A hierarchical fault diagnosis method for power distribution network

Abstract

The application particularly relates to a power distribution network hierarchical fault diagnosis method, and belongs to the field of intelligent fault diagnosis of power distribution networks, which comprises the following steps: analyzing protection action signals and circuit breaker action signals from alarm information when an actual fault of a power distribution network occurs, inputting the signals into a pre-constructed fuzzy time sequence Petri network fault diagnosis model to calculate a fault element confidence set, obtaining a fault cause set containing each possible fault cause and its confidence degree; based on a primary and secondary linkage simulation model consistent with the actual power distribution network, performing waveform simulation on each possible fault cause to generate simulation waveform data of each possible fault cause; and matching the simulation waveform data with actual fault recording waveform data in terms of fault characteristic waveforms, determining the possible fault cause corresponding to the simulation waveform data with the highest matching degree as the final diagnosis result. The application realizes rapid and accurate fault positioning and improves the efficiency and accuracy of fault processing of the power distribution network.

Description

Technical Field

[0001] This invention relates to the field of intelligent fault diagnosis technology for power system distribution networks, specifically to a hierarchical fault diagnosis method for distribution networks, applicable to complex fault analysis scenarios in digital and intelligent distribution networks. Background Technology

[0002] As power distribution networks evolve towards digitalization and intelligence, the structure of distribution network systems, especially secondary systems, is becoming increasingly complex. When a system failure occurs, maintenance personnel receive a massive amount of alarm information. The processing and judgment of these alarms heavily rely on the experience and skill of the maintenance personnel, which can easily lead to misdiagnosis and missed diagnosis, making it difficult to resolve the fault quickly. Therefore, researching computer-based fault diagnosis methods is of great practical significance in assisting maintenance personnel to conduct timely fault analysis and response, and to quickly restore the safe and reliable operation of the system.

[0003] Currently, common intelligent fault diagnosis methods mainly revolve around alarm information, constructing intelligent models to infer the cause of the fault. However, since different types of faults may produce the exact same alarm information—for example, single-phase grounding and two-phase grounding faults on the same busbar produce identical alarm information—this type of method may yield multiple different inferences, making it impossible to further determine the faulty phase and specific fault type, resulting in insufficient diagnostic accuracy.

[0004] In addition, there are some other problems in the existing technology: some methods based on communication message parsing are difficult to use protection action information to achieve automatic and rapid fault location after network failure; some methods based on port information evidence tables adopt single criteria and are difficult to accurately locate faults; some diagnostic models based on network message analyzers can achieve online diagnosis, but the model building is complex and the fault location accuracy is limited. Methods based on artificial intelligence, such as deep neural networks, require a large number of fault training samples and complex network structures, and are poorly adaptable to different topologies, making them difficult to apply uniformly.

[0005] Therefore, there is an urgent need for a power distribution network fault diagnosis method that can respond quickly, achieve precise location, and has good engineering adaptability. Summary of the Invention

[0006] The purpose of this invention is to solve the problems of vague diagnostic results and inability to accurately locate fault types and phases when relying solely on alarm information for fault diagnosis. It provides a hierarchical fault diagnosis method for distribution networks, which enables rapid and accurate fault location and improves the efficiency and accuracy of fault handling in distribution networks.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A hierarchical fault diagnosis method for distribution networks adopts a hierarchical diagnosis strategy, organically integrating "rapid inference based on alarm information" with "precise matching based on waveform data," and includes the following steps: Step 1: Construct a primary and secondary linkage simulation model on the simulation platform that is consistent with the actual power distribution network; Step 2: Obtain alarm information when the actual fault occurs in the distribution network; parse the protection action signal and circuit breaker action signal from the alarm information, and input them into the pre-constructed fuzzy time-series Petri net fault diagnosis model to calculate the confidence set of the fault element, and obtain a set of fault causes containing each possible fault cause and its confidence level; Step 3: Based on the aforementioned primary and secondary linkage simulation model, perform waveform simulation on each possible fault cause in the fault cause set to generate simulation waveform data for each possible fault cause. Step 4: Obtain actual fault waveform data when a fault occurs in the distribution network; perform fault characteristic waveform matching between the simulated waveform data of each possible fault cause and the actual fault waveform data, and determine the possible fault cause corresponding to the simulated waveform data with the highest matching degree with the actual fault waveform data as the final diagnostic result.

[0008] As one feasible approach, step one involves building a primary and secondary linkage simulation model consistent with the actual power distribution network on the CloudPSS Simstudio simulation platform, including the following steps: Real-time monitoring of actual measurement data in the distribution network, including bus voltage, branch current, and circuit breaker location; based on the aforementioned... Actual measurement data is used to calculate the outgoing load of each busbar and set it into a simulation model that is consistent with the actual distribution network, which is built on the simulation platform to obtain a primary and secondary linkage simulation model that is consistent with the actual distribution network.

