Primary and secondary fusion complete ring main unit fault diagnosis method and system

Abstract

The invention relates to the technical field of fault prediction and health management, in particular to a primary and secondary fusion complete ring main unit fault diagnosis method and system. Comprising the following steps: acquiring three-phase instantaneous voltage and current signals in real time, and converting the signals into digital transient data; performing time window preprocessing on the digital transient data, and executing wavelet packet decomposition to generate a transient feature vector; calculating and analyzing the transient feature vector through a transient zero-sequence power direction method and a support vector machine model to generate a local diagnosis result; when the local diagnosis result is that cooperative positioning needs to be started, a transient current similarity coefficient and a transient waveform intensity difference coefficient are calculated based on the digital transient data, and a comprehensive positioning result is generated; determining a fault line according to the comprehensive positioning result, and generating a remote control command; and executing a fault isolation operation based on the remote control command, and executing a PHM process. According to the method, the double local transient diagnosis and the cross-terminal cooperative positioning are deeply fused, so that the high-precision determination of the boundary of the fault section is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fault prediction and health management, in particular to a primary and secondary fusion integrated ring main unit fault diagnosis method and system. BACKGROUND

[0002] In the field of fault prediction and health management of ring main units, the existing technology mainly adopts centralized architecture for fault prediction and health management (PHM). Under this architecture, the main responsibility of the distribution terminal unit (DTU) configured inside the ring main unit is data acquisition and uploading. The terminal device only reports electrical information such as current and voltage to the master station system located in the substation or control center. The master station system is responsible for centralized data analysis, fault judgment, decision making, and issuing remote control commands to the terminal device to achieve fault isolation and recovery.

[0003] However, the existing fault line selection technology is fundamentally limited by the topology structure. Ring main units are applied to medium voltage distribution networks and belong to open-loop systems in electrical wiring. In widely used small current grounding systems (such as neutral point ungrounded or grounded through arc suppression coil), the traditional fault line selection principle (such as the principle based on steady-state zero sequence current) is designed for closed-loop systems such as main substations. Due to the open-loop topology structure of the ring main unit, the sum of the zero sequence currents of all outgoing lines is not zero, which makes any traditional method that only monitors the relationship between zero sequence currents or is based on steady-state quantities will draw incorrect diagnostic results in the scenarios of busbar fault, upstream fault or downstream fault of outgoing lines, forming a line selection blind area. Therefore, under the superimposed limitation of the existing centralized architecture and the open-loop topology of the ring main unit, the system cannot accurately select lines and independently isolate faults in the highest proportion of small current grounding faults.

[0004] Therefore, a primary and secondary fusion integrated ring main unit fault diagnosis method and system is proposed. SUMMARY

[0005] The purpose of the present application is to provide a primary and secondary fusion integrated ring main unit fault diagnosis method and system for fault diagnosis of ring main units.

[0006] To achieve the above purpose, the present application provides the following technical solutions: A primary and secondary fusion integrated ring main unit fault diagnosis method and system, comprising Real-time synchronous acquisition of three-phase instantaneous voltage and instantaneous current signals of each feeder port, and conversion into digitized transient data; Based on the time of fault occurrence, the digitized transient data is preprocessed by time window, and zero sequence waveform data is extracted; wavelet packet decomposition is performed on the zero sequence waveform data to generate a transient feature vector; The polarity result is generated by component extraction and integral calculation of the zero sequence waveform data through the transient zero sequence power direction method; the directivity index is output by regression analysis and weighted summation of the transient feature vector through the pre-trained support vector machine model; the local diagnosis result is generated based on the polarity result and the directivity index; When the local diagnosis result is to start cooperative positioning, the ring network box requests exchange of digitized transient data from adjacent ring network box terminals through the communication network; the transient current similarity coefficient and the transient waveform intensity difference coefficient are calculated based on the digitized transient data of the adjacent terminals, and the comprehensive positioning result is generated in combination with the local diagnosis result; The fault line is determined according to the comprehensive positioning result, and a remote control command is generated; the fault isolation operation is performed based on the remote control command, and the PHM process is executed.

[0007] Preferably, the specific generation process of the digitized transient data is as follows: real-time acquisition of three-phase instantaneous voltage and current signals is performed through the primary equipment sensors integrated by the ring network box; during the acquisition process, the three-phase instantaneous voltage and current signals are synchronized by a clock source at the hardware level and are digitized by an analog-to-digital converter to generate digitized transient data; the digitized transient data is stored in RAM and is locked and extracted by a fault transient triggering mechanism; the fault transient triggering mechanism is based on detection of line current, and when the absolute rate of change of zero sequence current and / or zero sequence voltage of any feeder port exceeds a threshold value, the fault transient triggering mechanism is triggered.

[0008] Preferably, the specific generation method of the transient feature vector includes: selecting a preset time window before and after the fault occurrence time; in the preset time window, superimposed calculation is performed on the three-phase instantaneous voltage and current to extract zero sequence waveform data containing instantaneous zero sequence voltage and zero sequence current; multi-layer wavelet packet decomposition is performed on the zero sequence waveform data, the multi-layer wavelet packet decomposition uses a mother wavelet base and sets the number of decomposition layers to divide the entire frequency band of the zero sequence waveform data into a plurality of non-overlapping sub-bands; in the offline stage, a power distribution network simulation model is constructed using electromagnetic transient simulation software, different fault mechanisms under different working conditions are simulated through the power distribution network simulation model, and electromagnetic transient simulation results are generated; based on the electromagnetic transient simulation results, transient energy analysis is performed on each sub-band, and the sub-bands are selected based on the transient energy concentration degree and the transient energy ratio; the wavelet packet energy value and the cross-correlation coefficient of each selected sub-band are calculated, and all selected sub-bands are arranged in frequency increasing order, and the finally generated transient feature vector consists of three parts: current energy sequence, voltage energy sequence, and voltage current cross-correlation coefficient sequence arranged in frequency increasing order.

[0009] Preferably, the generation process of the local diagnosis result comprises: generating a polarity result by component extraction and integral calculation of the zero sequence waveform data through a transient zero sequence power direction method; the specific process of the transient zero sequence power direction method is: extracting transient zero sequence components by digital filtering and differential processing of the zero sequence voltage and the zero sequence current of the zero sequence waveform data; the digital filtering and differential processing is realized through a digital differential filter; calculating a transient zero sequence instantaneous power integral value of the transient zero sequence components by an instantaneous power algorithm, and judging the polarity result; outputting a multi-frequency admittance feature vector with equal dimensions and a selected number of frequency bands by regression analysis of a transient feature vector through a support vector machine model; the support vector machine model adopts a radial basis function as a kernel function, and is trained based on different transient simulation data; each component of the multi-frequency admittance feature vector is a complex ratio of a transient current signal to a transient voltage signal in a corresponding frequency band; generating a directivity index by weighted summation of each component of the multi-frequency admittance feature vector; fusing the polarity result and the directivity index by logical AND to generate the local diagnosis result, wherein the local diagnosis result contains a direction indication and a result type; when the directivity index and the polarity result both indicate the same direction, the result type of the local diagnosis result is local isolation; only when one of the following conditions is met: the fault directions represented by the polarity result and the directivity index are inconsistent, the reliability of the polarity result is lower than a preset reliability threshold, and the reliability of the directivity index is lower than the preset reliability threshold, the result type of the local diagnosis result is cooperative positioning.

[0010] Preferably, the specific process of the exchange of the digitized transient data is: when the local diagnosis result is determined to be cooperative positioning, sending a request instruction of the exchange of the digitized transient data to a directly connected adjacent ring network box terminal through a communication network inside the power distribution network; the request instruction requires that the digitized transient data returned by the adjacent terminal is within the same time window of the fault occurrence time; the digitized transient data is losslessly encoded through differential encoding.

