloses a method, apparatus, device, and storage medium for multi-source heterogeneous data fusion processing based on an energy data platform, relating to the field of energy information technology. The method includes: acquiring three-phase current waveform data, circuit breaker coil current data, and load switch status signals from a power distribution terminal; analyzing the circuit breaker's operating characteristics based on the circuit breaker coil current data to obtain an operating feature vector; constructing a switch operating timing constraint matrix based on the load switch status signal and the operating feature vector; performing fusion analysis on the three-phase current waveform data based on the switch operating timing constraint matrix to obtain a fault feature tensor; and performing spatial location analysis on the fault feature tensor based on a traveling wave propagation model to obtain the fault point coordinates. This application can accurately locate fault points in power supply lines under low-cost equipment conditions, providing support for rapid fault repair.

Description

Technical Field

[0001] This application relates to the field of energy information technology, and in particular to a method, apparatus, equipment and storage medium for multi-source heterogeneous data fusion processing based on an energy data platform. Background Technology

[0002] As the digital transformation of the energy industry accelerates, the scale and complexity of energy data are constantly increasing. Energy data covers multiple stages, including power generation, transmission, distribution, and consumption. It is not only massive in volume but also comes from diverse sources and formats, exhibiting typical multi-source heterogeneous characteristics. To achieve efficient operation and optimized management of the energy system, effective integration and analysis of this data are necessary.

[0003] Currently, the energy industry typically employs traditional data integration methods, such as Extract, Transform, and Load (ETL) techniques, when processing multi-source heterogeneous data. These methods extract data from different sources into a centralized data warehouse, where it is cleaned, transformed, and integrated to achieve unified data management and analysis. In addition, some enterprises have adopted distributed computing frameworks, such as Hadoop and Spark, to process large-scale energy data.

[0004] Despite the widespread application of existing data processing technologies in the energy sector, several challenges remain. First, traditional ETL techniques are inefficient when processing large-scale, multi-source, heterogeneous data, making it difficult to meet real-time requirements. Second, while distributed computing frameworks can handle large-scale data, they face challenges in ensuring the accuracy and consistency of data fusion. More importantly, existing technologies struggle to balance equipment cost and fault location accuracy, particularly in power line fault location. Current methods often require expensive equipment to achieve high-precision fault location, limiting their widespread adoption in practice. Therefore, accurately locating fault points in power lines using low-cost equipment is a pressing issue that needs to be addressed. Summary of the Invention

[0005] The purpose of this application is to provide a multi-source heterogeneous data fusion processing method based on an energy data platform, which aims to solve the technical problem of how to accurately locate fault points in power supply lines under low-cost equipment conditions.

[0006] To achieve the above objectives, this application proposes a method for multi-source heterogeneous data fusion processing based on an energy data platform, the method comprising:

[0007] Acquire the three-phase current waveform data, circuit breaker coil current data, and load switch status signals of the power distribution terminal;

[0008] Based on the circuit breaker coil current data, the circuit breaker's operating characteristics are analyzed to obtain an operating feature vector.

[0009] Construct a switch action timing constraint matrix based on the load switch status signal and the action feature vector;

[0010] Based on the switching action timing constraint matrix, the three-phase current waveform data are fused and analyzed to obtain the fault feature tensor;

[0011] Based on the traveling wave propagation model, the fault feature tensor is spatially located and analyzed to obtain the coordinates of the fault point.

[0012] In one embodiment, the step of analyzing the circuit breaker's operating characteristics based on the circuit breaker coil current data to obtain an operating feature vector includes:

[0013] Multiple independent current pulses in the circuit breaker coil current data are identified using a multi-threshold gradient analysis method, and the rising edge boundary point and falling edge boundary point of the current pulse are obtained.

[0014] Based on the rising edge boundary point and the falling edge boundary point, calculate the pulse width characteristics, current amplitude extreme value characteristics, and pulse shape steepness characteristics of the current pulse;

[0015] Detect the continuous latch-up time characteristics and adjacent pulse interval characteristics among all the current pulses;

[0016] Based on the pulse width characteristics and the adjacent pulse interval characteristics, a set of effective pulses is obtained from the current pulses;

[0017] The current amplitude extreme value feature, pulse shape steepness feature, and continuous blocking time feature of the current pulse in the effective pulse set are encoded into a preset dimension by a segmented feature compression algorithm to obtain the action feature vector.

[0018] In one embodiment, the step of calculating the pulse width characteristics, current amplitude extreme value characteristics, and pulse kurtosis characteristics of the current pulse based on the rising edge boundary point and the falling edge boundary point includes:

[0019] The pulse width characteristics of the current pulse are calculated based on the rising edge boundary point and the falling edge boundary point;

[0020] The maximum value of the current value sequence in the current pulse is taken as the extreme value feature of the current amplitude;

[0021] The average rising gradient is calculated based on the current value sequence between the rising edge boundary point and the maximum value.

[0022] The average descent gradient is calculated based on the current value sequence between the descent edge boundary point and the maximum value.

[0023] The pulse morphology steepness characteristics of the current pulse are calculated based on the average rising gradient and the average falling gradient.

[0024] In one embodiment, the step of constructing the switching action timing constraint matrix based on the load switch state signal and the action feature vector includes:

[0025] Analyze the state transition time points in the load switch status signal to form a switch event time sequence;

[0026] Extract the rising edge start time of all current pulses from the action feature vector to form the circuit breaker action time sequence;

[0027] The time series of the switching events and the time series of the circuit breaker actions are paired according to time to obtain a set of event pairs;

[0028] Calculate the absolute time difference between the switching event time and the circuit breaker operating time for each event pair in the event pair set;

[0029] When the absolute time difference is greater than a preset tolerance threshold, an exception identifier is set for the event pair corresponding to the absolute time difference.

[0030] Construct a binary association matrix with switch events as rows and circuit breaker actions as columns, and assign values ​​to the matrix elements of the binary association matrix according to the exception identifier;

[0031] The spatiotemporal correlation normalization process is performed on the binary correlation matrix to obtain the switching action timing constraint matrix.

[0032] In one embodiment, the step of fusing and analyzing the three-phase current waveform data based on the switching action timing constraint matrix to obtain the fault feature tensor includes:

[0033] Bispectral analysis was performed on the three-phase current waveform data to extract the signal phase coupling characteristics in each frequency band, and a bispectral feature matrix was constructed based on the signal phase coupling characteristics.

[0034] The switching action timing constraint matrix is ​​sliced ​​along the time dimension to obtain multiple timing constraint submatrices;

[0035] The bispectral feature matrix is ​​aligned with the temporal constraint submatrix in a time-frequency manner.

[0036] Once the time-frequency alignment is complete, the correlation weights between the bispectral feature matrix and the temporal constraint submatrix are calculated using a cross-domain coupling algorithm.

[0037] The bispectral feature matrix is ​​reconstructed by weighting according to the correlation weights to form an initial three-dimensional feature space;

[0038] The initial three-dimensional feature space is reduced and normalized by tensor to obtain the fault feature tensor.

[0039] In one embodiment, the step of spatially locating and analyzing the fault feature tensor based on the traveling wave propagation model to obtain the fault point coordinates includes:

[0040] Obtain monitoring point location information and line electrical parameters from a pre-defined power grid topology database;

[0041] A traveling wave propagation model is constructed based on the electrical parameters of the line. The traveling wave propagation model includes a velocity calculation function and an attenuation correction function.

[0042] Based on the location information of the monitoring points, the traveling wave arrival time sequence and waveform distortion features of each monitoring point are extracted from the fault feature tensor.

[0043] A hyperbolic positioning equation system with the monitoring point as the focus is constructed, and the hyperbolic parameters of the hyperbolic positioning equation system are determined by the spatial distance difference characteristics;

[0044] Solving the hyperbolic positioning equations yields a set of intersection points, which are then mapped onto the power grid coordinate system to form an initial candidate region.

[0045] Based on the traveling wave arrival time sequence, the waveform distortion characteristics, the velocity calculation function, and the attenuation correction function, the initial candidate region is screened for matching degree to obtain the target candidate region;

[0046] With the goal of minimizing the travel wave arrival time residual, particle swarm optimization is performed in the target candidate region to obtain the fault point coordinates.

[0047] In one embodiment, the step of performing matching degree screening on the initial candidate region based on the traveling wave arrival time sequence, the waveform distortion characteristics, the velocity calculation function, and the attenuation correction function to obtain the target candidate region includes:

[0048] Based on the velocity calculation function, the arrival time series of the traveling wave is converted into spatial distance difference features;

[0049] The theoretical distortion value of the initial candidate region is calculated based on the attenuation correction function.

[0050] The theoretical distortion value is matched with the waveform distortion feature to obtain the matching result;

[0051] The target candidate region is obtained by filtering from the initial candidate region based on the matching results.

[0052] Furthermore, to achieve the above objectives, this application also proposes a multi-source heterogeneous data fusion processing device based on an energy data platform, the device comprising:

[0053] The data acquisition module is used to acquire three-phase current waveform data, circuit breaker coil current data and load switch status signals of the power distribution terminal;

[0054] The action analysis module is used to analyze the action characteristics of the circuit breaker based on the circuit breaker coil current data and obtain the action feature vector.

[0055] The constraint construction module is used to construct a switch action timing constraint matrix based on the load switch status signal and the action feature vector;

[0056] The fusion analysis module is used to perform fusion analysis on the three-phase current waveform data according to the switching action timing constraint matrix to obtain the fault feature tensor;

[0057] The location analysis module is used to perform spatial location analysis on the fault feature tensor based on the traveling wave propagation model to obtain the coordinates of the fault point.

[0058] Furthermore, to achieve the above objectives, this application also proposes a multi-source heterogeneous data fusion processing device based on an energy data platform. The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor. The computer program is configured to implement the steps of the multi-source heterogeneous data fusion processing method based on an energy data platform as described above.