[0009] As one possible approach, in step two, the fuzzy time-series Petri network fault diagnosis model is automatically generated based on the topology and protection configuration of the distribution network. The structure of the fuzzy time-series Petri net fault diagnosis model is divided into three layers: the first layer describes the fault combination of the target primary equipment and related secondary relay protection equipment; the second layer describes the correlation and time delay constraints between each fault combination and the corresponding protection action signal; and the third layer describes the control relationship and time delay constraints between the protection action signal and the controlled circuit breaker trip signal. The fuzzy time-series Petri net fault diagnosis model performs inference calculations based on the timing constraints and confidence levels of switching actions, outputs possible fault causes for each fault, including faulty components, fault types, and protection action states, and provides the confidence level for each possible fault cause, thus forming the fault cause set.

[0010] As one possible implementation, the fuzzy temporal Petri net fault diagnosis model is defined as a seven-tuple: ; in, It represents a finite set of locations used to describe the status of target primary equipment, secondary relay protection equipment, protection actions corresponding to faults, and circuit breaker status; It is the number of places in the set of places; It represents a finite set of transitions, used to describe the fault combination relationship between primary and secondary equipment, the time delay of fault-triggered protection actions, and the time delay of protection actions triggering circuit breaker actions. It is the number of transitions in the set of transitions; Defined The input matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There exist directed arcs. Then it means arrive There are no directed arcs between them; Defined The output matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There are directed arcs between them. Then it means arrive There are no directed arcs between them; This represents the time interval constraint vector of the location, used to constrain the timing of events; where Representative Treasury Time constraints; The confidence vector of a location represents the degree of credibility of an event occurring; where Representative Treasury Confidence level; This represents the time distance constraint vector between the associated databases of the transition, where Representing change Time distance constraints; In the above formula and .

[0011] As one feasible approach, the fuzzy temporal Petri net fault diagnosis model calculates the confidence set of faulty components, including the following steps: The fuzzy temporal Petri net fault diagnosis model performs inference calculations based on the confidence initialization rules of the location and the defined temporal constraint matrix. Specifically, the inference process proceeds step by step from the third layer to the first layer: first, the confidence of the protection action signal is updated according to the circuit breaker status; then, the confidence of the relay protection equipment status is updated according to the relationship between the protection action signal and the time delay; finally, the confidence of each target primary equipment incurring a fault is calculated. Possible causes of failure with a confidence level higher than a set threshold are selected to form a set of failure causes.

[0012] As one possible approach, in step three, based on the aforementioned primary and secondary linkage simulation model, the simulation script is traversed through a parallel fault enumeration tree to perform batch waveform simulation of each possible fault cause in the fault cause set.

[0013] As one possible approach, in step four, the simulated waveform data of each possible fault cause is matched with the actual fault recording data to determine the fault characteristic waveform. The possible fault cause corresponding to the simulated waveform data with the highest matching degree with the actual fault recording data is determined as the final diagnostic result. This specifically includes the following steps: The fundamental effective values ​​of the actual fault waveform data and the simulated waveform data of each possible fault cause are extracted by using the full-cycle Fourier algorithm and rolling calculation. The extracted fundamental effective value is differentially calculated to obtain the fault characteristic waveform reflecting the signal abrupt change characteristics; The Pearson correlation coefficient between the fault characteristic waveform of the actual fault recording and the fault characteristic waveform of each set of selected simulated waveform data is calculated as the matching degree. The possible causes of the fault corresponding to the set of simulated waveform data with the highest Pearson correlation coefficient are taken as the final diagnostic results.

[0014] As one possible implementation method, the full-cycle Fourier algorithm includes: Assuming the sampled analog signal x(t) It is a periodic time function containing the fundamental frequency, an unattenuated DC component, and all harmonics, expressed as: ; in, a n , b nThese represent the amplitudes of the sine and cosine terms for DC, fundamental frequency, and each harmonic, respectively. , , n =0,1,2…; By analyzing the analog signal x(t) A complete cycle T Integrate to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 : ; ; The integral is discretized and approximated using the trapezoidal method to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 for: ; ; in N This represents the number of sampling points per cycle of the fundamental frequency signal; x k It is the first k Sub-sample value; x 0、 x N They are respectively k =1 and k = N The sampled values; Based on the sinusoidal amplitude of the fundamental component a 1 Sum of cosine terms b 1 Calculate the fundamental effective value X 1 is: .

[0015] Compared with the prior art, the present invention has the following beneficial technical effects: Fast diagnosis: The hierarchical diagnosis mode is adopted. The first-level Petri net inference model quickly provides a narrowed fault range based on alarm information. The second-level waveform matching only needs to be compared within this small range, avoiding blind searching in the full sample and realizing rapid fault location as a whole.

[0016] High diagnostic accuracy: By introducing a second level of precise diagnosis based on fault waveform data, the sudden change characteristics of electrical quantities are used to effectively distinguish faults that cannot be identified at the alarm information level, and the diagnostic results are refined from "a set of possible faults" to "a unique fault cause", which greatly improves the accuracy of diagnosis.