[0011] Preferably, the generation process of the comprehensive positioning result is as follows: after the ring network box receives the synchronous zero sequence current waveform data of the adjacent terminal, the Pearson correlation coefficient between the zero sequence current waveform of the terminal and the zero sequence current waveform of the adjacent terminal is calculated as a transient current similarity coefficient; at the same time, the transient waveform intensity difference coefficient between the waveforms of the terminal and the adjacent terminal is calculated, and the transient waveform intensity difference coefficient is the absolute value of the difference between the transient waveform intensity indexes of the waveforms of the terminal and the adjacent terminal, and the transient waveform intensity index is the integral of the square of the zero sequence current waveform; the calculated transient current similarity coefficient and transient waveform intensity difference coefficient are logically judged with the local diagnosis result; if the transient current similarity coefficient is greater than a preset similarity threshold, and the transient waveform intensity difference coefficient is less than a preset energy difference threshold, it is determined that there is no fault section boundary between the two terminals; if the transient current similarity coefficient is less than the preset similarity threshold, and the transient waveform intensity difference coefficient is greater than the preset energy difference threshold, it is determined that there is a first type of candidate fault section boundary between the two terminals; all the sections not belonging to the above logical judgment are determined as second type of candidate fault section boundary; the ring network box determines the fault section boundary by comparing the transient current similarity coefficients and the transient waveform intensity difference coefficients of all adjacent sections, and combining the direction indication of the local diagnosis result; the process of determining the fault section boundary is as follows: excluding the sections determined as not having fault section boundary; taking the first type of candidate fault section boundary and the second type of candidate fault section boundary as candidate boundaries, and selecting the candidate boundary consistent with the direction indication of the local diagnosis result as the fault section boundary; generating a fault section identifier based on the fault section boundary, and the fault section identifier is the comprehensive positioning result.

[0012] Preferably, the specific process of executing the fault isolation operation and the PHM process based on the remote control command is as follows: according to the comprehensive positioning result, the section located is mapped to the physical fault line by querying the distribution network topology and the device connection table of the ring network box; the ring network box generates a remote control command according to the built-in fault handling strategy model; under the premise that the fault current has been cleared, the generated remote control command is directly sent to the intelligent operating mechanism of the circuit breaker integrated in the ring network box through the integrated digital interface, and the PHM process is started after the operation is completed, wherein the PHM process is as follows: the ring network box records the cumulative operation number, the fault current level and the operation time stamp of the device, and generates fault operation data; a device health degradation model is constructed based on the cumulative operation number and the current stress, the mechanical life and electrical life consumption of the device are evaluated based on the fault operation data through the device health degradation model, and the device health status index and the remaining service life are updated.

[0013] A primary and secondary fusion complete ring network box fault diagnosis system, comprising: The acquisition and perception module is used for synchronously acquiring three-phase instantaneous voltage and instantaneous current signals of each feeder port in real time and converting into digitized transient data; The data processing and synchronization module performs time window preprocessing on the digitized transient data based on a fault occurrence time, extracts zero sequence waveform data, performs wavelet packet decomposition on the zero sequence waveform data, and generates a transient feature vector; The local diagnosis decision module is used for extracting components and performing integral calculation on the zero sequence waveform data through a transient zero sequence power direction method, generating a polarity result, performing regression analysis and weighted summation on the transient feature vector through a pre-trained support vector machine model, and outputting a directivity index, and generating a local diagnosis result based on the polarity result and the directivity index; The cooperative positioning and control execution module requests neighboring ring network cabinet terminals to exchange digitized transient data through a communication network when the local diagnosis result is to start cooperative positioning; calculates transient current similarity coefficients and transient waveform intensity difference coefficients based on the digitized transient data of the neighboring terminals, and generates a comprehensive positioning result in combination with the local diagnosis result; determines a fault line according to the comprehensive positioning result, and generates a remote control command; and performs fault isolation operation and PHM process based on the remote control command.

[0014] Compared with the prior art, the present application has the following advantages: 1. When the local diagnosis is uncertain, the present application calculates transient current similarity coefficients and transient waveform intensity difference coefficients based on waveform data, and combines the direction indication of the local diagnosis result, which can accurately determine the fault section boundary. It overcomes the limitation of open-loop topology, avoids the situation that the ring network cabinet draws incorrect diagnosis results in the bus fault, upstream fault or downstream fault of the outgoing line, forms a line selection blind area, and realizes accurate line selection and local autonomous isolation of the highest proportion of small current grounding faults.

[0015] 2. The present application proposes a local diagnosis combining the transient zero sequence power direction method and the multi-frequency admittance method, and introduces a cooperative positioning mechanism, which can uniquely and accurately determine the section boundary where the fault current waveform changes dramatically; it overcomes the limitation that the prior art only relies on the steady-state zero sequence current relationship and is easy to misjudge, greatly improves the diagnosis accuracy of the fault detection method for complex small current grounding faults.

[0016] 3. The present application performs multi-layer wavelet packet decomposition on the zero sequence current waveform extracted in the preset time window of the fault occurrence time, and generates a transient feature vector containing the selected frequency band of the fault characteristics; so that the fault diagnosis model can capture the weak and high frequency characteristics of the small current grounding fault, thereby greatly improving the fault detection sensitivity and anti-interference ability of the system in the small current grounding system. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 A flow chart of a primary and secondary fusion complete ring network box fault diagnosis method of the present application; Figure 2 A structural schematic diagram of a primary and secondary fusion complete ring network box fault diagnosis system of the present application; Figure 3 A flow chart of generating a local diagnosis result of the present application; Figure 4 A flow chart of generating a comprehensive positioning result of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0019] Please refer to Figures 1 to 4 The present application provides a primary and secondary fusion complete ring network box fault diagnosis method and system, and the technical solutions are as follows: Embodiment 1

[0020] Real-time synchronous acquisition of three-phase instantaneous voltage and instantaneous current signals of each feeder port, and conversion into digitized transient data; Time window preprocessing of the digitized transient data based on the time of fault occurrence, extraction of zero sequence waveform data; wavelet packet decomposition of the zero sequence waveform data, generation of transient feature vectors; Component extraction and integral calculation of the zero sequence waveform data through the transient zero sequence power direction method, generation of polarity results; regression analysis and weighted summation of the transient feature vectors through a pre-trained support vector machine model, output of the directivity index; generation of a local diagnosis result based on the polarity results and the directivity index; When the local diagnosis result is to start cooperative positioning, the ring network box requests exchange of digitized transient data through a communication network to adjacent ring network box terminals; calculation of transient current similarity coefficients and transient waveform intensity difference coefficients based on the digitized transient data of the adjacent terminals, and generation of a comprehensive positioning result in combination with the local diagnosis result; Determination of the fault line according to the comprehensive positioning result, and generation of a remote control command; execution of fault isolation operation based on the remote control command, and execution of a PHM process.

[0021] Further, the specific generation process of the digitized transient data is: real-time acquisition of three-phase instantaneous voltage and current signals through the primary equipment sensor integrated by the ring network box; in the acquisition process, the three-phase instantaneous voltage and current signals are synchronized by a clock source at the hardware level and are digitized by an analog-to-digital converter to generate digitized transient data; the digitized transient data are stored in RAM and are locked and extracted by a fault transient triggering mechanism; the fault transient triggering mechanism is based on detection of line current, and when the absolute rate of change of zero sequence current and / or zero sequence voltage of any feeder port exceeds a threshold value, the fault transient triggering mechanism is triggered.

[0022] Specifically, in the embodiment, the primary equipment sensor adopts an electronic transformer based on Rogowski coils and a voltage divider, the frequency response bandwidth of the electronic transformer needs to reach 20 kHz or more, and the measurement accuracy level should reach 0.2S level, to ensure accurate acquisition of high-frequency transient signals. The sampling rate of the ADC is 80 kHz, and the bit depth is preferably 16 bits or more, to ensure accurate capture and digitization of weak transient signals. The high-precision clock source adopts a GPS / Beidou timing PTP master clock, and the synchronization accuracy is less than or equal to 1 microsecond. The trigger logic of the fault transient triggering mechanism is set to OR logic, that is, it is triggered when any of the following conditions is met: the absolute rate of change of zero sequence voltage of any feeder port exceeds the minimum trigger level that distinguishes normal operating conditions from minimum fault transients; the absolute rate of change of zero sequence current exceeds the minimum trigger level that distinguishes normal operating conditions from minimum fault transients. The minimum trigger level must be higher than the absolute rate of change of the maximum power frequency zero sequence component that may occur in normal operating state and lower than the maximum absolute rate of change of the transient traveling wave generated by the minimum expected fault. The fault transient triggering mechanism uses an independent high-speed logic circuit to calculate in real time, and after the mechanism is triggered, the system locks and extracts digitized transient data containing a preset time length before and after the fault occurs, the preset time length completely contains the whole process waveform information before and after the fault occurs, and the range is relatively long, which is set to 5 ms before the fault and 50 ms after the fault.

[0023] By configuring integrated sensors and using hardware-level timestamp synchronization and high-precision analog-to-digital conversion, the present application can effectively avoid errors in transient waveform correlation calculation caused by terminal clock deviation, and ensure the accuracy of the overall diagnostic method.