[0059] In addition, to achieve the above objectives, this application also proposes a storage medium, which is a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the multi-source heterogeneous data fusion processing method based on the energy data platform described above.

[0060] One or more technical solutions proposed in this application have at least the following technical effects:

[0061] First, the energy data platform system acquires three-phase current waveform data, circuit breaker coil current data, and load switch status signals from the power distribution terminal, providing comprehensive and real-time equipment operating status data for subsequent analysis. Next, the system analyzes the circuit breaker coil current data, extracting the circuit breaker's operating characteristics and forming an operating feature vector. This step, by quantifying the circuit breaker's operational performance, provides crucial evidence for fault early warning. Then, the system constructs a switch operating timing constraint matrix using the load switch status signal and the operating feature vector. By establishing a timing correlation between switch operations and circuit breaker operations, it provides clues for fault detection. Subsequently, the system performs fusion analysis on the three-phase current waveform data based on the switch operating timing constraint matrix to obtain a fault feature tensor. This fusion process more comprehensively reflects the system status, improving data processing efficiency and accuracy. Finally, the system performs spatial location analysis on the fault feature tensor based on a traveling wave propagation model to obtain the fault point coordinates. This application enables accurate location of fault points in power supply lines under low-cost equipment conditions, providing support for rapid fault repair. Attached Figure Description

[0062] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0063] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0064] Figure 1 This is a flowchart illustrating an embodiment of the multi-source heterogeneous data fusion processing method based on an energy data platform provided in this application.

[0065] Figure 2 This is a flowchart illustrating Embodiment 2 of the multi-source heterogeneous data fusion processing method based on an energy data platform in this application.

[0066] Figure 3 A simplified flowchart illustrating the multi-source heterogeneous data fusion processing method based on an energy data platform provided in Embodiment 2 of this application;

[0067] Figure 4 This is a schematic diagram of the module structure of the multi-source heterogeneous data fusion processing device based on the energy data platform according to an embodiment of this application;

[0068] Figure 5 This is a schematic diagram of the hardware operating environment involved in the multi-source heterogeneous data fusion processing method based on the energy data platform in the embodiments of this application.

[0069] The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0070] It should be understood that the specific embodiments described herein are merely illustrative of the technical solutions of this application and are not intended to limit this application.

[0071] To better understand the technical solution of this application, a detailed description will be provided below in conjunction with the accompanying drawings and specific implementation methods.

[0072] It should be noted that the executing entity of this application embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or an electronic device or energy data platform system capable of realizing the above functions. The following description uses an energy data platform system as an example to illustrate this embodiment and the subsequent embodiments.

[0073] Based on this, the embodiments of this application provide a method for multi-source heterogeneous data fusion processing based on an energy data platform, referring to... Figure 1 , Figure 1 This is a flowchart illustrating the first embodiment of the multi-source heterogeneous data fusion processing method based on an energy data platform in this application.

[0074] In this embodiment, the multi-source heterogeneous data fusion processing method based on the energy data platform includes steps S10 to S50:

[0075] Step S10: Obtain the three-phase current waveform data, circuit breaker coil current data, and load switch status signal of the power distribution terminal.

[0076] It should be noted that a distribution terminal unit (DTU) is an intelligent monitoring and control device in a power distribution network. It is usually deployed in substations or ring main units to collect information such as line electrical quantities (e.g., voltage, current) and switch status, and to perform protection and control functions.

[0077] Three-phase current waveform data refers to the waveform data of the current of each phase changing over time in a three-phase AC power system. It is a sequence of instantaneous three-phase current values ​​collected by the distribution terminal, reflecting the current amplitude, phase and harmonic characteristics when the line is under load or fault.

[0078] Circuit breaker coil current data refers to the current waveform in the drive coil during the opening and closing operation of the circuit breaker, which is used to analyze the circuit breaker's operating characteristics (such as operating time and jamming abnormalities).

[0079] Load switch status signals are digital signals that reflect the position of load switches on distribution lines (such as "open / closed" status), and are used to determine network topology changes and fault isolation range.

[0080] Understandably, firstly, the energy data platform system establishes a stable communication connection with the power distribution terminal to receive real-time three-phase current waveform data, circuit breaker coil current data, and load switch status signals collected by the terminal, ensuring data integrity and accuracy. Secondly, the system preprocesses this data, including data format conversion, noise filtering, and data verification, to improve data quality and provide a reliable data foundation for subsequent analysis. Finally, the system stores the processed data in the corresponding database, classifies and indexes it for later querying and analysis.

[0081] Step S20: Analyze the circuit breaker's operating characteristics based on the circuit breaker coil current data to obtain the operating feature vector.

[0082] It should be noted that the operating characteristics of a circuit breaker refer to the dynamic performance parameters of the circuit breaker during the opening or closing process, including key indicators reflecting the mechanical state and electrical response such as operating time, current change rate, and mechanical vibration, which are used to evaluate the reliability and abnormal conditions of the circuit breaker operation.

[0083] An action feature vector quantifies the operating characteristics of a circuit breaker into a fixed-dimensional numerical vector, typically containing time parameters (such as tripping duration), electrical parameters (such as peak current), and derived features (such as current rise slope). The formula for generating the action feature vector is as follows:

[0084]

[0085] Among them, T d This refers to the pulse duration; I p This refers to the peak current; S r This refers to the pulse steepness; t0 refers to the start time of the pulse current; I0 ​​refers to the current value at time t0; t p This refers to the moment when the pulse current reaches its peak value.

[0086] Understandably, the energy data platform system first preprocesses the collected circuit breaker coil current data to remove noise and interference signals, ensuring data accuracy and reliability. Second, the system uses signal processing techniques, such as Fourier transform or wavelet transform, to perform time-domain and frequency-domain analysis on the preprocessed current data, extracting key characteristic parameters of the current waveform, such as peak current, rise time, and duration. These parameters directly reflect the electrical characteristics of the circuit breaker during operation. Then, based on these key characteristic parameters, the system calculates the circuit breaker's operating time, operating speed, and other operating characteristic indicators. By comparing these with standard operating characteristic curves, the system evaluates the accuracy and reliability of the circuit breaker's operation. Finally, the system quantifies these operating characteristic indicators into a set of values, forming an operating feature vector, which is used for subsequent circuit breaker status assessment and fault diagnosis. This enables comprehensive monitoring and analysis of the circuit breaker's operating performance, providing data support for the safe operation of the power system.

[0087] Step S30: Construct a switch action timing constraint matrix based on the load switch status signal and the action feature vector.

[0088] It should be noted that the switch action timing constraint matrix is ​​a mathematical matrix that characterizes the correlation between the load switch state change and the circuit breaker action timing. Its row vectors correspond to switch state transition events (such as opening → closing), and column vectors correspond to circuit breaker action characteristics (such as opening time and current peak). The matrix element values ​​quantify the coordination relationship between the two in the time and space dimensions (0-1 normalized values), which are used to eliminate false fault signals with timing contradictions during fault diagnosis.

[0089] As an example, the step of constructing a switch action timing constraint matrix based on the load switch state signal and the action feature vector includes: parsing the state transition time points in the load switch state signal to form a switch event time sequence; extracting the rising edge start time points of all current pulses from the action feature vector to form a circuit breaker action time sequence; pairing the switch event time sequence with the circuit breaker action time sequence by time to obtain an event pair set; calculating the absolute time difference between the switch event time point and the circuit breaker action time point for each event pair in the event pair set; setting an anomaly identifier for the event pair corresponding to the absolute time difference when the absolute time difference is greater than a preset tolerance threshold; constructing a binary correlation matrix with switch events as rows and circuit breaker actions as columns, and assigning values ​​to the matrix elements of the binary correlation matrix according to the anomaly identifier; and performing spatiotemporal correlation normalization processing on the binary correlation matrix to obtain the switch action timing constraint matrix.

[0090] The state transition time point refers to the precise timestamp when the load switch state changes (such as the transition moment of "closing → opening"), triggered and recorded by a hardware interrupt signal or software edge detection. State transition detection formula:

[0091] ΔS(t)=|S(t)-S(t-1)|

[0092] Where S(t) refers to the load switch state value (0 or 1) at time t, and ΔS(t) refers to the state change. When ΔS(t) > 0, the jump time point is recorded.

[0093] A switch event time series refers to a set of load switch state transition events arranged chronologically. Each event includes a transition type (open / close) and a precise timestamp. Switch event time series T SW It is expressed as follows:

[0094]

[0095] in, It refers to the time of the i-th switch state transition, and N refers to the total number of load switch state transition events.

[0096] A current pulse refers to a transient current waveform triggered by operation in the circuit breaker coil current waveform, which includes rising edge, peak value and falling edge characteristics.

[0097] The rise edge start time refers to the moment when the current pulse waveform rises from the baseline to a preset threshold (e.g., 10% peak value), reflecting the start time of circuit breaker operation. Rise edge time extraction formula:

[0098]

[0099] in, This refers to the start time of the rising edge of the j-th current pulse, V. j It refers to the j-th element of the action feature vector, and M refers to the total number of pulses.

[0100] A circuit breaker operating time sequence refers to a set of circuit breaker operating events arranged chronologically, with each event based on the start time of the rising edge of a current pulse. Circuit breaker operating time sequence T ACT It is expressed as follows:

[0101]

[0102] An event pair set refers to a set of combinations of load switch tripping events and circuit breaker operating events paired according to their temporal proximity, representing the logical relationship between the two. The event pair set is represented as follows:

[0103]

[0104] The switching event time point refers to the specific moment when a single load switch state transition event occurs, with an accuracy down to the millisecond level.