[0017] High engineering adaptability: The simulation waveforms can be customized and generated by the simulation platform according to the actual distribution network topology and parameters. The fuzzy time series Petri net model can also be adjusted according to the updates of the distribution network topology and protection configuration. This avoids the problem that methods such as deep neural networks require a large amount of historical data for training and the model is fixed. It has good flexibility and adaptability. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the distribution network hierarchical fault diagnosis method of the present invention; Figure 2 This is the primary topology diagram of the 110kV substation in Example 2; Figure 3 This is a schematic diagram of the fault reasoning tree in Example 2; Figure 4 This is a comparison diagram of the fault characteristic waveforms of the fault recording and the simulated waveform in Example 2; Figure 5 A schematic diagram of a simple fuzzy Petri net fault diagnosis model; Figure 6 This is a schematic diagram of a simple fuzzy temporal Petri net fault diagnosis model of the present invention; Figure 7 This is a schematic diagram of a simple fuzzy temporal Petri net fault diagnosis model of the present invention; Figure 8 A schematic diagram of a fuzzy temporal Petri net fault diagnosis model of the present invention; Figure 9 This is a schematic diagram of fault feature waveform extraction according to the present invention. Detailed Implementation

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs; the terminology used herein in the specification of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] The distribution network hierarchical fault diagnosis method of the present invention combines actual operating data and simulation models to integrate information of the distribution network at multiple levels. It can respond quickly when a fault occurs, and provides a highly efficient and accurate fault analysis and location means for nuclear power plant distribution networks, ensuring the safe and stable operation of nuclear power plants.

[0022] The distribution network graded fault diagnosis method of the present invention includes the following steps: Step 1: Construct a primary and secondary linkage simulation model on the simulation platform that is consistent with the actual power distribution network; Step 2: Obtain alarm information when the actual fault occurs in the distribution network; parse the protection action signal and circuit breaker action signal from the alarm information, and input them into the pre-constructed fuzzy time-series Petri net fault diagnosis model to calculate the confidence set of the fault element, and obtain a set of fault causes containing each possible fault cause and its confidence level; Step 3: Based on the aforementioned primary and secondary linkage simulation model, perform waveform simulation on each possible fault cause in the fault cause set to generate simulation waveform data for each possible fault cause. Step 4: Obtain actual fault waveform data when a fault occurs in the distribution network; perform fault characteristic waveform matching between the simulated waveform data of each possible fault cause and the actual fault waveform data, and determine the possible fault cause corresponding to the simulated waveform data with the highest matching degree with the actual fault waveform data as the final diagnostic result.

[0023] In this invention, step one uses ClodPSS Simstudio as the simulation platform; constructing a primary and secondary linkage simulation model consistent with the actual power distribution network on the simulation platform includes the following steps: Real-time monitoring of actual measurement data in the distribution network, including bus voltage, branch current, and circuit breaker location; based on the aforementioned... Actual measurement data is used to calculate the outgoing load of each busbar and set it into a simulation model that is consistent with the actual distribution network, which is built on the simulation platform. This results in a primary and secondary linkage simulation model that is consistent with the actual distribution network. The primary and secondary linkage simulation model can reflect the actual operating state of the distribution network before the fault occurs, thus laying the foundation for fault diagnosis.

[0024] Fuzzy Petri net fault diagnosis models model system fault processes as dynamic systems of discrete events, making them easy to represent graphically and enabling fast reasoning based on matrix operations. In the research scenario of distribution network fault diagnosis, the main approach of applying fuzzy Petri net fault diagnosis models is to treat power grid equipment as storage locations, model their corresponding actions as transitions, and establish connection arcs based on protection configuration relationships. This allows for the creation of corresponding fuzzy Petri net fault diagnosis models for any equipment under study.

[0025] A fuzzy Petri net fault diagnosis model can be defined as the following quintuple: ; in, It represents a finite set of locations used to describe the status of target primary equipment, secondary relay protection equipment, protection actions corresponding to faults, and circuit breaker status; n It is the number of treasuries in the set of treasuries; It represents a finite set of transitions, used to describe the fault combination relationship between primary and secondary equipment, the time delay of fault-triggered protection actions, and the time delay of protection actions triggering circuit breaker actions. m It is the number of transitions in the set of transitions; Defined The input matrix represents the directed arc from the place to the transition; for any place P i and any change R j ,like represent P i arrive R j There are directed arcs between them. Then it means P i arrive R j There are no directed arcs between them; Defined The output matrix represents the directed arc from the place to the transition; for any place P i and any change R j ,like represent R j arrive P i There are directed arcs between them. Then it means R j arrive Pi There are no directed arcs between them; Denotes the confidence vector of a place, where Representative Treasury P i The confidence level; in the above formula and ; like Figure 5 This is a schematic diagram of a simple fuzzy Petri net fault diagnosis model. From right to left, the diagram represents the actual sequence of events, while from left to right, it represents the reasoning process.