[0024] Further, the specific generation manner of the transient feature vector comprises: locating a fault occurrence time, selecting a preset time window located before and after the fault occurrence time; in the preset time window, superimposition calculation is performed on three-phase instantaneous voltage and current, zero sequence waveform data containing instantaneous zero sequence voltage and zero sequence current is extracted; multi-layer wavelet packet decomposition is performed on the zero sequence waveform data, the multi-layer wavelet packet decomposition adopts a mother wavelet base, and a decomposition layer number is set, so that the entire frequency band of the zero sequence waveform data is divided into a plurality of non-overlapping sub-frequency bands; in an offline stage, an electromagnetic transient simulation software is used to construct a power distribution network simulation model, a fault mechanism under different working conditions is simulated through the power distribution network simulation model, and an electromagnetic transient simulation result is generated; according to the electromagnetic transient simulation result, transient energy analysis is performed on each sub-frequency band, and the sub-frequency bands are selected based on transient energy concentration degree and transient energy ratio; wavelet packet energy values and cross-correlation coefficients of each selected sub-frequency band are calculated, and all selected sub-frequency bands are arranged in frequency increasing order, and the finally generated transient feature vector is composed of three parts: current energy sequence, voltage energy sequence and voltage-current cross-correlation coefficient sequence arranged in frequency increasing order.

[0025] Specifically, in the embodiment, the preset time window is set to a duration that must be less than the round-trip propagation time of the fault traveling wave from the ring net box feeder port to the nearest electrical discontinuity point (for example, a busbar or a transformer). The selection of the duration strictly follows the first wave discrimination principle of the traveling wave, and ensures that the extracted transient feature vector only contains the propagation information of the first traveling wave; the starting time of the time window is set to a very short time interval after the fault transient triggering time, so as to eliminate the glitch interference caused by high-frequency digital differentiation at the triggering moment. The calculation of the zero sequence waveform data includes the calculation of instantaneous zero sequence current and the calculation of instantaneous zero sequence voltage , the calculation of the instantaneous zero sequence current is 1 / 3 of the arithmetic sum of three-phase instantaneous currents , , , the calculation of the instantaneous zero sequence voltage is 1 / 3 of the arithmetic sum of three-phase instantaneous voltages; The mother wavelet base adopted by the multi-layer wavelet packet decomposition is specifically db4, and the set decomposition layer number is specifically 4, so that the entire frequency band of the signal is divided into 16 non-overlapping sub-frequency bands; the selected frequency bands are specifically 5, and the selected frequency bands are the first 7 frequency bands with the highest transient energy concentration degree in the 16 sub-frequency bands. And the first 8 frequency bands with the highest transient energy ratio in the 16 sub-frequency bands. The finally selected 5 frequency bands are the intersection of the two groups of frequency bands satisfying the highest standard. The specific division of the 5 frequency bands is according to the 4-layer db4 wavelet packet decomposition result, the theoretical bandwidth of each sub-frequency band is 2.5 kHz, which is determined through energy entropy analysis on the zero sequence current waveform, and the specific frequency bands and frequency intervals (for example, the sampling rate is 80 kHz) are shown in Table 1: Table 1 Band frequency interval table

[0026] The dimension of the finally formed transient feature vector is consistent with the number of selected frequency bands, which is 5 dimensions.

[0027] The process of the electromagnetic transient simulation result specifically comprises: constructing a power distribution network simulation model by using an electromagnetic transient simulation software (PSCAD / EMTDC), simulating the fault mechanism of different working conditions and fault types (especially small current grounding faults) on the simulation model. Running the simulation and obtaining the voltage and current transient waveform data measured by the ring net box at the time of the fault. The transient data obtained by simulation is subjected to wavelet packet decomposition to obtain the wavelet packet energy of different sub-bands, so as to determine in which frequency bands the fault features are most prominent. The power distribution network simulation model constructed by the electromagnetic transient simulation software adopts a multi-phase, frequency-variable parameter line model to accurately simulate the propagation and attenuation characteristics of transient traveling waves at high frequencies. The grounding mode of the power distribution network system in the simulation model is a small current grounding system, and a constant impedance load model is adopted. The sampling rate of the simulation running should be consistent with the sampling rate of the actual collection system (preferably 80 kHz or higher).

[0028] The transient energy concentration degree is the proportion of the wavelet packet energy of the sub-band to the total energy, and the specific generation process of the transient energy ratio is: collecting data under normal operating conditions (no fault) and calculating the wavelet packet energy of the sub-band to generate normal working condition energy, collecting data under fault conditions and calculating the wavelet packet energy of the sub-band to generate fault working condition energy. The transient energy ratio is the ratio of the fault working condition energy to the normal working condition energy.

[0029] The specific process of calculating the wavelet packet energy value and the cross-correlation coefficient of each selected sub-band is: calculating the sum of squares of the zero sequence current decomposition coefficients in the frequency band to obtain the current frequency band energy; calculating the sum of squares of the zero sequence voltage decomposition coefficients in the frequency band to obtain the voltage frequency band energy; calculating the cross-correlation coefficient of the zero sequence voltage decomposition coefficients and the zero sequence current decomposition coefficients in the frequency band. The cross-correlation coefficient is obtained by multiplying the corresponding voltage coefficient and current coefficient, summing, and then dividing by the square root of the product of the voltage energy and the current energy. The coefficient reflects the similarity and polarity relationship of the voltage waveform and the current waveform in the time domain, thereby implicitly including the phase characteristics of the fault.

[0030] The finally generated transient feature vector is composed of three parts, namely the current frequency band energy sequence, the voltage frequency band energy sequence and the cross-correlation coefficient sequence arranged in ascending order of frequency. The combined feature vector can represent the energy size and the relative phase relationship of the voltage and current of the fault at the same time.

[0031] The application realizes the joint analysis of transient signals in time domain and frequency domain by multi-layer wavelet packet decomposition of zero sequence waveform data, effectively separates fault characteristics from background noise and power frequency interference, and avoids blind selection, thereby significantly improving the discriminability and sensitivity of the feature vector.

[0032] Further, the generation process of the local diagnosis result includes: generating a polarity result by component extraction and integral calculation of the zero sequence waveform data through the transient zero sequence power direction method; the specific process of the transient zero sequence power direction method is: extracting transient zero sequence components by performing digital filtering and differential processing on the zero sequence voltage and the zero sequence current of the zero sequence waveform data; the digital filtering and differential processing is realized by a digital differential filter; calculating transient zero sequence instantaneous power integral values of the transient zero sequence components by using an instantaneous power algorithm, and judging the polarity result; outputting a multi-frequency admittance feature vector with the same number of dimensions as the selected frequency bands by performing regression analysis on the transient feature vector through a support vector machine model; the support vector machine model uses a radial basis function as a kernel function, and the support vector machine model is trained based on different transient simulation data; each component of the multi-frequency admittance feature vector is a complex ratio of the transient current signal to the transient voltage signal in the corresponding frequency band; generating a directivity index by weighted summing each component of the multi-frequency admittance feature vector; generating the local diagnosis result by logically and fusing the polarity result and the directivity index, wherein the local diagnosis result includes a direction indication and a result type; when the directivity index and the polarity result both indicate the same direction, the result type of the local diagnosis result is local isolation; the result type of the local diagnosis result is to start cooperative positioning only when one of the following conditions is met: the fault directions represented by the polarity result and the directivity index are inconsistent, the reliability of the polarity result is lower than a preset reliability threshold, and the reliability of the directivity index is lower than the preset reliability threshold.

[0033] Specifically, the digital differential filter adopts an FIR (Finite Impulse Response) structure, and the filter order is N order (N is preferably 10) to realize approximate linear phase shift and minimum delay, and to ensure that the extraction of the transient zero sequence component does not interfere with the waveform polarity. The difference equation of the filter is: ; wherein is the input zero sequence waveform sample value, is the filtered transient zero sequence component, The coefficients of the differential filter are optimized by least square method, n represents the index of the discrete time series, indicating the data of the current sampling point, k represents the index of the order of the filter, indicating the delay of the input sequence. The coefficient design of the filter should ensure that the group delay is less than 50 microseconds, while the power frequency and its low harmonic components are suppressed to the maximum extent, and the high frequency band of the transient traveling wave energy is accurately linearly shifted and differentiated. The order of the differentiator must be designed through high-precision coefficient fitting (such as Lagrange interpolation method) to ensure that the amplitude error in the target transient frequency range (such as 3kHz to 15kHz) is less than 0.5%, so as to amplify the useful transient signal while effectively suppressing the negative impact of power frequency components and acquisition noise on direction judgment.