[0105] The circuit breaker operating time point refers to the starting moment of the rising edge of the current pulse corresponding to a single circuit breaker operation, which is aligned with the switching event time.

[0106] The absolute time difference refers to the absolute time difference between the switch switching moment and the circuit breaker operating moment in a paired event, used to quantify timing consistency. The absolute time difference is expressed as follows:

[0107]

[0108] Where, Δt pq It refers to the time difference between the switch event p and the action event q.

[0109] The preset tolerance threshold refers to the critical value (e.g., 150ms) that allows the maximum time deviation of the switch action, based on the relay protection action time setting.

[0110] An exception identifier is a logical flag (0 / 1) that indicates whether an event violates timing constraints; 0 indicates a timing exception, and 1 indicates normal behavior. Exception determination criteria:

[0111]

[0112] Among them, E pq τ refers to the anomaly identifier, and τ refers to the preset tolerance threshold.

[0113] Switching events refer to specific instances of state changes in load switches (such as "opening operation"), and include type and time attributes.

[0114] Circuit breaker operation refers specifically to the opening and closing operations of a circuit breaker, and is determined based on the characteristics of current pulses.

[0115] A binary correlation matrix is ​​a two-dimensional matrix in which rows represent switching events and columns represent circuit breaker actions. The element values ​​(0 or 1) reflect the timing validity of event pairs. The matrix element assignment formula is as follows:

[0116] B pq =1-E pq

[0117] Among them, B pq q refers to the element values ​​of the binary incidence matrix, where p is the row index (corresponding to a switch event) and q is the column index (corresponding to a circuit breaker action). The binary incidence matrix is ​​represented as follows:

[0118]

[0119] Here, B refers to the N×M binary incidence matrix.

[0120] First, the energy data platform system analyzes the load switch status signals point by point, identifying the exact time points when the signals transition from one state to another. These points are recorded and arranged chronologically to form a switch event time series, providing foundational data for subsequent timing analysis. Second, the system extracts the start time of the rising edge of each current pulse from the action feature vector. These time points mark the beginning of the circuit breaker's operation. These time points are also arranged chronologically to form a circuit breaker operation time series. This step is to obtain the specific time points of the circuit breaker operation for comparison with the switch event time series. Next, the system performs time matching between the switch event time series and the circuit breaker operation time series, finding the closest switch events and circuit breaker operations to form event pairs. This establishes the correspondence between switch operations and circuit breaker operations, providing a basis for analyzing their timing logic. Finally, the system calculates the absolute time difference between the switch event time point and the circuit breaker operation time point in each event pair to assess their synchronicity. When the absolute time difference exceeds the preset tolerance threshold, the system sets an anomaly identifier for the event pair, marking those event pairs whose time deviations exceed the normal range, so as to identify possible faults or abnormal situations in the future.

[0121] Subsequently, the system constructs a binary association matrix with switch events as rows and circuit breaker actions as columns, and assigns values ​​to matrix elements based on anomaly identifiers. Normal event pairs are assigned a normal state, and abnormal event pairs are assigned an abnormal state. This step is to represent the association between switch actions and circuit breaker actions in matrix form, facilitating subsequent matrix operations and analysis. Finally, the system performs spatiotemporal correlation normalization on the binary association matrix. Specifically, the system first calculates the temporal correlation weight of each element in the matrix, which is usually calculated based on the ratio of the time difference between event pairs to a preset time window; the smaller the time difference, the higher the weight. The formula for calculating the temporal correlation weight is as follows:

[0122]

[0123] Where, α pq This refers to the time-related weight, with a value range of (0,1] and Δt. pq τ refers to the time difference between event pairs, τ refers to the preset tolerance threshold, and λ refers to the attenuation factor, which is generally taken as λ = 5.

[0124] Then, the spatial relevance weight for each element is calculated. This is typically based on the physical distance or electrical connection between the switch and the circuit breaker; the closer the distance or the tighter the connection, the higher the weight. Spatial relevance weight W p Calculation formula:

[0125]

[0126] in, It refers to the position coordinate of the p-th switch, L ACT This refers to the circuit breaker position coordinates, and dist(·) refers to the Euclidean distance function (unit: m).

[0127] The system multiplies the temporal correlation weight and spatial correlation weight of each element to obtain a comprehensive weight. Next, the original value of each element is multiplied by its comprehensive weight to obtain a normalized value. Finally, all normalized values ​​are standardized so that the matrix element values ​​are distributed between 0 and 1, resulting in the switching action timing constraint matrix. This step is to eliminate differences in time scale and spatial distribution, enabling the matrix to more accurately reflect the timing constraints and logical relationships of switching actions, providing a more reliable basis for subsequent fault diagnosis and system optimization. The formula for spatiotemporal correlation normalization is as follows:

[0128]

[0129] Among them, M norm [p,q] refers to the element value in the p-th row and q-th column of the spatiotemporal normalized constraint matrix, W i This refers to the spatial correlation weight of the i-th switch, where N is the total number of load switch state transition events, and α is the weight of the i-th switch. pj This refers to the time correlation weight between the p-th switch and the j-th action, and M refers to the total number of circuit breaker action events.

[0130] Step S40: Based on the switching action timing constraint matrix, perform fusion analysis on the three-phase current waveform data to obtain the fault feature tensor.

[0131] It should be noted that the fault feature tensor is a high-dimensional data structure that integrates temporal constraints and current waveform features. Its three dimensions correspond to: time series (waveform sampling points), spatial location (monitoring point topology), and feature channels (electrical features such as harmonics, phase, and energy). This tensor can extract the traveling wave propagation pattern caused by the fault using a spatiotemporal convolution kernel, providing a multi-dimensional feature space containing temporal constraints for the localization algorithm. The formula for generating the fault feature tensor is as follows:

[0132] As an example, the step of fusing and analyzing the three-phase current waveform data according to the switching action timing constraint matrix to obtain the fault feature tensor includes: performing bispectral analysis on the three-phase current waveform data, extracting signal phase coupling features in each frequency band, and constructing a bispectral feature matrix based on the signal phase coupling features; slicing the switching action timing constraint matrix in the time dimension to obtain multiple timing constraint sub-matrices; aligning the bispectral feature matrix with the timing constraint sub-matrices in time and frequency; calculating the correlation weights between the bispectral feature matrix and the timing constraint sub-matrices using a cross-domain coupling algorithm after the time and frequency alignment is completed; reconstructing the bispectral feature matrix with weights based on the correlation weights to form an initial three-dimensional feature space; and performing tensor dimensionality reduction and standardization on the initial three-dimensional feature space to obtain the fault feature tensor.

[0133] Signal phase coupling characteristics refer to the nonlinear phase correlation characteristics between different frequency components in a three-phase current signal, reflecting the harmonic interaction caused by a fault. The complex features extracted through bispectral analysis include amplitude and phase information.

[0134] The bispectral characteristic matrix is ​​a two-dimensional complex matrix output from bispectral analysis, where rows and columns correspond to frequency pairs (f1, f2), and element values ​​represent the coupling strength and phase relationship of that frequency pair. It is used to quantify the nonlinear characteristics of fault features. The formula for calculating the bispectral amplitude is as follows:

[0135]

[0136] Wherein, B(f1,f2) refers to the bispectral complex value (unit: A). 3 / Hz 2 R(τ1,τ2) refers to the third-order cumulative quantity of the three-phase current signal (unit: A). 3 f1 and f2 refer to frequency variables, representing two independent frequency components of the analyzed signal, used to detect phase coupling relationships. τ1 and τ2 refer to time delay variables (unit: s), used to calculate the integration kernel of the third-order cumulant R(τ1, τ2). i is the imaginary unit, used for the complex representation of the Fourier transform.

[0137] The bispectral characteristic matrix B is represented as follows:

[0138]

[0139] Among them, f max This refers to the highest analysis frequency, |B(f i f j )| refers to the frequency pair (f i f j The bispectral amplitude at ().

[0140] A timing constraint submatrix is ​​a sub-matrix obtained by dividing the switching action timing constraint matrix along the time axis. Each submatrix covers the switch-circuit breaker relationship within a specific time window, preserving the spatiotemporal constraint characteristics of the original matrix. The formula for the timing constraint submatrix is ​​as follows:

[0141] M k =M[(k-1)·△T:k·△T,:]

[0142] Where M refers to the switching action timing constraint matrix (size T×M), ΔT refers to the single window time length, k refers to the window index (k=1,2,...,K), and M k It refers to the k-th temporal constraint submatrix (size ΔT×M).

[0143] Cross-Domain Coupling Algorithm (CDCA) is an algorithm used to calculate the correlation weights between different data domains. This algorithm is used to fuse electrical features and temporal constraints. By calculating the mutual information entropy or coherence coefficient of bispectral features and temporal constraints, it generates weight parameters that reflect the strength of the correlation between the two.

[0144] The correlation weight is a scalar value ranging from [0,1] that quantifies the correlation between bispectral features and temporal constraints. It is used to adjust the contribution of different frequency components during weighted fusion to eliminate noise interference. The formula for calculating the correlation weight is as follows:

[0145]

[0146] Among them, w kl This refers to the association weight, B k This refers to the bispectral feature of the k-th subband, M. l This refers to the l-th temporal constraint submatrix. This refers to tensor inner product operations.

[0147] The initial three-dimensional feature space refers to the original high-dimensional space composed of weighted bispectral features. The three dimensions are time, frequency, and space (monitoring point), and it contains uncompressed complete fault feature information.