[0026] Existing literature discloses a power system fault diagnosis method based on temporal-constrained fuzzy Petri Nets (Xu B, Yin X, Yin X, et al. Fault Diagnosis of Power Systems Based on TemporalConstrained Fuzzy Petri Nets[J / OL]. IEEE Access, 2019, 7: 101895-101904.). This method adds a confidence vector for locations to the fuzzy Petri net fault diagnosis model. The elements of the confidence vector range from 0 to 1, which can effectively handle the uncertainty in alarm information. In addition, locations and transitions in the fuzzy Petri net fault diagnosis model are given time attributes to represent the temporal constraints of alarm information, improving fault diagnosis performance. However, for systems with complex network topologies and protection configurations, it is difficult to achieve online adaptive modeling, resulting in poor adaptability and engineering practicality. Therefore, it is necessary to find a method that can achieve automatic modeling based on the characteristics of fault diagnosis.

[0027] This invention improves upon the existing fuzzy Petri net fault diagnosis model, enabling the automatic modeling of a fuzzy temporal Petri net fault diagnosis model through a fault enumeration tree traversal simulation method.

[0028] In this invention, a fuzzy temporal Petri net fault diagnosis model can be defined as the following seven-tuple: ; in, It represents a finite set of locations used to describe the status of target primary equipment, relay protection equipment, protection actions corresponding to faults, and circuit breaker status; It is the number of places in the set of places; It represents a finite set of transitions, used to describe the fault combination relationship between primary and secondary equipment, the time delay of fault-triggered protection actions, and the time delay of protection actions triggering circuit breaker actions. It is the number of transitions in the set of transitions; Defined The input matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There are directed arcs between them. Then it means arrive There are no directed arcs between them; Defined The output matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There are directed arcs between them. Then it means arrive There are no directed arcs between them; This represents the time interval constraint vector of the location, used to constrain the timing of events; where Representative Treasury Time constraints; The confidence vector of a location represents the degree of credibility of an event occurring; where Representative Treasury Confidence level; This represents the time distance constraint vector between the associated databases of the transition, where Representing change Time distance constraints; In the above formula and .

[0029] Because the topology of a distribution network may differ under different operating conditions, the direction of fault propagation may also differ when a fault occurs. Therefore, the fault diagnosis model needs to be adaptively updated according to changes in the topology. Thus, a topology analysis of the distribution network is necessary before establishing a fuzzy time-series Petri net fault diagnosis model. Under different operating conditions, the power supply path differs, resulting in different fault current paths and triggered protection actions when a short-circuit fault occurs. After setting the initial state of the simulation model based on distribution network measurement data and running the simulation, the power flow direction of each device component can be used to determine the power supply path, thereby analyzing the possible path of fault propagation during a fault.

[0030] Because the distribution network is a radial network, if the protection at the fault point fails to operate, the fault will propagate to the upstream power supply bus, causing the upstream equipment protection to operate. However, only the alarm information can be obtained from the alarms showing the operation of each circuit breaker protection, and the type of protection operation is unknown. Therefore, this invention can only establish a simple fuzzy time-series Petri net fault diagnosis model, such as... Figure 6 and 7 As shown in the diagram, the sequence of events from right to left represents the actual order in which they occurred, while the process of reasoning is represented from left to right.

[0031] The topology of existing fuzzy Petri net fault diagnosis models actually describes the possible paths of event evolution after a suspected fault in the target device. When establishing a fuzzy Petri net fault diagnosis model for a suspected faulty device, all possible fault propagation directions need to be considered. However, the fuzzy temporal Petri net fault diagnosis model proposed in this invention simulates various fault propagation scenarios caused by protection failure to operate by considering fault combinations of secondary relay protection devices. This invention employs... Figure 2 Taking the 110kV substation topology shown as an example, the improved fuzzy time-series Petri net fault diagnosis model structure is illustrated. Using the 110kV I line as an example, when a single-phase ground fault occurs, the protection and circuit breaker actions differ depending on the fault condition of the relay protection equipment. Based on the traversal results of the fault enumeration tree, the fuzzy time-series Petri net fault diagnosis model can be obtained as follows: Figure 8 As shown.

[0032] Depend on Figure 8 It can be seen that the fuzzy time-series Petri net fault diagnosis model can be divided into three layers. The first layer describes the fault combination of the target primary equipment and related relay protection equipment. The second layer describes the association and time delay constraints between the fault combination and the corresponding protection action. The third layer describes the control relationship and time delay constraints between the protection action signal and the tripping signal of the controlled circuit breaker.

[0033] Due to the hierarchical nature of the fuzzy temporal Petri net fault diagnosis model, places and transitions can be divided in the following ways: Places can be divided into... ,in Indicates the target device. This indicates the fault combination of the relay protection equipment. Indicates the protection action corresponding to the fault combination. Circuit breakers representing protection and control; transitions can be categorized as follows: ,in This indicates the relationship between a primary faulty device and a secondary relay protection device forming a fault combination. This indicates the delay in the protection action triggered by each fault combination. This represents the time delay from protection action to circuit breaker action; after the topology of the fuzzy timing Petri net fault diagnosis model is determined, its input matrix... and output matrix Then it can be obtained according to its definition.