[0034] The instantaneous power algorithm is used to calculate the transient zero sequence instantaneous power The value of the transient zero sequence instantaneous power is equal to the product of the transient zero sequence voltage and the transient zero sequence current obtained after digital filtering and differentiation. The calculation of the transient zero sequence instantaneous power integral value is based on the calculation of the above-mentioned instantaneous power , The calculation of the transient zero sequence instantaneous power integral value is the cumulative summation (integration) of the instantaneous power in a specific time interval. The integration interval starts from the fault occurrence time and lasts for the fault occurrence time plus the integration duration . Among them, the integration duration should be determined according to the shortest section length of the adjacent line of the ring box in the distribution network; the integration duration must be strictly shorter than the round-trip propagation time of the fault traveling wave from the measurement point to the nearest reflection point (the opposite side of the shortest adjacent line section), so as to ensure that only the first wave information is included in the integration time, effectively excluding the interference of reflected waves on power direction judgment.

[0035] The specific process for determining the polarity result is as follows: Before determining the polarity, the system first performs a validity check: calculates the maximum instantaneous value of the transient zero-sequence current within a preset time window at the time of the fault occurrence; only when the maximum instantaneous value exceeds the fault current amplitude threshold is the subsequent polarity determination performed. The fault current amplitude threshold should be at least 10% higher than the maximum instantaneous value of the unbalanced current and background noise of the normal system; otherwise, if the validity check fails, no polarity result is generated. Since the local diagnostic result lacks direction indication information at this time, the reliability of the polarity result is determined to be lower than the preset reliability threshold, and the result type of the local diagnostic result is determined to require the activation of collaborative positioning, and the direction indication is empty or remains in the previous state. For data that passes the verification, the direction from the ring network box bus side to the feeder side is set as the positive reference direction. If the calculated polarity result is negative, it means that the power flows to the bus, and the fault is determined to be on the ring network box bus or its upstream power supply side; if the polarity result is positive, it means that the power flows to the line, and the fault is determined to be on the feeder side.

[0036] The support vector machine model constructs two independent regression sub-models for each selected frequency band: one for predicting the magnitude of the admittance in that band, and the other for predicting the phase angle of the admittance in that band. Although the input transient feature vector only contains energy coefficients, due to the nonlinear inherent mapping relationship between the energy distribution of the transient traveling wave and the distance to the fault point and the path impedance (complex), the model can implicitly learn this mapping through training with a large amount of full-phase simulation data, thereby regressing the magnitude and phase angle respectively.

[0037] The support vector machine model uses a radial basis function (RBF) as its kernel function. The RBF kernel function has the following form: ; in, and It is the input transient feature vector. The kernel function parameter is a crucial parameter that controls the distribution range of data in high-dimensional space. Model training is based on a large amount of transient simulation data with different fault types and locations. The training data must cover: different fault types (e.g., single-phase grounding, two-phase grounding, phase-to-phase short circuit), different fault distances (from the ring network busbar to the end of the line), and different transient operating conditions (e.g., different loads, different grounding resistances). Label generation: the output label (multi-frequency admittance feature vector) corresponding to each set of simulation data (input transient feature vector) is pre-calculated based on the actual physical parameters of the simulation model and the fault location. During model training, cross-validation methods must be used to adjust the regularization parameter C and the kernel function parameter. Two key hyperparameters were optimized.

[0038] The components of the multi-frequency admittance feature vector are generated by calculating the complex ratio of the transient current signal to the transient voltage signal in each selected frequency band. Specifically, the complex ratio is obtained by performing short-time Fourier transform (STFT) on the zero-sequence waveform data, extracting the phasors of the current and voltage in the corresponding frequency band, and calculating the phasor ratio. In this way, a multi-frequency admittance feature vector with dimensions equal to the number of selected frequency bands is generated. To ensure the integrity of the electrical physical information, each component of the multi-frequency transient admittance feature vector must contain both the modulus and the phase angle information of the admittance. Specifically, each component of the feature vector is composed of its modulus and phase angle. The modulus is the ratio of the envelope amplitude of the transient current signal to the envelope amplitude of the transient voltage signal in the corresponding frequency band; the phase angle is the phase difference between the transient current signal and the transient voltage signal in the corresponding frequency band.

[0039] For the training of the support vector machine model, the output label (i.e. the real multi-frequency admittance feature vector) is generated by calculating the inverse of the theoretical equivalent impedance from the fault point to the measurement point at the center frequency of the frequency band. The modulus and phase angle of the complex admittance are taken as the two components of the standard output label of the training sample. Since the admittance feature vector is a vector with 2K real components (where K is the number of frequency bands), the support vector machine model is configured as a multi-output regression model, which aims to simultaneously regress the modulus and phase angle of each frequency band. In this way, high-precision directional regression analysis of fault distance and direction is achieved.

[0040] The directional index is generated by weighted sum of the components of the multi-frequency admittance feature vector, which consists of five components, each representing the comprehensive feature of the equivalent transient admittance in a selected frequency band. The five admittance feature components are multiplied by their respective weight coefficients , respectively, and then all the product results are accumulated to obtain the final directional index, the weight coefficients are set for each frequency band, and their values are determined based on electromagnetic transient simulation results and fault sensitivity analysis, and the sum of all weight coefficients should be equal to 1.

[0041] The weight coefficients The establishment process is specifically: based on electromagnetic transient simulation data, the Pearson correlation coefficient between the transient energy change amount and the fault distance change amount of each selected frequency band is calculated; the absolute value of the Pearson correlation coefficient is taken as the sensitivity factor of the frequency band; the weight is normalized, and the specific normalization process is: the sensitivity factor values of all selected selected frequency bands are counted; the arithmetic sum of the sensitivity factor values is calculated; the sensitivity factor value of each frequency band is divided by the arithmetic sum, and the quotient is the weight coefficient corresponding to the frequency band. The calculation logic ensures that the sum of all frequency band weight coefficients is strictly equal to 1, so that the frequency band with higher sensitivity occupies a larger proportion in the final diagnosis result.

[0042] The calculation of the polarity result reliability is based on the integral value of the transient zero sequence instantaneous power The reliability is defined as: comparing the absolute value of the integral value of the instantaneous power with the equivalent noise energy level estimated by the system background noise, signal acquisition error and other factors to obtain a ratio. The larger the ratio, the stronger the energy of the transient traveling wave signal, the better the ability to drown out noise, and the higher the reliability of direction judgment. The directivity index reliability is obtained according to the prediction residual statistical value or cross-validation error accumulated by the model during the training process. Specifically, the reliability is inversely proportional to the average prediction error of the model under different fault scenarios. The higher the reliability value, the closer the output prediction of the model to the current transient feature vector to the true physical label of the training data, and the higher the reliability of the directivity judgment. The reliability threshold values of the polarity result and the directivity index are empirical parameters determined by analyzing and counting a large amount of power distribution network electromagnetic transient simulation and field test data. The setting principle of these threshold values is: when the reliability of the polarity result is higher than the set threshold value, and the reliability of the directivity index is higher than the set threshold value, the correctness of the local diagnosis result can meet the established engineering requirements.

[0043] As an enhancement scheme of the support vector machine model, an online adaptive correction mechanism is introduced to fine-tune the key parameters of the SVM model using real-time collected confirmed non-fault event data (such as normal switching operation, lightning arrester action). The specific implementation is: When the system detects a high-energy transient event that is not a fault (for example, the zero sequence current rate of change exceeds the threshold value but there is no subsequent isolation operation or the transient polarity result is determined to be a non-fault direction), the transient feature vector of the event is taken as the input, and the theoretical line impedance admittance value under normal operating state calculated in real time according to the power grid topology is taken as the standard output label to construct a correction training set The goal of the correction mechanism is to fine-tune the weight coefficients and the kernel function parameters of the SVM model to minimize the correction training set The sum of squares of the previous regression errors. An incremental learning algorithm, such as incremental support vector regression (ISVR) or a small-batch online learning method based on gradient descent, is adopted to avoid time-consuming global model retraining at each correction. The algorithm only slightly adjusts the current support vector set and kernel function parameters based on the data The correction process is completed within 50 milliseconds, and does not cause significant delays to the main diagnostic process. Correction frequency limit: To ensure model stability, a minimum time interval for the correction mechanism is set (e.g., a maximum of one trigger every 24 hours), preventing frequent model drift due to abnormal data collection.

[0044] By introducing an online adaptive correction mechanism, the main problem of applying static models in dynamic power grid environments is solved, significantly improving the long-term reliability and continuous accuracy of local diagnosis.

[0045] The invention combines two complementary transient analysis techniques for local diagnosis, greatly reducing the risk of misjudgment by a single method, ensuring high reliability when the ring main unit performs autonomous isolation operations.

[0046] Further, the specific process of exchanging digitalized transient data is as follows: when the local diagnosis result is determined to start cooperative positioning, a request instruction for exchanging digitalized transient data is sent to the adjacent ring main unit terminal directly connected through the communication network within the distribution network; the request instruction requires the digitalized transient data returned by the adjacent terminal to be within the same time window of the fault occurrence time; the digitalized transient data is losslessly encoded through differential encoding.