[0148] First, the energy data platform system performs bispectral analysis on the three-phase current waveform data. Specifically, it decomposes the current signal into multiple frequency components and calculates the bispectral density between each component. This involves obtaining nonlinear phase information by calculating the third-order cumulant of the signal, thereby extracting the signal phase coupling characteristics within each frequency band. These characteristics reflect the interaction between different frequency components, providing information on the nonlinear characteristics of the signal for subsequent analysis. Next, a bispectral feature matrix is ​​constructed based on the extracted signal phase coupling characteristics. Specifically, the bispectral density values ​​of each frequency component are arranged in frequency order to form a matrix. Rows of the matrix correspond to different frequency components, and columns correspond to different time points or signal samples. This allows the phase coupling information of the signal to be stored in matrix form, facilitating subsequent processing and analysis. Then, the switching action timing constraint matrix is ​​sliced ​​in the time dimension. This involves dividing the entire timing constraint matrix into multiple sub-matrices according to preset time intervals. Each sub-matrix contains switching action constraint information within a specific time interval. The purpose of this is to decompose the complex timing constraint information into multiple small, easily processed parts, facilitating subsequent time-frequency alignment with the bispectral feature matrix.

[0149] Then, during time-frequency alignment, the frequency components in the bispectral feature matrix need to be matched with the time intervals corresponding to each time-constrained sub-matrix. Specifically, this is done by adjusting the time axis of the bispectral feature matrix to ensure a one-to-one correspondence with the time intervals of the time-constrained sub-matrix, guaranteeing consistency in both time and frequency. This is to enable data from different sources to be fused and analyzed under the same time and frequency benchmark. After time-frequency alignment, the correlation weights between the bispectral feature matrix and the time-constrained sub-matrix are calculated using a cross-domain coupling algorithm. Specifically, statistical models or machine learning methods within the algorithm are used to analyze the data correlation between the two matrices within the corresponding time intervals, calculating the weight value at each time point or frequency component. These weight values ​​reflect the similarity and correlation between the two matrices, providing a scientific basis for weight allocation in subsequent feature fusion. Based on the calculated correlation weights, the bispectral feature matrix is ​​reconstructed using weighted averages. This involves multiplying each element of the bispectral feature matrix by its corresponding correlation weight, and then recombining the weighted data to form an initial three-dimensional feature space. This space contains information in three dimensions: time, frequency, and feature weights. Weighted reconstruction effectively combines the timing information of switching actions with the phase coupling characteristics of the current signal, providing a more comprehensive data foundation for subsequent feature extraction and analysis. Finally, the initial three-dimensional feature space undergoes tensor dimensionality reduction and standardization. Specifically, dimensionality reduction techniques such as Principal Component Analysis (PCA) or Singular Value Decomposition (SVD) are applied to reduce the data dimensionality and remove redundant information. Simultaneously, the dimensionality-reduced data is standardized to unify the numerical range and dimensions of different features, ultimately yielding a fault feature tensor. This process aims to obtain a more concise and standardized feature representation, facilitating subsequent fault diagnosis and analysis.

[0150] Step S50: Spatial localization analysis of the fault feature tensor based on the traveling wave propagation model to obtain the coordinates of the fault point.

[0151] It should be noted that the Traveling Wave Propagation Model (TWPM) is a mathematical model that describes the propagation law of fault traveling waves in the power grid. It includes wave velocity calculation, attenuation characteristics, and quantitative relationships of reflection / refraction coefficients. Based on Maxwell's equations, it considers the electrical parameters of the power line (such as resistance, inductance, capacitance, etc.) and the geometric structure of the line. It is used to convert the time-domain traveling wave signal into spatial distance difference characteristics. This model dynamically corrects the time delay and distortion of the propagation path through line parameters.

[0152] Fault point coordinates refer to the two-dimensional geographic positioning results (latitude and longitude or grid coordinates) based on the distribution network GIS coordinate system, with positioning accuracy down to the meter level, used to guide fault repair.

[0153] As an example, the step of spatially locating and analyzing the fault feature tensor based on the traveling wave propagation model to obtain the fault point coordinates includes: obtaining monitoring point location information and line electrical parameters from a preset power grid topology database; constructing a traveling wave propagation model based on the line electrical parameters, the traveling wave propagation model including a velocity calculation function and an attenuation correction function; extracting the traveling wave arrival time series and waveform distortion features of each monitoring point from the fault feature tensor based on the monitoring point location information; constructing a hyperbolic positioning equation system with the monitoring point as the focus, the hyperbolic parameters of the hyperbolic positioning equation system being determined by spatial distance difference features; solving the hyperbolic positioning equation system to obtain an intersection set, and mapping the intersection set to the power grid coordinate system to form an initial candidate region; performing matching degree screening on the initial candidate region based on the traveling wave arrival time series, the waveform distortion features, the velocity calculation function, and the attenuation correction function to obtain a target candidate region; and performing particle swarm optimization in the target candidate region with the goal of minimizing the traveling wave arrival time residual to obtain the fault point coordinates.

[0154] The pre-defined power grid topology database is a dedicated database that stores the distribution network nodes, line connection relationships, and equipment parameters. It contains structured data such as line type (overhead / cable), length, and impedance, which is used for traveling wave propagation path modeling.

[0155] The location information of the monitoring point refers to the geographical coordinates (latitude and longitude or grid coordinates) of the traveling wave monitoring device installed at the key nodes of the power distribution network, with an accuracy of sub-meter level, and is used for spatial positioning calculation.

[0156] Monitoring points refer to intelligent monitoring terminals deployed in locations such as substations and ring main units, equipped with high-precision clock synchronization modules (such as GPS / BeiDou), used to collect traveling wave signals and record arrival times.

[0157] Line electrical parameters refer to the basic parameters that describe the electromagnetic characteristics of a line, including resistance (R), inductance (L), capacitance (C), and characteristic impedance (Z0) per unit length, which are used to calculate the traveling wave propagation speed and attenuation coefficient.

[0158] The traveling wave propagation model is a mathematical model that quantifies the propagation law of traveling waves in the power grid. It includes wave velocity calculation, attenuation compensation and reflection characteristics, and is used to convert time-domain signals into spatial distance differences.

[0159] The velocity calculation function is a formula for calculating the propagation velocity of traveling waves based on line parameters, taking into account line type (0.97c for overhead lines, 0.5c for cables) and aging correction factors. The velocity calculation function is expressed as follows:

[0160]

[0161] Where L′ refers to the inductance per unit length of the line (unit: H / m), C′ refers to the capacitance per unit length of the line (unit: F / m), η refers to the wave speed correction factor (0.95~1.0 for overhead lines and 0.6~0.7 for cables), and v refers to the traveling wave propagation speed (unit: m / s).

[0162] The attenuation correction function is a mathematical relationship used to compensate for the energy attenuation during traveling wave propagation, and is used to restore the original waveform amplitude at the fault point. The attenuation correction function is expressed as follows:

[0163]

[0164] Where α(f) refers to the frequency-dependent attenuation coefficient (unit: Np / m), R′(f) refers to the resistance per unit length (frequency-dependent, unit: Ω / m), G′(f) refers to the conductance per unit length (frequency-dependent, unit: S / m), and Z0 refers to the characteristic impedance of the line (unit: Ω).

[0165] A traveling wave arrival time series is an ordered set of arrival times of the traveling wave front recorded at various monitoring points, with time synchronization accuracy down to the nanosecond level, used to calculate spatial distance differences. The formula for traveling wave arrival time detection is as follows:

[0166]

[0167] Among them, t p This refers to the time it takes for the traveling wave to reach the monitoring point p, I p (t) refers to the current traveling wave signal at monitoring point p (unit: A).

[0168] Waveform distortion characteristics refer to the waveform distortion features caused by line dispersion effects and reflection superposition during traveling wave propagation, including amplitude attenuation, phase shift, and harmonic distortion rate. The formula for calculating waveform distortion characteristics is as follows:

[0169]

[0170] Among them, D p It refers to the waveform distortion rate, I ref (t) refers to the standard traveling wave template (unit: A), I p (t) Same as above.

[0171] The spatial distance difference feature refers to the theoretical distance difference (Δd = v × Δt) between the fault point and different monitoring points, where Δd is calculated from the traveling wave propagation velocity (v) and the time difference of arrival (Δt), and is used to construct the hyperbolic positioning equation. This feature reflects the spatial geometric relationship between the fault location and the monitoring point, and is the core input parameter for positioning.

[0172] The hyperbolic location equations are a set of mathematical equations constructed with monitoring points as the foci. Each hyperbola represents the trajectory of a possible fault point corresponding to the time difference between the arrival times of a traveling wave at two monitoring points. The formulas for the hyperbolic location equations are as follows:

[0173]

[0174] Among them, (x p ,y p (x,y) refers to the coordinates of monitoring point p (unit: m), (x,y) refers to the coordinates of the fault point (the target to be solved), and Δd refers to the spatial distance difference.

[0175] The set of intersection points refers to the solution set of a system of hyperbolic equations, that is, the intersection points of multiple hyperbolas on a plane, representing the theoretically calculated values ​​of candidate fault locations.

[0176] The power grid coordinate system refers to a spatial reference system for power distribution networks based on GIS, which maps electrical nodes into two-dimensional / three-dimensional coordinates for visualizing positioning results.

[0177] The initial candidate region refers to a continuous polygonal region generated by the convex hull algorithm from the set of hyperbola intersection points, covering all possible fault points.

[0178] The target candidate region refers to the precise range narrowed down after waveform matching and time difference filtering of the initial candidate region, excluding false intersections caused by line branches.

[0179] Minimizing the time-of-arrival residual of a traveling wave refers to optimizing the objective function by minimizing the sum of squared errors between the theoretical and measured arrival times (minΣ(Δt)). 2 This improves positioning accuracy.

[0180] Particle swarm optimization is an intelligent algorithm that iteratively searches for the optimal solution within a target candidate region. The particle position represents the candidate fault point, and the fitness function is the time residual.