[0034] In this invention, the inference and calculation process of the fuzzy temporal Petri net fault diagnosis model is as follows: 1. Initialize the confidence level based on the received alarm information: When a fault occurs in the distribution network, the protection action signal and circuit breaker action signal are parsed from the received alarm information and input into the fuzzy time-series Petri net fault diagnosis model. Considering the uncertainty of alarm information, that is, the possibility of missed alarms or false alarms, the confidence of the circuit breaker and relay protection equipment library is initialized according to the assignment rules in Table 1. Table 1 Initialization Assignment Rules for Circuit Breaker and Relay Protection Equipment Library 2. The third-layer inference calculation and update of the fuzzy temporal Petri net fault diagnosis model. P B3 : First, the confidence levels of the circuit breaker libraries connected to the same transition are averaged to obtain a new confidence level, expressed by the following formula: ,in, Indicates belonging to Layer transition R j The confidence level obtained from the reasoning; Indicates input to transition R j The number of warehouses; Indicates the location belonging to the layer; Represents the input matrix I 3. Zhongku Institute P i With change R j The corresponding element is 1; Indicates the place of storage P i Initial confidence level.

[0035] Then update the database of connected relay protection devices according to the following formula: ,in, Indicates belonging to P B3 Layered warehouse P i Updated confidence level; and These represent the confidence weights of alarm information from relay protection devices and alarm information from circuit breakers, respectively. Indicates the place of storage P i Initial confidence level; Indicates the previous step of change R j The confidence level obtained from the reasoning; Represents the output matrix O 3 changes R j With the warehouse P i The corresponding element is 1.

[0036] 3. The second-layer inference calculation of the fuzzy temporal Petri net fault diagnosis model yields... P B2 Confidence level: P B3 Push forward to the one connected to it R B2 With the confidence level unchanged, it can be expressed as follows: ,in, Indicates belonging to R B2 Layer changes R j Updated confidence level; Represents the input matrix I 2. Zhongku Institute P i With change R j The corresponding element is 1; Then it represents the relationship between the reasoning obtained in the previous step and the transition. R j Connected warehouses P i Confidence level.

[0037] The confidence level of PB2 is obtained by averaging the confidence levels of RB2 cells connected to the same library in PB2, and is expressed as follows: ;in, Indicates belonging to P B2 Layered warehouse P i Updated confidence level; Indicates output to the repository P i The number of changes; Represents the output matrix O 2 changes R j With the warehouse P iThe corresponding element is 1; Indicates the previous step of change R j The confidence level obtained from the reasoning.

[0038] 4. The first-layer inference calculation of the fuzzy temporal Petri net fault diagnosis model yields the fault confidence of suspected components: The first layer of inference in the fuzzy temporal Petri net fault diagnosis model only needs to filter out... P B2 The repository with the highest confidence level is then multiplied by its confidence weight. This is because the probability of relay protection equipment failing to operate is relatively low, therefore the confidence level of the main protection failure should be reduced to avoid misjudgment. The reasoning process is expressed by the following formula: ,in, Obtained through reasoning P B1 Confidence level of the layer library; For the treasury P i Confidence weights; Indicates taking P B2 The repository with the highest confidence level; 5. Filter out faulty components with a fault confidence level higher than a set threshold: After traversing and reasoning to calculate the fuzzy temporal Petri net fault diagnosis model for each faulty component to obtain the fault confidence of each component, suspicious components with fault confidence higher than a set threshold are screened out, and the entire preliminary fault diagnosis process ends.

[0039] Because the same alarm information may occur when different faults occur in a distribution network, the fuzzy time-series Petri net fault diagnosis model may yield multiple different inferences. Therefore, relying solely on alarm information for fault diagnosis cannot determine the faulty phase and type. This necessitates the introduction of electrical quantities from the distribution network, i.e., fault waveform data, for further fault diagnosis. Since electrical quantities undergo abrupt changes when a short-circuit fault occurs in a distribution network, these abrupt changes can be used as fault characteristics to diagnose different fault types.

