Power grid fault identification method, system and device based on fuzzy theory and improved D-S evidence theory, and storage medium

Abstract

The invention discloses a power grid fault identification method, system and device based on a fuzzy theory and an improved D-S evidence theory, and a storage medium, and relates to the technical field of power system fault diagnosis, and the method comprises the steps: obtaining the original data of a power grid, including fault voltage and current, and the action state information of a protection relay and a circuit breaker; and carrying out noise reduction on the fault voltage and current, and classifying the denoised data by using a support vector machine classification model optimized by a fruit fly algorithm to obtain an electrical quantity fault diagnosis result. Switching value information is mined through timing constraint screening, and error information is discriminated and corrected; matrix iteration is carried out on the corrected switching value to obtain a fault probability, and finally, an electrical quantity diagnosis result and the switching value fault probability are fused to accurately judge a fault. According to the method, the electrical quantity information with relatively good timeliness and accuracy and the switching quantity information with relatively good redundancy are subjected to information synthesis, so that the characteristics of the information can be brought into full play, and a fault element can be accurately and quickly judged.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power system fault diagnosis, and in particular to a power grid fault identification method, system and device based on fuzzy theory and improved D-S evidence theory and a storage medium. BACKGROUND

[0002] The power grid in high-altitude complex mountainous areas is influenced by geographical environment, climate conditions and new energy access, and has complex fault modes. Traditional fault identification methods based on a single data source have great limitations in accuracy, real-time performance and robustness.

[0003] At present, traditional fault identification mainly adopts a method based on a single data source, relies on observation of the state of a protection device for fault diagnosis, and determines a fault region according to switch information. When a system element fails, a relay protection device acts, a circuit breaker trips, and a traditional analysis determines a fault according to these conditions. However, this method has obvious limitations and performs poorly in accuracy, real-time performance and robustness, and cannot solve problems such as misoperation or refusal of a protection device, damage to an information channel and incomplete information. In particular, when multiple faults occur and multiple alarm signals are generated, it is difficult to adapt to a complex power system. SUMMARY

[0004] In view of the above problems, the present application provides a power grid fault identification method, system, device and storage medium based on fuzzy theory and improved D-S evidence theory.

[0005] Therefore, the technical problem solved by the present application is how to provide an intelligent fault identification method that fuses multiple data sources and considers complex working conditions to improve the reliability and safety of power grid operation.

[0006] To solve the above technical problems, the present application provides the following technical solutions. In a first aspect, the present application provides a power grid fault identification method based on fuzzy theory and improved D-S evidence theory, comprising: obtaining power grid raw data including fault voltage and current information and action state information of protection relays and circuit breakers; performing noise reduction processing on the obtained fault voltage and current information, training a support vector machine classification model optimized by a fruit fly algorithm using historical fault sample data, and classifying real-time fault electrical quantity features after noise reduction to obtain an electrical quantity fault diagnosis result; using a time sequence constraint relationship to screen and mine the action state information of protection relays and circuit breakers, automatically identifying and correcting false fault information in switch information through reverse and forward time sequence reasoning; performing matrix iteration operation on the corrected switch information to obtain a switch fault probability; The electrical quantity fault diagnosis result and the switching quantity fault probability are fused to realize accurate discrimination of the fault.

[0007] As an optimal solution of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, wherein: The obtained fault voltage and current information are subjected to noise reduction processing, a support vector machine classification model optimized by a fruit fly algorithm is trained using historical fault sample data, and the real-time fault electrical quantity features after noise reduction are classified to obtain the electrical quantity fault diagnosis result, which includes: The fault voltage and current information are subjected to noise reduction processing, the modal component, the center frequency and the Lagrange multiplier are updated by iteration, and it is judged whether the iteration is stopped according to the preset convergence error to obtain the noise reduction result. The support vector machine classification model optimized by the fruit fly algorithm is trained using historical fault sample data, the initial position of the fruit fly population is determined, the fruit fly population searches for the optimal position through multiple iterations, and the parameter training result is output.