[0047] Specifically, the communication network is based on IEC 61850 GOOSE messages; the digitalized transient data is losslessly encoded through differential encoding to minimize data transmission volume and reduce cooperative positioning communication delay. After receiving the instruction, the adjacent ring main unit terminal returns the differential encoded waveform record file (such as based on the COMTRADE format) to the current ring main unit through the IEC 61850 MMS service.

[0048] The invention exchanges data when cooperative positioning is started, ensuring efficient startup and accurate calculation of the cooperative positioning mechanism. On-demand request (triggered only when local diagnosis is "need to start cooperative positioning") avoids a large amount of unnecessary communication load, optimizing the utilization rate of communication resources in the distribution network.

[0049] Further, the generation process of the comprehensive positioning result is: after the ring network box receives the synchronous zero sequence current waveform data of the adjacent terminal, the Pearson correlation coefficient between the zero sequence current waveform of the terminal and the zero sequence current waveform of the adjacent terminal is calculated as a transient current similarity coefficient; at the same time, the transient waveform intensity difference coefficient between the waveforms of the terminal and the adjacent terminal is calculated, the transient waveform intensity difference coefficient is the absolute value of the difference between the transient waveform intensity indexes of the waveforms of the terminal and the adjacent terminal, and the transient waveform intensity index is the integral of the square of the zero sequence current waveform; the calculated transient current similarity coefficient and transient waveform intensity difference coefficient are logically judged with the local diagnosis result; if the transient current similarity coefficient is greater than a preset similarity threshold, and the transient waveform intensity difference coefficient is less than a preset energy difference threshold, it is determined that there is no fault section boundary between the two terminals; if the transient current similarity coefficient is less than the preset similarity threshold, and the transient waveform intensity difference coefficient is greater than the preset energy difference threshold, it is determined that there is a first type of candidate fault section boundary between the two terminals; all the sections not belonging to the above logical judgment are determined as the second type of candidate fault section boundary; the ring network box determines the fault section boundary by comparing the transient current similarity coefficients and the transient waveform intensity difference coefficients of all adjacent sections, and combining the direction indication of the local diagnosis result; the determination process of the fault section boundary is: excluding the sections determined as not having the fault section boundary; taking the first type of candidate fault section boundary and the second type of candidate fault section boundary as candidate boundaries, and selecting the candidate boundary consistent with the direction indication of the local diagnosis result as the fault section boundary; generating a fault section identifier based on the fault section boundary, and the fault section identifier is the comprehensive positioning result.

[0050] Specifically, the calculation process of the Pearson correlation coefficient is: the average values of all sampling points in the zero sequence transient current waveform sequence collected by the ring network box and the adjacent ring network box are calculated respectively. For each sampling point in the sequence, the original value of the sampling point is subtracted from the average value of the sequence. The corresponding sampling point values after the centering processing are multiplied. Then, all the product results are accumulated and summed. The result of the accumulation and summation is divided by a standardization factor reflecting the fluctuation amplitude of the two waveforms, and the standardization factor is obtained by multiplying the fluctuation square sums of the two waveform sequences and taking the square root of the product result. The specific process of the fluctuation square sum is: the square of the difference between all sampling points and their average value is calculated, and all the square values are accumulated and summed. The finally calculated value is the transient current similarity coefficient.

[0051] The calculation of the transient waveform intensity index is the integral of the square of the zero sequence current waveform, and the integral time range should be the integral duration of the transient zero sequence instantaneous power integral value in the local diagnosis result. Consistency is ensured to capture only the energy of the initial fault traveling wave.

[0052] In the aforementioned logical judgment, the preset similarity threshold is set to 0.95 to ensure a high correlation between the waveforms; the preset energy difference threshold is set to 0.05 pu to ensure minimal transient energy difference between the two terminals. These thresholds are determined through statistical analysis and optimization of the electromagnetic transient simulation results of the distribution network simulation model. The specific process of the statistical analysis and optimization is as follows: Simulate two sample sets in the simulation model—intra-area faults and extra-area faults—and calculate the transient current similarity coefficient for each sample; plot the probability density curves of the coefficient distributions of the two types of samples, and select the value corresponding to the intersection of the two curves as the initial threshold; further, through receiver operating characteristic curve (ROC curve) analysis, select the value corresponding to the point that minimizes the sum of the false alarm rate and the missed alarm rate as the final preset similarity threshold.

[0053] The specific logic for determining the fault segment boundary is as follows: Segments deemed to lack a fault segment boundary are excluded; the first and second types of candidate fault segment boundaries are selected as candidate boundaries. Prioritize selecting the boundary from the first type of candidate fault segment boundaries that matches the direction indication of the local diagnostic results as the fault segment boundary; if there are no matching items in the first type of candidate boundaries or the first type of candidate boundaries is empty, then select the candidate boundary from the second type of candidate fault segment boundaries that matches the direction indication of the local diagnostic results as the fault segment boundary. If multiple candidate boundaries matching the direction indication exist, further compare the transient current similarity coefficients corresponding to these candidate boundaries, and select the segment with the smallest similarity coefficient as the fault segment boundary; if there is more than one segment with the smallest similarity coefficient, compare the corresponding transient waveform intensity difference coefficients, and select the segment with the largest energy difference coefficient as the unique fault segment boundary. This logic is based on the principle of lowest waveform correlation and largest energy difference within the fault segment.

[0054] As an enhancement to the comprehensive positioning results, a third cooperative positioning factor is introduced: the high-frequency energy ratio difference coefficient (FCRD). This coefficient is used to quantify the essential difference between fault points and non-fault points in high-frequency characteristic attenuation. The specific implementation method is as follows: After performing wavelet packet decomposition to generate transient eigenvectors, the high-frequency band energy is calculated using the decomposition results. The high-frequency band energy Defined as the sum of squares of all wavelet packet coefficients in the transient eigenvector with frequencies ranging from 5 kHz to 12.5 kHz. Calculate the normalized high-frequency energy ratio. High-frequency energy is about to be generated. Divide by the sum of the squares of all wavelet packet coefficients in the transient eigenvector. Calculate the high-frequency energy ratio difference coefficient between this terminal and its neighboring terminals. . defined as the absolute difference of values. The logical judgment of the integrated positioning result, the condition of determining the fault-free boundary in the section is enhanced as: the transient current similarity coefficient is greater than 0.95, and the transient waveform intensity difference coefficient is less than 0.05 p.u., and the high-frequency energy ratio difference coefficient is less than the preset threshold value of 0.02. The condition of determining the first type of candidate fault section boundary is enhanced as: the transient current similarity coefficient is less than 0.95, and the transient waveform intensity difference coefficient is greater than 0.05 p.u., and the high-frequency energy ratio difference coefficient is greater than the preset threshold value of 0.10. When multiple second type of candidate fault section boundaries need to be finally distinguished, the largest section is preferentially selected as the final fault section boundary.

[0055] By introducing the high-frequency energy ratio difference coefficient (FCRD), a third orthogonal criterion is added on the basis of the existing waveform similarity and transient waveform intensity difference, thereby significantly improving the robustness and accuracy of the cooperative positioning.

[0056] By calculating the Pearson correlation coefficient (similarity) and the transient waveform intensity difference coefficient of the transient zero sequence current waveforms of adjacent terminals, quantitative and cross-terminal verification of the fault section boundary is realized. The uncertain local diagnosis result is converted into a unique and reliable integrated positioning result, which provides accurate coordinates for the final fault isolation.

[0057] Further, the specific process of executing the fault isolation operation and the PHM process based on the remote control command is as follows: according to the integrated positioning result, the section located is mapped to the physical fault line by querying the distribution network topology and the equipment connection table of the ring network box; the ring network box generates a remote control command according to the built-in fault handling strategy model; under the premise that the fault current has been cleared, the generated remote control command is directly sent to the intelligent operating mechanism of the circuit breaker integrated in the ring network box through the integrated digital interface, and the PHM process is started after the operation is completed. The PHM process is as follows: the ring network box records the cumulative operation number, fault current level and operation time stamp of the equipment, and generates fault operation data; a device health degradation model is constructed based on the cumulative operation number and current stress, and the mechanical life and electrical life consumption of the equipment are evaluated based on the fault operation data through the device health degradation model, and the device health status index and remaining service life are updated.