[0040] This invention uses the abrupt change in the fundamental effective value of the waveform in fault recordings and simulations as the fault characteristic. To analyze the amplitude of the fundamental frequency components of voltage and current, a Fourier algorithm based on a periodic function model is typically used in practice. Depending on the sampling period used, the Fourier algorithm can be divided into full-cycle Fourier algorithms and half-cycle Fourier algorithms. The full-cycle Fourier algorithm uses samples from a continuous cycle to calculate the signal amplitude and phase angle, with a data window of 20ms. The half-cycle Fourier algorithm, on the other hand, uses only half a cycle of sampled data to calculate the signal amplitude and phase angle, with a data window of 10ms. Although the half-cycle Fourier algorithm is half a cycle faster than the full-cycle Fourier algorithm in terms of calculation speed, in terms of filtering characteristics, it can only filter out odd harmonic components and cannot filter out constant DC components and even harmonic components. Since the signal after a fault often contains attenuated DC components, the calculation error of the half-cycle Fourier algorithm is relatively large. In comparison, the full-cycle Fourier algorithm not only completely filters out various integer harmonic components and pure DC components, but also has a certain ability to suppress low-frequency components contained in non-integer high-frequency components and exponentially decaying aperiodic components. Therefore, in order to achieve more accurate calculation and analysis of signals, this invention uses the full-cycle Fourier algorithm to analyze the waveform amplitude. The implementation process of the algorithm is as follows: Assuming the sampled analog signal x(t) It is a periodic time function containing the fundamental frequency, an unattenuated DC component, and all harmonics, expressed as: ; in, x ( t () represents the sampled periodic analog signal; t It is a time variable; n For harmonic order, n =0 corresponds to the DC component, n=1 corresponds to the fundamental frequency, and n≥2 corresponds to each harmonic. ω 1 represents the fundamental angular frequency; X n For the first n The amplitude of the second harmonic; α n For the first n The initial phase of the subharmonic; a n represents the coefficients of the sine term in the Fourier series; b n Let be the coefficients of the cosine terms in the Fourier series, and satisfy the relation , ; Represents the sinusoidal component of the signal. Represents the cosine component of the signal; Based on the principle of Fourier series, by analyzing the analog signal... x(t) A complete cycle T Integrate to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 : ; ; The integral is discretized and approximated using the trapezoidal method to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 for: ; ; in N This represents the number of sampling points per cycle of the fundamental signal; x k It is the first k Sub-sample value; x 0、 x N They are respectively k =1 and k = N The sampled values; The integration process shows that the amplitudes of the fundamental sine and cosine components have eliminated the influence of the DC component and the integer harmonic component. Combined with the phasor form of the signal, the effective value of the fundamental frequency can be obtained. X 1 is: .

[0041] In practice, the analysis of phase components using the full-cycle Fourier transform algorithm can be approached in two ways. One approach is to divide the entire stage into units based on one cycle length, and then use the full-cycle Fourier transform algorithm to analyze and extract phase components for each stage. The other approach is to use a rolling method to achieve continuous effective value analysis, which takes each sampling point as the starting unit and recursively analyzes the phase components by moving backward or forward through one cycle of data windows. Since the latter method can more completely reflect the characteristics of the recorded waveform data, this invention adopts the latter method to achieve effective value analysis in order to improve the accuracy of the analysis.

[0042] After obtaining the fundamental RMS value of the waveform, the fault characteristic waveform can be obtained by taking the difference. For example... Figure 9The figure shows the original waveform of a fault recording, the fundamental amplitude waveform obtained after Fourier analysis, and the fault characteristic waveform obtained after differentiation. In the figure, 10kV_1M_Ua represents the A-phase voltage of the 1M bus of the 10kV substation. The original waveform represents the time-domain voltage signal directly acquired by the fault recording device, reflecting the instantaneous changes in electrical quantities before and after the fault. The fundamental amplitude represents the amplitude of the power frequency fundamental component extracted after Fourier analysis of the original waveform, used to characterize the steady-state change characteristics of the voltage signal. The change in fundamental amplitude represents the difference result of the fundamental amplitude at adjacent moments, used to characterize the degree of abrupt change of the fundamental component over time. The fault characteristic waveform effectively highlights the signal abrupt change characteristics caused by the fault. After diagnosis by the fuzzy time-series Petri net fault diagnosis model, comparing the fault characteristic waveforms of the fault recording and the simulated waveform can effectively diagnose the specific fault type and location.

[0043] To compare the fault characteristic waveforms of fault recordings and simulated waveforms, this invention uses the Pearson coefficient to quantify the similarity between the two differential waveforms.

[0044] The Pearson coefficient, also known as the Pearson product-moment correlation coefficient, is a statistical indicator used to measure the degree of linear correlation between two variables. The Pearson coefficient ranges from -1 to 1, where 1 indicates a perfect positive linear correlation (an increase in one variable results in a linear increase in the other); -1 indicates a perfect negative linear correlation (an increase in one variable results in a linear decrease in the other); and 0 indicates no linear correlation (no linear relationship between the two variables). The formula for calculating the Pearson coefficient is: ; in, r It is the Pearson coefficient. x i and y i They are variables X and Y The i One value; and They are variables X and Y The mean.

[0045] The Pearson coefficient is commonly used in statistics, data analysis, and machine learning to determine the relationships and correlations between variables. This coefficient effectively determines whether two variables have a linear relationship, thus providing a basis for further analysis. In this invention, it is used to compare the similarity of fault characteristic waveforms between recorded fault waveforms and simulated waveforms, thereby identifying the most probable cause of the fault.

[0046] The distribution network graded fault diagnosis method of the present invention will be described in detail below through specific embodiments.