[0181] First, the energy data platform system obtains the latitude and longitude coordinates (WGS84 standard) and line electrical parameters (R, L, C, G) of all monitoring points in real time by calling the RESTful API interface of the power grid topology database, and uses CRC-32 verification to ensure data integrity. The system automatically matches a preset wave velocity calculation model according to the line type: a modified speed of light model is used for overhead lines, and a frequency-varying model is used for cables. Simultaneously, an attenuation correction function is constructed based on line impedance and resistance to achieve accurate modeling of the traveling wave propagation path. This eliminates the influence of line parameter differences on wave velocity and attenuation, ensuring subsequent positioning accuracy.

[0182] Secondly, the system analyzes the three-dimensional structure (time × space × feature) of the fault feature tensor, extracts the arrival timestamps of the traveling waves at each monitoring point from the time dimension (based on GPS PTP protocol synchronization, accuracy ±50ns), and separates waveform distortion parameters from the feature dimension: obtaining the harmonic distortion rate (THD) in the 0.1-1MHz frequency band through wavelet packet decomposition, calculating the wavefront steepness (dV / dt) using the five-point difference method, and recording the amplitude attenuation of the first wave peak. This process is accelerated by FPGA to achieve microsecond-level feature extraction, meeting real-time requirements.

[0183] Then, the system constructs a hyperbolic equation for each pair of monitoring points: using the latitude and longitude of the two monitoring points as foci, the WGS84 coordinates are converted to a local UTM Cartesian coordinate system, and a system of equations is established. The Levenberg-Marquardt algorithm is used to iteratively solve the system of equations, with the convergence condition set to residual <1e-6. The solution set is then mapped back to the geographic coordinate system using a convex hull algorithm (implemented by Graham scan) to form an initial candidate region polygon. Introducing convex hull processing can eliminate isolated spurious solutions caused by measurement errors.

[0184] Next, the system performs a three-level matching and screening process: 1) Time difference consistency check, eliminating candidate points with residuals exceeding 1μs; 2) Waveform distortion matching, calculating the normalized cross-correlation coefficient (NCC) between the measured waveform and the theoretical waveform (expected waveform after attenuation correction) of the candidate point, retaining points with NCC > 0.9; 3) Spatial topology verification, combining GIS route path to exclude candidate points more than 5m away from the nearest route. This screening strategy can reduce the false alarm rate by more than 90%.

[0185] Finally, the system initializes the particle swarm in the selected target region (usually a 50m × 50m rectangle): 100 particles are set, randomly distributed within the target region, with a velocity constraint of ±5m / iteration. The residual objective function is designed as follows:

[0186]

[0187] Where J(x,y) refers to the objective function (sum of squared time residuals), d p (x,y) refers to the spatial distance from the fault point (x,y) to the monitoring point p, N refers to the total number of monitoring points, and t p t refers to the measured arrival time of the traveling wave at the p-th monitoring point, and v refers to the propagation speed of the traveling wave.

[0188] After each iteration, the global optimal solution is updated. The process terminates when the optimal solution changes by less than 0.1m for 10 consecutive iterations, and the final fault point coordinates are output (format: longitude ±0.5m, latitude ±0.5m).

[0189] As an example, the step of filtering the initial candidate region based on the traveling wave arrival time series, the waveform distortion features, the velocity calculation function, and the attenuation correction function to obtain the target candidate region includes: converting the traveling wave arrival time series into spatial distance difference features according to the velocity calculation function; calculating the theoretical distortion value of the initial candidate region according to the attenuation correction function; performing similarity matching between the theoretical distortion value and the waveform distortion features to obtain a matching result; and filtering the target candidate region from the initial candidate region according to the matching result.

[0190] The theoretical distortion value refers to the expected waveform distortion calculated based on the attenuation correction function, including amplitude attenuation rate and phase shift. This value is used to compare with the measured waveform distortion characteristics to verify the rationality of the candidate region. The formula for calculating the theoretical distortion value is as follows:

[0191]

[0192] in, This refers to the theoretical distortion value of candidate point k, d k This refers to the distance from candidate point k to the monitoring point, D0 refers to the original distortion rate of the fault point (assumed to be 1), and α(f k ) refers to the frequency-dependent attenuation coefficient function (see α(f) above), f k It refers to the characteristic frequency component of the traveling wave signal, that is, the kth dominant frequency in the fault traveling wave.

[0193] The matching result refers to the quantitative score output by the similarity matching algorithm (such as normalized cross-correlation coefficient NCC or dynamic time warping DTW), ranging from [0,1], which represents the degree of agreement between theoretical distortion and measured distortion, and is used to screen high-confidence candidate points.

[0194] First, the energy data platform system calls a speed calculation function to convert the arrival timestamps of the precisely synchronized traveling waves at each monitoring point into corresponding spatial distance differences, while dynamically compensating for wave speed differences in different line sections. Specifically, the system automatically loads a preset wave speed correction coefficient based on the line type (overhead or cable) and dynamically adjusts the calculation model in conjunction with real-time collected temperature and humidity data to ensure that the distance conversion error is controlled within 0.1%. This is done to eliminate the influence of environmental factors on wave speed and improve the accuracy of subsequent spatial positioning. Second, based on the line attenuation characteristics, the system calculates the expected waveform distortion pattern for each candidate fault point. Specific operations include: 1) calculating the theoretical amplitude attenuation curve based on the distance from the fault point to each monitoring point; 2) predicting the distortion characteristics of each frequency band of the waveform using a line parameter model; and 3) generating a theoretical waveform template containing two-dimensional time-frequency information.

[0195] Then, the system performs waveform similarity matching. The specific implementation process is as follows: 1) Synchronously align the measured waveform and the theoretical waveform to ensure consistency of the time axis; 2) Use sliding window technology to divide the waveform into multiple analysis segments; 3) Compare the waveform shape, extreme point position, and trend within each window; 4) Combine the similarity scores of each window and obtain the overall matching degree through weighted average. The key to this design is that segmented comparison can capture both overall features and identify local anomalies. Finally, the system filters target candidate regions based on the matching results. The specific steps are: 1) Set a dynamic matching degree threshold (initial value is 0.85); 2) Retain all candidate points with matching degrees exceeding the threshold; 3) Analyze the spatial distribution of candidate points using the density clustering algorithm (DBSCAN); 4) Remove isolated points and retain regions forming dense clusters; 5) Calculate the geometric center of the cluster region as the final target region. This filtering strategy can effectively eliminate interference signals and improve the reliability of the positioning results. The entire processing flow is completed within 200ms, meeting the requirements of real-time fault diagnosis.

[0196] This embodiment provides a multi-source heterogeneous data fusion processing method based on an energy data platform. First, the energy data platform system acquires three-phase current waveform data, circuit breaker coil current data, and load switch status signals from the power distribution terminal, providing comprehensive and real-time equipment operating status data for subsequent analysis. Next, the system analyzes the circuit breaker coil current data, extracting the circuit breaker's operating characteristics and forming an operating feature vector. This step, by quantifying the circuit breaker's operational performance, provides crucial information for fault early warning. Then, the system constructs a switch operating timing constraint matrix using the load switch status signal and the operating feature vector. By establishing a timing correlation between switch operations and circuit breaker operations, it provides clues for fault detection. Subsequently, the system performs fusion analysis on the three-phase current waveform data based on the switch operating timing constraint matrix to obtain a fault feature tensor. This fusion process more comprehensively reflects the system status, improving the efficiency and accuracy of data processing. Finally, the system performs spatial location analysis on the fault feature tensor based on a traveling wave propagation model to obtain the fault point coordinates. This embodiment can accurately locate fault points in power supply lines under low-cost equipment conditions, providing support for rapid fault repair.

[0197] Based on the first embodiment of this application, in the second embodiment of this application, the content that is the same as or similar to that in Embodiment 1 above can be referred to the above description, and will not be repeated hereafter. Based on this, please refer to... Figure 2 , Figure 2 This is a flowchart illustrating the second embodiment of the multi-source heterogeneous data fusion processing method based on an energy data platform according to this application. Step S20 of the multi-source heterogeneous data fusion processing method based on an energy data platform includes steps S21 to S25:

[0198] Step S21: Use multi-threshold gradient analysis to identify multiple independent current pulses in the circuit breaker coil current data, and obtain the rising edge boundary point and falling edge boundary point of the current pulse.

[0199] It should be noted that the multi-threshold gradient analysis method is a pulse detection algorithm that combines dynamic thresholds and gradient changes. By setting gradient thresholds for different amplitude ranges (such as the 10%-90% amplitude range on the rising edge and the 90%-10% amplitude range on the falling edge), it can accurately capture the start and end positions of the current pulse and avoid misjudgments caused by noise interference.

[0200] The rising edge boundary point refers to the initial moment when the current pulse rises from the baseline (steady-state value) to a preset low threshold (e.g., 10% peak value), reflecting the initial energization of the circuit breaker's operating coil. The formula for detecting the rising edge boundary point is as follows:

[0201]

[0202] Among them, I c (t) refers to the circuit breaker coil current (unit: A); t refers to time (unit: s); γ refers to the preset current change rate threshold (unit: A / s). The t that satisfies this formula is the rising edge boundary point.

[0203] The falling edge boundary point refers to the point at which the current pulse drops from its peak value to a preset low threshold value, representing the moment when the circuit breaker's operating coil is de-energized.