[0058] The distribution network topology and the equipment connection table are stored in the form of an adjacency list, which records the logical section identifier of each feeder port of the ring network box, the connected physical line number, and the associated circuit breaker device ID and communication address, so as to realize accurate mapping from the logical positioning result to the physical equipment operation object.​​​​

[0059] The fault processing strategy model is an expert system in the form of a rule base, and its input includes a fault section identifier, a fault type (for example, a small current ground fault), and a current device operating state. The core decision rule of the model is: based on the upstream and downstream physical connection relationship of the fault section and the line protection coordination principle, a trip or opening remote control command for the upstream circuit breaker and the downstream load switch of the fault section is generated, if the integrated positioning result indicates that the fault is located at the feeder side of the ring net box, a opening command is generated to isolate the feeder; if the integrated positioning result indicates that the fault section is located at the bus side of the ring net box or the upstream power supply side, a closing command or the existing state is maintained, and diagnostic information is reported to the master station.

[0060] Before generating the remote control command, the system first performs a safe opening capability check, and the specific process is as follows: querying the distribution network device account, identifying whether the connected switch device type on the current fault line is 'circuit breaker' or 'load switch', and reading the rated short-circuit breaking current value of the device. The real-time monitored fault current amplitude is obtained, which is compared with the rated short-circuit breaking current of the device. If it is identified that the device is a circuit breaker, and the real-time fault current amplitude is less than its rated short-circuit breaking current, the system determines that it has safe opening capability, generates a remote control command to immediately open, and directly cuts off the fault line. If it is identified that the device is a load switch, or although it is a circuit breaker but the real-time fault current exceeds its rated breaking capacity, the system determines that it does not have direct opening capability. At this time, the system automatically locks the immediate opening function, and instead executes the no-voltage opening logic. That is: the system waits for the action of the upper-level substation protection device to trip, and when it is detected that the line voltage disappears and the current returns to zero, a opening command is generated to open the switch, and the isolation operation is completed before the upper-level power supply recloses.

[0061] The integrated digital interface is a special hardware circuit board for realizing digital communication between a distribution station terminal (DTU) and a primary equipment bay, and includes: a fiber transceiver (SFP / SFP+), a communication rate of 100 / 1000 Mbps, and no less than 2 optical ports; a master control chip (FPGA or high-performance DSP) that meets the real-time analysis and generation of SV messages (sample values) and GOOSE messages (control commands). A PTP slave clock chip with a synchronization accuracy of better than 1 microsecond, a buffer memory (RAM) with a cache capacity meeting the high-speed message cache requirement of at least 100 milliseconds, a power isolation and filtering circuit with an isolation voltage meeting the anti-interference level required by the IEC standard.

[0062] The circuit breaker intelligent operating mechanism is an integrated device integrating mechanical execution, control and monitoring functions. It contains: an operating executor, which is a permanent magnet operating mechanism or a high-energy electric spring mechanism, and the total opening and closing time is less than 30 milliseconds; an embedded microcontroller (single-chip microcomputer / MCU) adopting a dual-core heterogeneous architecture of "MCU+DSP" or "MCU+FPGA". Among them, the MCU has a main frequency of more than 300 MHz and is responsible for protocol stack and logic control; the DSP / FPGA acts as a coprocessor and is specially used for performing wavelet packet decomposition and parallel floating-point operation acceleration of SVM kernel functions to ensure that the processing time delay of the whole process of fault diagnosis is strictly controlled within 20 ms; a contact stroke displacement sensor with a detection accuracy of better than 0.1 millimeter; an electric parameter acquisition unit with a sampling frequency meeting the demand of fault current waveform monitoring; a non-volatile storage chip (such as Flash or FRAM) with a storage capacity meeting the demand of operation data storage for more than ten years; and a driving power switch circuit with rated current / voltage meeting the demand of instantaneous high-power driving of the operating mechanism coil or motor. In the embodiment, for the ring network box configured with a vacuum circuit breaker, the operating mechanism is configured to perform short-circuit current breaking operation; and for the ring network box configured with a sulfur hexafluoride load switch, the operating mechanism is configured to perform only opening operation when it is detected that the line has no current and no voltage, so as to cooperate with the above-mentioned no-voltage opening logic and ensure the safety of device operation.

[0063] The device health degradation model adopts a contact wear calculation method based on the weighted accumulation of breaking current. The specific calculation logic is: for each tripping operation of the circuit breaker, record its breaking current value, and perform exponential operation on the current value, and the value of the index is usually set between 1.5 and 2.0 (for example, for sulfur hexafluoride load switch, the value is usually 1.9 to 2.0, and for vacuum circuit breaker, the value is usually 1.5 to 1.7); multiply the calculated value by a preset wear coefficient to obtain the equivalent wear amount of a single operation, and the wear coefficient is determined according to the ablation resistance experimental data of the contact alloy material, and its typical value range is 0.0001 to 0.0005, so as to quantify the irreversible physical damage caused by single large current breaking to the device. The system quantifies the electrical life consumption rate of the device by calculating the proportion of the cumulative equivalent wear amount to the maximum cumulative wear threshold, and further calculates the percentage of the remaining service life of the device. The maximum cumulative wear threshold allowed by the device is the total equivalent wear reference value set by the manufacturer according to the contact material, design thickness and durability test results. For example, for a selected model of sulfur hexafluoride load switch, the maximum cumulative wear threshold is preferably set to 10,000 equivalent wear units. The device health status indicator includes mechanical life consumption rate and electrical life consumption rate. The mechanical life consumption rate is the ratio of the cumulative number of operations of the device to the designed mechanical life number of the device. The electrical life consumption rate is based on the fault current level and the cumulative number of operations to evaluate the arc damage to the contacts and arc chamber when breaking the fault current. The final device health status indicator and the determination of the remaining service life of the device should be based on the principle of conservative minimum value. That is, after calculating the mechanical life consumption rate and the electrical life consumption rate respectively, first calculate the respective remaining life percentage (i.e. subtract the respective consumption rate from 1). Then, the final percentage of the remaining service life of the device takes the lower value of the mechanical remaining life percentage and the electrical remaining life percentage.

[0064] As an enhanced solution of the PHM process, the PHM results are integrated with the device stress model, spare parts inventory information and maintenance work order system to automatically generate optimized maintenance work orders with time windows, which are specifically implemented as: Adding the calculation of the operating stress factor in the PHM process, which is proportional to the square of the breaking current when the circuit breaker is operating, that is: ; wherein is the fault current, is the operation time, is a material property constant. This factor is used to more accurately quantify the damage to the contacts and arc chamber of the device caused by each operation. Two maintenance trigger thresholds are set: the life threshold Triggered when the remaining useful life (RUL) of the circuit breaker is less than 60%. High stress accumulation threshold. : The cumulative operating stress factor of the equipment over three consecutive months The system triggers when the stress exceeds a preset 1.5 times the baseline stress. If either of the two aforementioned maintenance trigger thresholds is met, an electronic maintenance work order will be automatically created, detailing the equipment ID, the reason for triggering (insufficient RUL or excessive stress), and recommended maintenance items (e.g., contact inspection / replacement). The system then checks the spare parts inventory management system to confirm the availability of required spare parts (e.g., arc-extinguishing chamber, operating mechanism springs); simultaneously, it checks the maintenance personnel scheduling system to match a suitable maintenance time window (e.g., planned power outage window). Based on the remaining percentage of RUL (lower RUL, higher priority) and the severity of the stress exceedance, the work order is assigned a high, medium, or low maintenance priority and pushed to the maintenance personnel's mobile terminal.

[0065] By transforming passive PHM assessment results into proactive maintenance actions and unplanned power outages into planned maintenance, a leap from condition monitoring to predictive maintenance has been achieved, resulting in significant operational benefits.

[0066] This invention ensures rapid and accurate fault isolation through isolation operations and a PHM (Prognostics and Health Management) process. It achieves a closed-loop fault handling system and intelligent operation and maintenance, effectively extending equipment lifespan and reducing maintenance costs.

[0067] This invention deeply integrates dual local transient diagnostics (power direction and multi-frequency admittance) with cross-terminal collaborative positioning, ensuring high-precision determination of fault segment boundaries. It overcomes the blind spots and uncertainties of traditional steady-state fault location methods in low-current grounding systems and open-loop topologies. Example 2

[0068] This application provides a primary and secondary integrated ring network box fault diagnosis system, referring to... Figure 2 This is a schematic diagram of a primary and secondary integrated ring network enclosure fault diagnosis system. The primary and secondary integrated ring network enclosure fault diagnosis system includes: an acquisition and sensing module, a data processing and synchronization module, a local diagnosis and decision module, and a collaborative positioning and control execution module.