[0047] Example 1 Taking a three-phase short-circuit fault occurring on busbar 5ESB as an example, the received alarm message is: a three-phase short-circuit fault has occurred on busbar 5ESB.

[0048] The protection action signal and circuit breaker position change signal are parsed from the alarm information and input into a pre-constructed fuzzy time-series Petri net fault diagnosis model for calculating the confidence set of fault elements. The resulting set of fault causes, including various possible fault causes and their confidence levels, is as follows: a fault occurs in bus 5ESB and the main protection of bus 5ESB operates, with a confidence level of 0.87875; a fault occurs in an outgoing branch of bus 5ESB, but the main protection of that branch fails to operate, causing the fault to propagate and causing the main protection of bus 5ESB to operate, with a confidence level of 0.74000; a fault occurs in bus 5EMA and the main protection of bus 5EMA fails to operate, causing the fault to propagate and causing the main protection of the upstream bus 5ESB to operate, with a confidence level of 0.74000.

[0049] Based on a primary and secondary linkage simulation model built on a simulation platform that is consistent with the actual substation topology, waveform simulation is performed on each possible fault cause in the fault cause set to generate simulation waveform data for each possible fault cause.

[0050] The simulated waveform data of each possible fault cause was matched with the actual fault recording data to determine the fault cause with the highest matching degree with the actual fault recording data as "bus 5ESB fault occurred and bus 5ESB main protection was activated". This was determined as the final accurate diagnosis result, which was completely consistent with the actual fault.

[0051] Example 2 Taking a typical 10kV sectionalizing switch failure in a 110kV substation as an example. The primary topology of this substation is as follows: Figure 2 As shown in Table 1, the received alarm information is as follows.

[0052] Table 1 Alarm Information The protection action signals and circuit breaker position change signals are parsed from the alarm information and input into a pre-constructed fuzzy time-series Petri net fault diagnosis model for calculating the confidence set of the fault element. This yields a fault cause set containing all possible fault causes and their confidence levels. In this fault cause set, the fault element with the highest confidence level is the 10kV sectionalizing switch 532, which causes the backup protection of the low-voltage branch 2 of main transformer #2 and the low-voltage side of main transformer #3 to operate. The fault reasoning tree is as follows: Figure 3 As shown.

[0053] Based on a primary and secondary linkage simulation model built on a simulation platform that is consistent with the actual substation topology, waveform simulation is performed on each possible fault cause in the fault cause set to generate simulation waveform data for each possible fault cause. Specifically, in the primary and secondary linkage simulation model, the fault point is set as the 10kV section 532 switch, and the fault causes are set as single-phase short-circuit to ground, two-phase short circuit, two-phase short-circuit to ground, and three-phase short circuit in sequence. Then, the primary and secondary linkage simulation model is run to obtain the simulation waveform data for each possible fault cause.

[0054] By matching the simulated waveform data of various possible fault causes with the actual fault recording data, the most likely fault cause with the highest matching degree with the actual fault recording data was determined to be a two-phase short circuit to ground in phases A and B of the 10kV sectional 532 switch. Some of the fault characteristic waveforms were compared to... Figure 4 As shown, the waveforms of the two fault characteristics fit well, and this is determined to be the final accurate diagnosis result. The diagnosis result is completely consistent with the actual fault, which can provide some guidance for maintenance personnel to analyze the waveform data.

[0055] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, all of which fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

Claims

1. A method for hierarchical fault diagnosis in a power distribution network, characterized in that, Includes the following steps: Step 1: Construct a primary and secondary linkage simulation model on the simulation platform that is consistent with the actual power distribution network; Step 2: Obtain alarm information when an actual fault occurs in the distribution network; The protection action signal and circuit breaker action signal are parsed from the alarm information and input into the pre-built fuzzy time-series Petri net fault diagnosis model to calculate the confidence set of the fault element, thereby obtaining a set of fault causes containing each possible fault cause and its confidence level. Step 3: Based on the aforementioned primary and secondary linkage simulation model, perform waveform simulation on each possible fault cause in the fault cause set to generate simulation waveform data for each possible fault cause. Step 4: Obtain actual fault waveform data when a fault occurs in the distribution network; perform fault characteristic waveform matching between the simulated waveform data of each possible fault cause and the actual fault waveform data, and determine the possible fault cause corresponding to the simulated waveform data with the highest matching degree with the actual fault waveform data as the final diagnostic result.

2. The method for hierarchical fault diagnosis of distribution networks according to claim 1, characterized in that, Step one involves building a primary and secondary linkage simulation model on the simulation platform that is consistent with the actual power distribution network, including the following steps: Real-time monitoring of actual measurement data in the distribution network, including bus voltage, branch current, and circuit breaker location; based on the aforementioned... Actual measurement data is used to calculate the outgoing load of each busbar and set it into a simulation model that is consistent with the actual distribution network, which is built on the simulation platform to obtain a primary and secondary linkage simulation model that is consistent with the actual distribution network.