[0204] Understandably, firstly, the energy data platform system preprocesses the raw circuit breaker coil current data, using a Savitzky-Golay filter with a window length of 21 points for smoothing. This preserves the steep edge characteristics of the pulses while suppressing high-frequency noise (interference >2kHz can be eliminated at a sampling rate of 10kHz). This is because electromagnetic interference from the coil current is severe during circuit breaker operation, and it is essential to ensure that the pulse characteristics are not overwhelmed by noise. Secondly, the system performs multi-threshold gradient detection: 1) A 5ms sliding window is used when calculating the first-order difference sequence, with the rising edge detection threshold set at 0.5A / ms (to avoid false triggering of electromagnetic interference) and the falling edge threshold at -0.3A / ms (considering the arc sustaining current); 2) Within the initially marked pulse interval, the time resolution is improved to 1μs through third-order spline interpolation, and the zero-crossing point of the 10%-90% amplitude range is located as the precise boundary; 3) The rise / fall time ratio of each pulse is checked (normal value 0.8-1.2), and distorted pulses caused by line induction are removed. The dynamic threshold strategy can be adapted to the characteristics of different types of circuit breakers.

[0205] Then, the system performs pulse characteristic verification: 1) Establishing a decision tree model including peak current (2-15A), rise time (0.5-5ms), and pulse width (5-20ms) to automatically identify and eliminate abnormal pulses exceeding the normal range; 2) Eliminating duplicated pseudo-pulses through adjacent pulse interval analysis (normal >100ms); 3) Recording the boundary timestamps of verified pulses, with time synchronization error controlled within ±20μs (based on IEEE 1588 clock synchronization). This verification process ensures that each pulse truly reflects the circuit breaker operation. Finally, the system outputs structured pulse data: each valid pulse includes a rising edge timestamp (10% amplitude point), a falling edge timestamp (10% amplitude point), a peak current value, and a pulse quality score (0-100%).

[0206] Step S22: Calculate the pulse width characteristics, current amplitude extreme value characteristics, and pulse morphology steepness characteristics of the current pulse based on the rising edge boundary point and the falling edge boundary point.

[0207] It should be noted that the pulse width characteristic refers to the time span (unit: milliseconds) from the rising edge boundary point (10% amplitude) to the falling edge boundary point (10% amplitude) of the current pulse. It reflects the mechanical action speed of the circuit breaker operating mechanism and is used to determine whether the opening and closing is stuck or timed out.

[0208] The extreme value characteristic of current amplitude refers to the peak current intensity (unit: ampere) in the pulse waveform, which characterizes the magnitude of the electromagnetic driving force of the circuit breaker operating coil. Abnormal values ​​may indicate a short circuit in the coil or a power supply failure.

[0209] Pulse morphology steepness refers to the rate of change of the slope of the quantized pulse rise / fall times (unit: A / ms). 2 The second derivative is obtained by calculating the second derivative of the amplitude range of the rising edge from 20% to 80%, which reflects the dynamic response characteristics of the operating mechanism and is used to detect mechanical wear or poor lubrication.

[0210] As an example, the step of calculating the pulse width feature, current amplitude extreme value feature, and pulse shape steepness feature of the current pulse based on the rising edge boundary point and the falling edge boundary point includes: calculating the pulse width feature of the current pulse based on the rising edge boundary point and the falling edge boundary point; taking the maximum value of the current value sequence in the current pulse as the current amplitude extreme value feature; calculating the average rising gradient based on the current value sequence between the rising edge boundary point and the maximum value; calculating the average falling gradient based on the current value sequence between the falling edge boundary point and the maximum value; and calculating the pulse shape steepness feature of the current pulse based on the average rising gradient and the average falling gradient.

[0211] A current value sequence refers to a set of discrete sampled values ​​of the circuit breaker coil current within a single pulse time interval, arranged in chronological order, with a sampling rate ≥10kHz, and including the entire process data including the rising edge, peak value, and falling edge.

[0212] The average rise gradient refers to the rate of change of current (ΔI / Δt) per unit time from the rising edge boundary point to the peak point of a current pulse. It is calculated by fitting the rising segment data points through linear regression and reflects the starting acceleration of the operating mechanism. The formula for calculating the average rise gradient is as follows:

[0213]

[0214] Where, k r This refers to the average slope of the ascending segment, i.e., the average gradient of ascent; I max This refers to the peak value of the pulse current, i.e., the extreme value characteristic of the current amplitude; t p This refers to the current reaching I. max Time; t r This refers to the time at the rising edge boundary point; I c (t r ) refers to t r Current value at any given time.

[0215] The average descent gradient refers to the rate of change of current per unit time from the peak point to the descent boundary point of the current pulse. It is calculated by fitting the descent segment using an exponential decay model and characterizes the energy dissipation rate of the arc extinguishing process.

[0216] First, the system uses the rising edge boundary point and falling edge boundary point as a reference to extract the time difference between them as a pulse width feature. Simultaneously, a high-precision clock counter (1μs resolution) is used for calibration to ensure that the time measurement error is less than ±0.1ms. This is done because the mechanical action time of the circuit breaker directly reflects the health status of its operating mechanism; accurate measurement can effectively detect jamming or delayed faults. Second, the system scans the current value sequence, locates the maximum value using a sliding extreme value detection algorithm (window width of 5 sampling points), and corrects the peak current value using cubic spline interpolation. Finally, it records the current amplitude extreme value feature, thus eliminating ADC (analog-to-digital converter) quantization error and ensuring peak detection accuracy of ±0.5A.

[0217] Then, the system performs least-squares linear fitting on the current sequence during the rising edge phase (from the rising edge boundary point to the peak point) and calculates the average slope as the average rising gradient. The falling edge phase (from the peak point to the falling edge boundary point) is handled symmetrically, and an exponential decay model is used to fit the average falling gradient. Finally, the system divides the sum of the absolute values ​​of the average rising gradient and the average falling gradient by the peak current value to obtain a dimensionless pulse morphology steepness feature, which is then normalized and converted into a standard score of 0-100%. This feature can sensitively reflect the dynamic response characteristics of the circuit breaker operating mechanism and is used for early mechanical fault warning.

[0218] Step S23: Detect the continuous latch-up time characteristics and adjacent pulse interval characteristics among all the current pulses.

[0219] It should be noted that the continuous interlocking time characteristic refers to the duration for which the drive coil remains current-free after the circuit breaker completes one opening or closing operation. This reflects the time required for the mechanical reset or cooling of the operating mechanism. An abnormally short duration may indicate mechanical jamming or a control circuit malfunction. The formula for calculating the continuous interlocking time is as follows:

[0220]

[0221] in, This refers to the interlocking time between the i-th and i+1-th pulses. This refers to the falling edge time of the i-th pulse. It refers to the rising time of the (i+1)th pulse.

[0222] The adjacent pulse interval characteristic refers to the time difference between the rising edges of two consecutive current pulses. It is used to analyze the frequency characteristics of circuit breaker operation. If the interval is too short, it may trigger the protection blocking logic.

[0223] Understandably, firstly, the system iterates through all identified current pulse sequences, extracts the falling edge termination timestamp of each pulse and the rising edge start timestamp of the next pulse, and calculates the difference between the two using a high-precision clock counter as a continuous blocking time feature. Secondly, the system constructs a sequence according to the pulse time order and calculates the time difference between the start times of the rising edges of adjacent pulses as an adjacent pulse interval feature.

[0224] Step S24: Based on the pulse width feature and the adjacent pulse interval feature, a set of effective pulses is selected from the current pulses.

[0225] It should be noted that the effective pulse set refers to the set of legal circuit breaker operation pulses selected by the preset action rules, which must simultaneously meet the following: (1) Pulse width characteristics: located within the preset action time window (e.g., 300-450ms for opening operation, 200-350ms for closing operation). (2) Adjacent pulse interval characteristics: conforming to the action continuity rules (e.g., continuous opening interval ≥ 100ms, opening delay after closing ≥ 50ms).

[0226] Understandably, the system iterates through all current pulses, compares the pulse width characteristics of each pulse with the preset action time window, and simultaneously checks whether the interval characteristics of adjacent pulses conform to the continuity rule; only when a pulse meets both the width condition and the interval condition is the pulse retained, and finally all the pulses that meet the requirements are encapsulated into a structured effective pulse set according to the time sequence and output.

[0227] Step S25: The current amplitude extreme value feature, the pulse shape steepness feature, and the continuous blocking time feature of the current pulse in the effective pulse set are encoded into a preset dimension by a segmented feature compression algorithm to obtain the action feature vector.

[0228] It should be noted that the Segmented Feature Compression Algorithm (SFCA) is a data processing technique used to simplify and compress complex signal features. In this embodiment, it is used to encode pulse feature sequences of different lengths into fixed-dimensional vectors, including three steps: feature normalization, principal component analysis dimensionality reduction, and segmented quantization encoding, to ensure that the circuit breaker pulse features of different operation counts can be uniformly expressed.

[0229] The preset dimension refers to the feature vector length (such as 64-dimensional or 128-dimensional) pre-set according to the circuit breaker model. It is determined through offline training and optimization to minimize storage space while retaining 95% of the original information, thus meeting the requirements of real-time processing.

[0230] Understandably, the system first extracts the extreme value features of current amplitude (floating-point number), pulse morphology steepness features (percentage), and continuous latch-up time features (milliseconds) from each effective pulse. Then, min-max normalization is used to scale each feature to the [0,1] interval. Principal component analysis is then used to reduce the dimensionality of the normalized feature matrix, retaining 95% of the variance components. Next, a segmented quantization encoder converts the dimensionality-reduced data into an 8-bit integer sequence: the floating-point feature values ​​are discretized to 256 levels with a precision of 0.01. A 512-bit Bloom filter code is generated for each segment of 32 pulses, and the code sequence is padded with zeros according to a preset 64-dimensional length. Finally, a standardized 64-dimensional motion feature vector is output.