[0069] Further, the acquisition and sensing module contains a high-frequency electronic transformer unit, a synchronous digitizing unit and a fault transient triggering and storage unit. The high-frequency electronic transformer unit is integrated based on Rogowski coil current transformer and voltage transformer technology based on voltage divider technology, which is used to real-time induction of three-phase transient voltage and current signals of each feeder port. The synchronous digitizing unit contains a high-precision GPS / Beidou timing PTP master clock source, a 16-bit high-precision analog-to-digital converter, which uses the clock source to stamp the collected signals with a hardware-level timestamp, converts the analog signals into digitized transient data, and ensures the timing accuracy of the stored waveforms. The fault transient triggering and storage unit is composed of a high-speed logic comparison circuit and a circular RAM memory, which is used to monitor the line state in real time and extract complete data from 5ms before fault to 50ms after fault from the RAM.

[0070] Further, the data processing and synchronization module contains a zero sequence waveform extraction unit, a multi-layer wavelet packet decomposition unit and a feature vector optimization selection unit. The zero sequence waveform extraction unit is composed of a digital signal processing chip, which is used to locate the fault occurrence time, intercept the preset time window data from 1ms to 5ms after the fault, and extract the instantaneous zero sequence voltage and current waveforms through superposition calculation. The multi-layer wavelet packet decomposition unit is an operation unit based on wavelet transform algorithm, which uses Daubechies4 (db4) mother wavelet basis to perform 4-layer wavelet packet decomposition on the zero sequence data, and divides the frequency band into 16 sub-bands. The feature vector optimization selection unit is composed of a distribution network simulation model interface, which is used to calculate the transient energy concentration and energy ratio of each sub-band in combination with the electromagnetic transient simulation results, automatically select 5 selected frequency bands reflecting the fault characteristics, and generate a 5-dimensional transient feature vector in ascending order of frequency.

[0071] Further, the local diagnostic decision module contains a transient polarity determination unit, a multi-frequency admittance regression analysis unit, and a decision logic fusion unit. The transient polarity determination unit is composed of a low-delay digital differential filter and an instantaneous power integrator, which is used to calculate and integrate the transient zero sequence instantaneous power. If the integration result is negative, it is determined to be a bus side fault; if it is positive, it is determined to be a feeder side fault. The multi-frequency admittance regression analysis unit is composed of a pre-trained support vector machine model, which uses a radial basis function as a kernel function, inputs the transient feature vector, and outputs a multi-frequency admittance feature vector. By weighted sum of each component, a directivity index is generated. The decision logic fusion unit is composed of a logic comparator, which is used to perform logical and fusion of the polarity result and the directivity index.

[0072] Further, the cooperative positioning and control execution module includes a differential encoding communication unit, a comprehensive positioning calculation unit, a topology mapping and control unit, and a PHM intelligent operation and maintenance unit. The differential encoding communication unit is composed of an IEC61850 GOOSE-based communication interface. When a cooperative request is received, the transient data is differentially losslessly encoded and sent to the adjacent DTU. At the same time, the data returned by the adjacent terminal is received. The comprehensive positioning calculation unit is composed of a correlation analysis and difference coefficient calculation engine. It is used to calculate the transient current similarity coefficient and transient waveform intensity difference coefficient of the terminal and the adjacent terminal waveforms, and logically determine whether there is a fault section boundary between the two terminals. The topology mapping and control unit is composed of a topology storage and integrated digital interface, which is used to query the distribution network topology and map the determined fault section to the physical line. According to the fault handling strategy model, a remote control command is generated to directly drive the intelligent operating mechanism of the circuit breaker for isolation. The PHM intelligent operation and maintenance unit is composed of a life assessment model, which is used to update the remaining useful life (RUL) of the equipment, automatically trigger maintenance work orders and match spare parts inventory. Embodiment 3

[0073] The underground cable in the embodiment of the present application is damaged due to aging insulation, and a high resistance grounding fault occurs. The generation process of digital transient data is focused on, and the specific implementation of front-end sensing and triggering is described in detail to ensure the accuracy of source data.

[0074] An integrated primary equipment sensor is configured at each feeder port of the ring network box D. The sensor uses a current transformer based on Rogowski Coil and a voltage transformer based on a voltage divider technique. The frequency response bandwidth of the sensor is set to be above 20 kHz, and the measurement accuracy level reaches 0.2S level, so as to ensure that the high-frequency transient signals at the fault instant are captured without distortion. In the signal acquisition process, a high-precision clock source (GPS / Beidou timing PTP master clock) is introduced. The clock source stamps hardware-level time stamps on the three-phase instantaneous voltage and current signals, and the synchronization accuracy is controlled within 1 microsecond, so as to eliminate the time deviation between multiple terminals. The analog signals are digitized by an analog-to-digital converter (ADC). The sampling rate of the ADC is set to be 80 kHz, and the bit depth is 16 bits. The generated digital transient data is stored in a loop RAM (random access memory) in real time for storage, and waits for a trigger instruction. The system runs an independent fault transient triggering mechanism, which monitors the line state in real time based on an independent high-speed logic circuit. The trigger logic is set to be or logic, that is, any one of the following conditions is met to trigger: the absolute rate of change of zero sequence voltage of any feeder port is detected to be more than 5000 kV / s; the absolute rate of change of zero sequence current is detected to be more than 500 kA / s. Once triggered, the system immediately locks and extracts the data in the RAM, and the data segment range covers the complete waveform from 5 ms before the fault occurs to 50 ms after the fault.

[0075] Through the high-precision hardware-level synchronous acquisition of transient data, the time of fault occurrence and the complete waveform data before and after the fault occurrence can be accurately captured, thereby providing high-quality original input for subsequent analysis. Embodiment 4

[0076] The embodiment of the present application refers to the specific scene of Embodiment 3, and focuses on the local diagnosis method of transient power direction and multi-frequency admittance fusion, which is described in detail in Figure 3 , and a flowchart for generating a comprehensive positioning result.

[0077] The data processing and synchronization module of the ring box D has extracted accurate zero sequence voltage and current waveforms, and generated a 5-dimensional transient feature vector based on the selected high-frequency band. The transient polarity determination unit integrates the zero sequence power in a very short time window (such as 13.3 microseconds) after the fault occurs. Since the fault occurs at the downstream feeder side of D, the power integration result is positive (the power flows from the bus to the feeder), which preliminarily determines that it is a feeder side fault. The multi-frequency admittance regression analysis unit inputs the 5-dimensional feature vector into the SVM model. The SVM model combines the energy weights of each frequency band, and the output directivity index is significantly greater than zero, strongly pointing to the feeder side. The decision logic fusion unit performs logical AND judgment on the directivity index and the polarity result. Both are highly consistent and have high reliability. The system generates a "locally isolable" local diagnosis result in a very short time (for example, 20 ms), and immediately starts the tripping process.

[0078] Through the rapid fusion of double criteria, the ring box D accurately judges the fault direction in a very short time, avoids the expansion of the fault or unnecessary full-line tripping, and greatly improves the reliability and recovery speed of power supply. Embodiment 5

[0079] The embodiment of the present application refers to the specific scene of Embodiment 3, and focuses on the local diagnosis result of "need to start cooperative positioning", which is focused on the cooperative positioning of waveform similarity and energy difference, which is described in detail in Figure 4 , and a flowchart for generating a comprehensive positioning result.

[0080] The fault occurs on a branch cable of G-H section. After the local diagnosis of ring net box G, the directive index is ambiguous due to the similar distance between the bus side and the feeder side, and the decision is to start the cooperative positioning. Ring net box G immediately requests the synchronous and differentially encoded fault transient data from adjacent terminals F and H through GOOSE message. For F-G section: the calculation result shows that the Pearson similarity coefficient is 0.98 (very high similarity), and the transient waveform intensity difference coefficient is 0.02 p.u. (very small difference); for G-H section: the calculation result shows that the Pearson similarity coefficient is 0.65 (similarity drops sharply), and the transient waveform intensity difference coefficient is 0.40 p.u. (significant difference). The system starts the comprehensive judgment, the transient characteristics of F-G section are consistent, so it is not a fault section; the transient waveform and energy characteristics of G-H section are very different, and it is determined as the fault section boundary.

[0081] Through the comparison of multi-terminal data and multi-dimensional feature analysis, the ambiguity of the local judgment of G terminal is successfully excluded, and the fault is accurately segmented and locked in G-H line, avoiding misoperation or blind road pulling.