3. The method for hierarchical fault diagnosis of distribution networks according to claim 1, characterized in that, In step two, the fuzzy time-series Petri network fault diagnosis model is automatically generated based on the topology and protection configuration of the distribution network. The structure of the fuzzy time-series Petri net fault diagnosis model is divided into three layers: the first layer describes the fault combination of the target primary equipment and related secondary relay protection equipment; the second layer describes the correlation and time delay constraints between each fault combination and the corresponding protection action signal; and the third layer describes the control relationship and time delay constraints between the protection action signal and the controlled circuit breaker trip signal. The fuzzy time-series Petri net fault diagnosis model performs inference calculations based on the timing constraints and confidence levels of switching actions, outputs possible fault causes for each fault, including faulty components, fault types, and protection action states, and provides the confidence level for each possible fault cause, thus forming the fault cause set.

4. The method for hierarchical fault diagnosis of distribution networks according to claim 3, characterized in that, The fuzzy temporal Petri net fault diagnosis model is defined as a seven-tuple: ； in, It represents a finite set of locations used to describe the status of target primary equipment, secondary relay protection equipment, protection actions corresponding to faults, and circuit breaker status; It is the number of places in the set of places; It represents a finite set of transitions, used to describe the fault combination relationship between primary and secondary equipment, the time delay of fault-triggered protection actions, and the time delay of protection actions triggering circuit breaker actions. It is the number of transitions in the set of transitions; Defined The input matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There exist directed arcs. Then it means arrive There are no directed arcs between them; Defined The output matrix represents the directed arc from the place to the transition; for any place and any change ,like represent arrive There exist directed arcs. Then it means arrive There are no directed arcs between them; This represents the time interval constraint vector of the location, used to constrain the timing of events; where Representative Treasury Time constraints; The confidence vector of a location represents the degree of credibility of an event occurring; where Representative Treasury Confidence level; This represents the time distance constraint vector between the associated databases of the transition, where Representing change Time distance constraints; In the above formula and .

5. The method for hierarchical fault diagnosis of distribution networks according to claim 3, characterized in that, The calculation of the confidence set of faulty components in the fuzzy time-series Petri net fault diagnosis model includes the following steps: The fuzzy temporal Petri net fault diagnosis model performs inference calculations based on the confidence initialization rules of the location and the defined temporal constraint matrix. Specifically, the inference process proceeds step by step from the third layer to the first layer: first, the confidence of the protection action signal is updated according to the circuit breaker status; then, the confidence of the relay protection equipment status is updated according to the relationship between the protection action signal and the time delay; finally, the confidence of each target primary equipment incurring a fault is calculated. Possible causes of failure with a confidence level higher than a set threshold are selected to form a set of failure causes.

6. The method for hierarchical fault diagnosis of distribution networks according to claim 1, characterized in that, In step three, based on the first- and second-stage linkage simulation model, the simulation script is traversed by parallel fault enumeration tree to perform batch waveform simulation of each possible fault cause in the fault cause set.

7. The method for hierarchical fault diagnosis of distribution networks according to claim 1, characterized in that, In step four, the simulated waveform data of each possible fault cause is matched with the actual fault recording data to determine the fault characteristic waveform. The possible fault cause corresponding to the simulated waveform data with the highest matching degree with the actual fault recording data is determined as the final diagnostic result. This step includes the following steps: The fundamental effective values ​​of the actual fault waveform data and the simulation waveform data of each possible fault cause are extracted by using the full-cycle Fourier algorithm in a rolling calculation manner. The extracted fundamental effective value is differentially calculated to obtain the fault characteristic waveform reflecting the signal abrupt change characteristics; The Pearson correlation coefficient between the fault characteristic waveform of the actual fault recording and the fault characteristic waveform of each set of selected simulated waveform data is calculated as the matching degree. The possible causes of the fault corresponding to the set of simulated waveform data with the highest Pearson correlation coefficient are taken as the final diagnostic results.

8. The method for hierarchical fault diagnosis of distribution networks according to claim 7, characterized in that, The full-cycle Fourier algorithm includes: Assuming the sampled analog signal x(t) It is a periodic time function containing the fundamental frequency, an unattenuated DC component, and all harmonics, expressed as: in, a n , b n These represent the amplitudes of the sine and cosine terms for DC, fundamental frequency, and each harmonic, respectively. , , n =0,1,2…; By analyzing the analog signal x(t) A complete cycle T Integrate to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 : ； ； The integral is discretized and approximated using the trapezoidal method to obtain the amplitude of the sinusoidal term of the fundamental component. a 1 Sum of cosine terms b 1 for: ； ； in N This represents the number of sampling points per cycle of the fundamental signal; x k It is the first k Sub-sample value; x 0、 x N They are respectively k =1 and k = N The sampled values; Based on the sinusoidal amplitude of the fundamental component a 1 Sum of cosine terms b 1 Calculate the fundamental effective value X 1 is: 。

Images