[0231] This embodiment first employs multi-threshold gradient analysis to identify multiple independent current pulses in the circuit breaker coil current data and obtains the rising and falling edge boundary points of these pulses, providing an accurate time reference for subsequent feature extraction. Next, the system calculates the pulse width characteristics, current amplitude extreme value characteristics, and pulse kurtosis characteristics of the current pulses based on these boundary points. These characteristics comprehensively reflect the shape and intensity of the current pulses, providing crucial information for subsequent fault diagnosis. Then, the system detects the continuous blocking time characteristics and adjacent pulse interval characteristics among all current pulses, which helps to understand the timing relationships between pulses and further reveal the operating characteristics of the circuit breaker. Finally, based on the pulse width characteristics and adjacent pulse interval characteristics, the system filters out the effective pulse set from the current pulses, removing noise and interference pulses to improve the accuracy of the analysis. Finally, the current amplitude extreme value features, pulse morphology steepness features, and continuous blocking time features of the current pulses in the effective pulse set are encoded into action feature vectors of a preset dimension by a segmented feature compression algorithm. This simplifies the complex pulse features into vectors of a fixed dimension, which facilitates subsequent processing and analysis, improves computational efficiency, and ultimately achieves an efficient and accurate description of the circuit breaker's operating characteristics.

[0232] For example, to help understand the implementation process of the multi-source heterogeneous data fusion processing method based on the energy data platform obtained by combining this embodiment with the above embodiment one, please refer to... Figure 3 , Figure 3 A simplified flowchart of a multi-source heterogeneous data fusion processing method based on an energy data platform is provided, specifically:

[0233] This diagram illustrates the complete workflow of a multi-source heterogeneous data fusion processing method based on an energy data platform. It begins by acquiring three-phase current waveform data, circuit breaker coil current data, and load switch status signals from the distribution terminal. This data is collected by a data acquisition module and then cleaned and pre-processed by a data preprocessing module. The pre-processed data is divided into two parts: one part is sent to a circuit breaker action analysis module for further analysis of the circuit breaker's operating characteristics and extraction of action feature vectors; the other part is sent to a timing constraint construction module to construct a switch action timing constraint matrix in conjunction with the load switch status signals. The action feature vectors and timing constraint matrix are then input into a fusion analysis module, combined with the pre-processed three-phase current waveform data, to generate a fault feature tensor through complex fusion analysis. Finally, the fault feature tensor is sent to a traveling wave propagation model module. This module obtains monitoring point location information and line electrical parameters from the power grid topology database, performs spatial location analysis of the fault feature tensor based on the traveling wave propagation model, and ultimately outputs the fault point coordinates. The entire process is interconnected, encompassing the collection, preprocessing, feature extraction, fusion analysis, and fault location of multi-source data, aiming to achieve efficient and accurate power system fault diagnosis.

[0234] This application also provides a multi-source heterogeneous data fusion processing device based on an energy data platform. Please refer to [link / reference]. Figure 4 The multi-source heterogeneous data fusion processing device based on the energy data platform includes:

[0235] Data acquisition module 10 is used to acquire three-phase current waveform data, circuit breaker coil current data and load switch status signals of the power distribution terminal;

[0236] Action analysis module 20 is used to analyze the action characteristics of the circuit breaker based on the circuit breaker coil current data and obtain action feature vector;

[0237] The constraint construction module 30 is used to construct a switch action timing constraint matrix based on the load switch status signal and the action feature vector;

[0238] The fusion analysis module 40 is used to perform fusion analysis on the three-phase current waveform data according to the switching action timing constraint matrix to obtain the fault feature tensor;

[0239] The positioning and analysis module 50 is used to perform spatial positioning and analysis on the fault feature tensor based on the traveling wave propagation model to obtain the coordinates of the fault point.

[0240] The multi-source heterogeneous data fusion processing device based on an energy data platform provided in this application, employing the multi-source heterogeneous data fusion processing method based on an energy data platform as described in the above embodiments, can solve the technical problem of accurately locating fault points in power supply lines under low-cost equipment conditions. Compared with the prior art, the beneficial effects of the multi-source heterogeneous data fusion processing device based on an energy data platform provided in this application are the same as those of the multi-source heterogeneous data fusion processing method based on an energy data platform provided in the above embodiments, and other technical features in the multi-source heterogeneous data fusion processing device based on an energy data platform are the same as those disclosed in the methods of the above embodiments, and will not be repeated here.

[0241] This application provides a multi-source heterogeneous data fusion processing device based on an energy data platform. The multi-source heterogeneous data fusion processing device based on an energy data platform includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the multi-source heterogeneous data fusion processing method based on the energy data platform in the first embodiment described above.

[0242] The following is for reference. Figure 5This document illustrates a structural schematic diagram of a multi-source heterogeneous data fusion processing device based on an energy data platform, suitable for implementing embodiments of this application. The multi-source heterogeneous data fusion processing device based on an energy data platform in this application embodiment may include, but is not limited to, mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), and in-vehicle terminals (e.g., in-vehicle navigation terminals), as well as fixed terminals such as digital TVs and desktop computers. Figure 5 The multi-source heterogeneous data fusion processing device based on the energy data platform shown is merely an example and should not impose any limitations on the functionality and scope of use of the embodiments of this application.

[0243] like Figure 5 As shown, the multi-source heterogeneous data fusion processing device based on the energy data platform may include a processing unit 1001 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to the program stored in ROM (Read Only Memory) 1002 or the program loaded from storage device 1003 into RAM (Random Access Memory) 1004. The RAM 1004 also stores various programs and data required for the operation of the multi-source heterogeneous data fusion processing device based on the energy data platform. The processing unit 1001, ROM 1002, and RAM 1004 are interconnected via bus 1005. Input / output (I / O) interface 1006 is also connected to the bus. Typically, the following systems can be connected to I / O interface 1006: input devices 1007 including, for example, touchscreens, touchpads, keyboards, mice, image sensors, microphones, accelerometers, gyroscopes, etc.; output devices 1008 including, for example, LCDs (Liquid Crystal Displays), speakers, vibrators, etc.; storage devices 1003 including, for example, magnetic tapes, hard disks, etc.; and communication devices 1009. Communication device 1009 allows the multi-source heterogeneous data fusion processing equipment based on the energy data platform to wirelessly or wiredly communicate with other devices to exchange data. Although the figure shows a multi-source heterogeneous data fusion processing equipment based on the energy data platform with various systems, it should be understood that it is not required to implement or possess all the systems shown. More or fewer systems can be implemented alternatively.

[0244] Specifically, according to the embodiments disclosed in this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments disclosed in this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device, or installed from storage device 1003, or installed from ROM 1002. When the computer program is executed by processing device 1001, it performs the functions defined in the methods of the embodiments disclosed in this application.

[0245] The multi-source heterogeneous data fusion processing device based on an energy data platform provided in this application, employing the multi-source heterogeneous data fusion processing method based on an energy data platform as described in the above embodiments, can solve the technical problem of accurately locating fault points in power supply lines under low-cost equipment conditions. Compared with the prior art, the beneficial effects of the multi-source heterogeneous data fusion processing device based on an energy data platform provided in this application are the same as those of the multi-source heterogeneous data fusion processing method based on an energy data platform provided in the above embodiments, and other technical features in this multi-source heterogeneous data fusion processing device based on an energy data platform are the same as those disclosed in the previous embodiment method, and will not be repeated here.

[0246] It should be understood that the various parts disclosed in this application can be implemented using hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples.

[0247] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0248] This application provides a computer-readable storage medium having computer-readable program instructions (i.e., a computer program) stored thereon, which are used to execute the multi-source heterogeneous data fusion processing method based on the energy data platform in the above embodiments.

[0249] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, RAM (Random Access Memory), ROM (Read Only Memory), EPROM (Erasable Programmable Read Only Memory or Flash Memory), optical fibers, CD-ROM (CD-Read Only Memory), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0250] The aforementioned computer-readable storage medium may be included in a multi-source heterogeneous data fusion processing device based on an energy data platform; or it may exist independently and not be assembled into a multi-source heterogeneous data fusion processing device based on an energy data platform.

[0251] The aforementioned computer-readable storage medium carries one or more programs. When these programs are executed by a multi-source heterogeneous data fusion processing device based on an energy data platform, the multi-source heterogeneous data fusion processing device based on the energy data platform performs the following actions: acquires three-phase current waveform data, circuit breaker coil current data, and load switch status signals from the distribution terminal; analyzes the circuit breaker operating characteristics based on the circuit breaker coil current data to obtain an operating feature vector; constructs a switch operating timing constraint matrix based on the load switch status signal and the operating feature vector; performs fusion analysis on the three-phase current waveform data based on the switch operating timing constraint matrix to obtain a fault feature tensor; and performs spatial location analysis on the fault feature tensor based on a traveling wave propagation model to obtain the fault point coordinates.

[0252] Computer program code for performing the operations of this application can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, as well as conventional procedural programming languages ​​such as "C" or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including LAN (Local Area Network) or WAN (Wide Area Network)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0253] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0254] The modules described in the embodiments of this application can be implemented in software or hardware. The names of the modules do not necessarily limit the functionality of the unit itself.

[0255] The readable storage medium provided in this application is a computer-readable storage medium that stores computer-readable program instructions (i.e., a computer program) for executing the above-described multi-source heterogeneous data fusion processing method based on an energy data platform. This solves the technical problem of accurately locating fault points in power supply lines under low-cost equipment conditions. Compared with the prior art, the beneficial effects of the computer-readable storage medium provided in this application are the same as those of the multi-source heterogeneous data fusion processing method based on an energy data platform provided in the above embodiments, and will not be repeated here.