[0082] Although the embodiments of the present application have been shown and described, it can be understood by those skilled in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A primary and secondary fusion ring main-tainer fault diagnosis method, characterized in that, The application relates to a method for realizing fault location of a power distribution network. The method comprises the following steps: Real-time synchronous acquisition of three-phase instantaneous voltage and instantaneous current signals of each feeder port and conversion into digitized transient data; Time window preprocessing of the digitized transient data based on a fault occurrence time, and extraction of zero sequence waveform data; Wavelet packet decomposition of the zero sequence waveform data to generate a transient feature vector; Component extraction and integral calculation of the zero sequence waveform data through a transient zero sequence power direction method to generate a polarity result; regression analysis and weighted summation of the transient feature vector through a pre-trained support vector machine model to output a directivity index; generation of a local diagnosis result based on the polarity result and the directivity index; When the local diagnosis result is that cooperative positioning needs to be started, the ring network box requests exchange of digitized transient data from adjacent ring network box terminals through a communication network; calculation of a transient current similarity coefficient and a transient waveform intensity difference coefficient based on the digitized transient data of the adjacent terminals, and generation of a comprehensive positioning result in combination with the local diagnosis result; Determination of a fault line according to the comprehensive positioning result, and generation of a remote control command; 2. The method according to claim 1, characterized in that, Execution of fault isolation operation based on the remote control command, and execution of a PHM process. The specific generation process of the digitized transient data is as follows: real-time acquisition of three-phase instantaneous voltage and current signals through a primary equipment sensor integrated in the ring network box; in the acquisition process, hardware level time stamp synchronization of the three-phase instantaneous voltage and current signals through a clock source, and ADC digitization of the three-phase instantaneous voltage and current signals through an analog-to-digital converter to generate digitized transient data; the digitized transient data is stored in RAM and is locked and extracted through a fault transient triggering mechanism.

3. The method according to claim 1, characterized in that, The fault transient triggering mechanism is based on detection of line current, and when the absolute change rate of zero sequence current and / or zero sequence voltage of any feeder port exceeds a threshold value, the fault transient triggering mechanism is triggered. The specific generation mode of the transient feature vector comprises the following steps: positioning of a fault occurrence time, selection of a preset time window before and after the fault occurrence time; superposition calculation of three-phase instantaneous voltage and current in the preset time window to extract zero sequence waveform data containing instantaneous zero sequence voltage and zero sequence current; multi-layer wavelet packet decomposition of the zero sequence waveform data, wherein a mother wavelet base is used and the number of decomposition layers is set to divide the entire frequency band of the zero sequence waveform data into a plurality of non-overlapping sub-bands; in the offline stage, a power distribution network simulation model is constructed by using an electromagnetic transient simulation software, different fault mechanisms under different working conditions are simulated through the power distribution network simulation model, and electromagnetic transient simulation results are generated; according to the electromagnetic transient simulation results, transient energy analysis is performed on each sub-band, and the sub-bands are selected based on transient energy concentration degree and transient energy ratio; wavelet packet energy values and cross-correlation coefficients of each selected sub-band are calculated, and all selected sub-bands are sorted in frequency increasing order, and finally the generated transient feature vector is composed of three parts: current energy sequence, voltage energy sequence and voltage current cross-correlation coefficient sequence sorted in frequency increasing order.

4. The method according to claim 1, characterized in that, The process of the local diagnosis result comprises: component extraction and integral calculation of zero sequence waveform data by a transient zero sequence power direction method to generate a polarity result; the specific process of the transient zero sequence power direction method is: digital filtering and differential processing of zero sequence voltage and zero sequence current of the zero sequence waveform data to extract transient zero sequence components; the differential processing is realized by a low-delay digital differential filter; transient zero sequence instantaneous power integral value calculation of the transient zero sequence components is performed by using an instantaneous power algorithm, and a polarity result is judged; regression analysis of transient feature vectors is performed by using a support vector machine model to output a multi-frequency admittance feature vector with the same dimension as the selected frequency band number; the support vector machine model uses a radial basis function as a kernel function, and the support vector machine model is trained based on different transient simulation data; each component of the multi-frequency admittance feature vector is a complex ratio of transient current signals and transient voltage signals in a corresponding frequency band; each component of the multi-frequency admittance feature vector is weighted and summed to generate a directivity index; the polarity result and the directivity index are logically fused to generate the local diagnosis result, and the local diagnosis result comprises a direction indication and a result type; when the directivity index and the polarity result both indicate the same direction, the result type of the local diagnosis result is local isolation; only when one of the following conditions is met: the fault directions represented by the polarity result and the directivity index are inconsistent, the reliability of the polarity result is lower than a preset reliability threshold, and the reliability of the directivity index is lower than the preset reliability threshold, the result type of the local diagnosis result is cooperative positioning.

5. The method according to claim 1, characterized in that, The specific process of the exchanged digitized transient data is: when the local diagnosis result is determined to be cooperative positioning, a request instruction for exchanging digitized transient data is sent to a physically directly connected adjacent ring network box terminal through a communication network inside the power distribution network; the request instruction requires that the digitized transient data returned by the adjacent terminal is in the same time window of the fault occurrence time; the digitized transient data is losslessly encoded by differential encoding.

6. The method according to claim 1, characterized in that, The generation process of the comprehensive positioning result is as follows: after the ring network box receives the synchronous zero sequence current waveform data of the adjacent terminal, the Pearson correlation coefficient between the zero sequence current waveform of the terminal and the zero sequence current waveform of the adjacent terminal is calculated as a transient current similarity coefficient; at the same time, the transient waveform intensity difference coefficient between the waveforms of the terminal and the adjacent terminal is calculated, the transient waveform intensity difference coefficient being the absolute value of the difference between the transient waveform intensity indexes of the waveforms of the terminal and the adjacent terminal, and the transient waveform intensity index being the integral of the square of the zero sequence current waveform; the calculated transient current similarity coefficient and transient waveform intensity difference coefficient are logically judged with the local diagnosis result; if the transient current similarity coefficient is greater than a preset similarity threshold value and the transient waveform intensity difference coefficient is less than a preset energy difference threshold value, it is determined that there is no fault section boundary between the two terminals; if the transient current similarity coefficient is less than the preset similarity threshold value and the transient waveform intensity difference coefficient is greater than the preset energy difference threshold value, it is determined that there is a first type of candidate fault section boundary between the two terminals; all the sections not belonging to the above logical judgment are determined as second type of candidate fault section boundaries; the ring network box determines the fault section boundary by comparing the transient current similarity coefficients and the transient waveform intensity difference coefficients of all adjacent sections and combining the direction indication of the local diagnosis result; the process of determining the fault section boundary is as follows: the sections determined as not having fault section boundaries are excluded; the first type of candidate fault section boundary and the second type of candidate fault section boundary are taken as candidate boundaries, and the candidate boundary consistent with the direction indication of the local diagnosis result is selected as the fault section boundary; a fault section identifier is generated based on the fault section boundary, and the fault section identifier is the comprehensive positioning result.

7. The method according to claim 1, characterized in that, The specific process of executing the fault isolation operation based on the remote control command is as follows: according to the comprehensive positioning result, the section located is mapped to a physical fault line by querying the distribution network topology and the equipment connection table of the ring network box; The ring network box generates a remote control command according to the built-in fault handling strategy model; Under the premise that the fault current has been cleared, the generated remote control command is directly sent to the intelligent operating mechanism of the circuit breaker integrated in the ring network box through the integrated digital interface, and the PHM process is started after the operation is completed, wherein the PHM process is as follows: the ring network box records the cumulative operation number, fault current level and operation time stamp of the equipment, and generates fault operation data; Based on the cumulative operation number and the current stress, a device health degradation model is constructed, the mechanical life and electrical life consumption of the equipment are evaluated based on the fault operation number through the device health degradation model, and the device health status index and the remaining service life are updated.

8. A primary and secondary fusion ring main-tainer fault diagnosis system, characterized in that, It comprises: A collection and perception module for synchronously collecting three-phase instantaneous voltage and instantaneous current signals of each feeder port in real time and converting them into digital transient data; A data processing and synchronization module for performing time window preprocessing on the digital transient data at the fault occurrence time, extracting zero sequence waveform data, and performing wavelet packet decomposition on the zero sequence waveform data to generate a transient feature vector; The local diagnosis decision module is used for component extraction and integral calculation of the zero sequence waveform data through the transient zero sequence power direction method to generate a polarity result; regression analysis and weighted summation of the transient feature vector are performed through a pre-trained support vector machine model to output a directivity index; and the local diagnosis result is generated based on the polarity result and the directivity index; The cooperative positioning and control execution module is used for requesting exchange of digitized transient data from adjacent ring network box terminals through a communication network when the local diagnosis result is that cooperative positioning needs to be started; transient current similarity coefficients and transient waveform intensity difference coefficients are calculated based on the digitized transient data of the adjacent terminals, and a comprehensive positioning result is generated in combination with the local diagnosis result; The fault line is determined according to the comprehensive positioning result, and a remote control command is generated; The fault isolation operation is performed based on the remote control command, and the PHM process is performed.

Images