[0256] The above description is only a part of the embodiments of this application and does not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A method for multi-source heterogeneous data fusion processing based on an energy data platform, characterized in that, The method includes: Acquire the three-phase current waveform data, circuit breaker coil current data, and load switch status signals of the power distribution terminal; Based on the circuit breaker coil current data, the circuit breaker's operating characteristics are analyzed to obtain an operating feature vector. Construct a switch action timing constraint matrix based on the load switch status signal and the action feature vector; Based on the switching action timing constraint matrix, the three-phase current waveform data are fused and analyzed to obtain the fault feature tensor; Based on the traveling wave propagation model, the fault feature tensor is spatially located and analyzed to obtain the coordinates of the fault point. The step of analyzing the circuit breaker's operating characteristics based on the circuit breaker coil current data to obtain the operating feature vector includes: Multiple independent current pulses in the circuit breaker coil current data are identified using a multi-threshold gradient analysis method, and the rising edge boundary point and falling edge boundary point of the current pulse are obtained. Based on the rising edge boundary point and the falling edge boundary point, calculate the pulse width characteristics, current amplitude extreme value characteristics, and pulse shape steepness characteristics of the current pulse; Detect the continuous latch-up time characteristics and adjacent pulse interval characteristics among all the current pulses; Based on the pulse width characteristics and the adjacent pulse interval characteristics, a set of effective pulses is obtained from the current pulses; The current amplitude extreme value feature, pulse shape steepness feature, and continuous blocking time feature of the current pulse in the effective pulse set are encoded into a preset dimension by a segmented feature compression algorithm to obtain an action feature vector; The step of constructing the switch action timing constraint matrix based on the load switch status signal and the action feature vector includes: Analyze the state transition time points in the load switch status signal to form a switch event time sequence; Extract the rising edge start time of all current pulses from the action feature vector to form the circuit breaker action time sequence; The time series of the switching events and the time series of the circuit breaker actions are paired according to time to obtain a set of event pairs; Calculate the absolute time difference between the switching event time and the circuit breaker operating time for each event pair in the event pair set; When the absolute time difference is greater than a preset tolerance threshold, an exception identifier is set for the event pair corresponding to the absolute time difference. Construct a binary association matrix with switch events as rows and circuit breaker actions as columns, and assign values ​​to the matrix elements of the binary association matrix according to the exception identifier; The spatiotemporal correlation normalization process is performed on the binary correlation matrix to obtain the switching action timing constraint matrix; The step of fusing and analyzing the three-phase current waveform data based on the switching action timing constraint matrix to obtain the fault feature tensor includes: Bispectral analysis was performed on the three-phase current waveform data to extract the signal phase coupling characteristics in each frequency band, and a bispectral feature matrix was constructed based on the signal phase coupling characteristics. The switching action timing constraint matrix is ​​sliced ​​along the time dimension to obtain multiple timing constraint submatrices; The bispectral feature matrix is ​​aligned with the temporal constraint submatrix in a time-frequency manner. Once the time-frequency alignment is complete, the correlation weights between the bispectral feature matrix and the temporal constraint submatrix are calculated using a cross-domain coupling algorithm. The bispectral feature matrix is ​​reconstructed by weighting according to the correlation weights to form an initial three-dimensional feature space; The initial three-dimensional feature space is reduced and normalized by tensor to obtain the fault feature tensor; The step of performing spatial localization analysis on the fault feature tensor based on the traveling wave propagation model to obtain the fault point coordinates includes: Obtain monitoring point location information and line electrical parameters from a pre-defined power grid topology database; A traveling wave propagation model is constructed based on the electrical parameters of the line. The traveling wave propagation model includes a velocity calculation function and an attenuation correction function. Based on the location information of the monitoring points, the traveling wave arrival time sequence and waveform distortion features of each monitoring point are extracted from the fault feature tensor. A hyperbolic positioning equation system with the monitoring point as the focus is constructed, and the hyperbolic parameters of the hyperbolic positioning equation system are determined by the spatial distance difference characteristics; Solving the hyperbolic positioning equations yields a set of intersection points, which are then mapped onto the power grid coordinate system to form an initial candidate region. Based on the traveling wave arrival time sequence, the waveform distortion characteristics, the velocity calculation function, and the attenuation correction function, the initial candidate region is screened for matching degree to obtain the target candidate region; With the goal of minimizing the travel wave arrival time residual, particle swarm optimization is performed in the target candidate region to obtain the fault point coordinates.

2. The method as described in claim 1, characterized in that, The step of calculating the pulse width characteristics, current amplitude extreme value characteristics, and pulse shape steepness characteristics of the current pulse based on the rising edge boundary point and the falling edge boundary point includes: The pulse width characteristics of the current pulse are calculated based on the rising edge boundary point and the falling edge boundary point; The maximum value of the current value sequence in the current pulse is taken as the extreme value feature of the current amplitude; The average rising gradient is calculated based on the current value sequence between the rising edge boundary point and the maximum value. The average descent gradient is calculated based on the current value sequence between the descent edge boundary point and the maximum value. The pulse morphology steepness characteristics of the current pulse are calculated based on the average rising gradient and the average falling gradient.

3. The method as described in claim 1, characterized in that, The step of performing matching degree screening on the initial candidate region based on the traveling wave arrival time series, the waveform distortion characteristics, the velocity calculation function, and the attenuation correction function to obtain the target candidate region includes: Based on the velocity calculation function, the arrival time series of the traveling wave is converted into spatial distance difference features; The theoretical distortion value of the initial candidate region is calculated based on the attenuation correction function. The theoretical distortion value is matched with the waveform distortion feature to obtain the matching result; The target candidate region is obtained by filtering from the initial candidate region based on the matching results.

4. A multi-source heterogeneous data fusion processing device based on an energy data platform, characterized in that, The device includes: The data acquisition module is used to acquire three-phase current waveform data, circuit breaker coil current data and load switch status signals of the power distribution terminal; The action analysis module is used to analyze the action characteristics of the circuit breaker based on the circuit breaker coil current data to obtain an action feature vector. The steps of analyzing the circuit breaker action characteristics based on the circuit breaker coil current data to obtain the action feature vector include: using a multi-threshold gradient analysis method to identify multiple independent current pulses in the circuit breaker coil current data, and obtaining the rising edge boundary point and falling edge boundary point of the current pulse; calculating the pulse width feature, current amplitude extreme value feature, and pulse shape steepness feature of the current pulse based on the rising edge boundary point and the falling edge boundary point; detecting the continuous blocking time feature and adjacent pulse interval feature between all the current pulses; filtering out a valid pulse set from the current pulses based on the pulse width feature and the adjacent pulse interval feature; and encoding the current amplitude extreme value feature, pulse shape steepness feature, and continuous blocking time feature of the current pulses in the valid pulse set into a preset dimension using a segmented feature compression algorithm to obtain the action feature vector. A constraint construction module is used to construct a switch action timing constraint matrix based on the load switch state signal and the action feature vector. The step of constructing the switch action timing constraint matrix based on the load switch state signal and the action feature vector includes: parsing the state transition time points in the load switch state signal to form a switch event time sequence; extracting the rising edge start time points of all current pulses from the action feature vector to form a circuit breaker action time sequence; pairing the switch event time sequence with the circuit breaker action time sequence by time to obtain an event pair set; calculating the absolute time difference between the switch event time point and the circuit breaker action time point for each event pair in the event pair set; setting an anomaly identifier for the event pair corresponding to the absolute time difference when the absolute time difference is greater than a preset tolerance threshold; constructing a binary correlation matrix with switch events as rows and circuit breaker actions as columns, and assigning values ​​to the matrix elements of the binary correlation matrix according to the anomaly identifier; and performing spatiotemporal correlation normalization processing on the binary correlation matrix to obtain the switch action timing constraint matrix. The fusion analysis module is used to perform fusion analysis on the three-phase current waveform data according to the switching action timing constraint matrix to obtain a fault feature tensor. The steps of performing fusion analysis on the three-phase current waveform data according to the switching action timing constraint matrix to obtain the fault feature tensor include: performing bispectral analysis on the three-phase current waveform data to extract signal phase coupling features in each frequency band, and constructing a bispectral feature matrix based on the signal phase coupling features; slicing the switching action timing constraint matrix in the time dimension to obtain multiple timing constraint sub-matrices; aligning the bispectral feature matrix with the timing constraint sub-matrices in time and frequency; calculating the correlation weights between the bispectral feature matrix and the timing constraint sub-matrices using a cross-domain coupling algorithm after the time and frequency alignment is completed; reconstructing the bispectral feature matrix with weights based on the correlation weights to form an initial three-dimensional feature space; and performing tensor dimensionality reduction and standardization on the initial three-dimensional feature space to obtain the fault feature tensor. The location analysis module is used to perform spatial location analysis on the fault feature tensor based on the traveling wave propagation model to obtain the fault point coordinates. The steps of performing spatial location analysis on the fault feature tensor based on the traveling wave propagation model to obtain the fault point coordinates include: obtaining monitoring point location information and line electrical parameters from a preset power grid topology database; constructing a traveling wave propagation model based on the line electrical parameters, the traveling wave propagation model including a velocity calculation function and an attenuation correction function; extracting the traveling wave arrival time series and waveform distortion features of each monitoring point from the fault feature tensor based on the monitoring point location information; constructing a hyperbolic location equation system with the monitoring point as the focus, the hyperbolic parameters of the hyperbolic location equation system being determined by spatial distance difference features; solving the hyperbolic location equation system to obtain an intersection set, and mapping the intersection set to the power grid coordinate system to form an initial candidate region; performing matching degree screening on the initial candidate region based on the traveling wave arrival time series, the waveform distortion features, the velocity calculation function, and the attenuation correction function to obtain a target candidate region; and performing particle swarm optimization in the target candidate region with the goal of minimizing the traveling wave arrival time residual to obtain the fault point coordinates.

5. A multi-source heterogeneous data fusion processing device based on an energy data platform, characterized in that, The device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the multi-source heterogeneous data fusion processing method based on an energy data platform as described in any one of claims 1 to 3.

6. A storage medium, characterized in that, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the multi-source heterogeneous data fusion processing method based on the energy data platform as described in any one of claims 1 to 3.

Images