[0008] As an optimal solution of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, wherein: The obtained fault voltage and current information are subjected to noise reduction processing, a support vector machine classification model optimized by a fruit fly algorithm is trained using historical fault sample data, and the real-time fault electrical quantity features after noise reduction are classified to obtain the electrical quantity fault diagnosis result, which includes: The real-time fault electrical quantity features after noise reduction are classified using the trained support vector machine classification model, the combined deviation of the sequences on the left and right of a certain point in the fault data sequence is calculated and minimized, the fault features are extracted, the power grid fault is classified according to the fault features, the classification result is matched with the fault type, and the electrical quantity fault diagnosis result is obtained.

[0009] As an optimal solution of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, wherein: The action state information of the protection relay and the circuit breaker is screened and mined using the time sequence constraint relationship, the false fault information in the switching quantity is automatically identified and corrected through reverse and forward time sequence reasoning, which includes: The power outage area is determined, and the elements in the area are regarded as suspicious fault elements; the related alarm information is divided for each suspicious fault element to form a set; the one-time point constraint of the fault occurrence is obtained by reverse time sequence reasoning on each set using the actual alarm time sequence information.

[0010] As an optimal solution of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, wherein: The action state information of the protection relay and the circuit breaker is screened and mined by using the time constraint relationship, and the false fault information in the switch quantity is automatically identified and corrected through reverse and forward time sequence reasoning, and the method further comprises the following steps of: The constraints of the sets are combined to obtain total constraints of the monadic time point of the fault occurrence, and the total constraints are subjected to forward time sequence reasoning to obtain the monadic time point constraints of each library, and the actual acquired time sequence information is compared with the monadic time point constraints of each library through specific operators to identify the false alarm information that does not satisfy the time sequence information.

[0011] As an optimal scheme of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, the method comprises the following steps of: The electrical quantity fault diagnosis result and the switch quantity fault probability are fused to realize accurate identification of the fault, and the method comprises the following steps of: The electrical quantity fault diagnosis result and the switch quantity fault probability are fused to realize accurate identification of the fault, and the method comprises the following steps of: The average evidence of the total evidence set is calculated, the deviation of each evidence from the average evidence is calculated, and the improved distance function is used to measure the deviation.

[0012] The beneficial effects of the optimal technical scheme are as follows: the two types of evidence are integrated under the same identification framework, the different types of evidence can be processed under a unified system, the subsequent fusion operation is facilitated, and the efficiency and accuracy of the fusion are improved.

[0013] As an optimal scheme of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, the method comprises the following steps of: The electrical quantity fault diagnosis result and the switch quantity fault probability are fused to realize accurate identification of the fault, and the method comprises the following steps of: The average value of the deviation of each evidence from the average evidence is calculated, and the evidence with a deviation value greater than the average value is regarded as abnormal evidence. The data source is corrected, the average evidence is used to replace the abnormal evidence to obtain a corrected evidence set, and the D-S synthesis formula is used to fuse the information of the corrected data source.

[0014] The beneficial effects of the optimal technical scheme are as follows: the average evidence is used to replace the abnormal evidence to correct the data source and remove unreliable information; the D-S synthesis formula is used for information fusion to effectively integrate the corrected evidence and obtain a more accurate fault identification result.

[0015] In a second aspect, the present application provides a power grid fault identification system based on the fuzzy theory and the improved D-S evidence theory, which comprises: A data acquisition module is configured to acquire original power grid data including fault voltage, current information, and action state information of protection relays and circuit breakers. An electrical quantity diagnosis module is configured to perform noise reduction processing on the acquired fault voltage and current information, train a support vector machine classification model optimized by a fruit fly algorithm using historical fault sample data, and classify real-time fault electrical quantity features after noise reduction to obtain an electrical quantity fault diagnosis result. A switch quantity correction module is configured to filter and mine the action state information of protection relays and circuit breakers using a time sequence constraint relationship, automatically identify and correct error fault information in the switch quantity through reverse and forward time sequence reasoning. A probability calculation module is configured to perform matrix iteration operation on the corrected switch quantity information to obtain a switch quantity fault probability. A result fusion module is configured to fuse the electrical quantity fault diagnosis result and the switch quantity fault probability to realize accurate fault discrimination.

[0016] In a third aspect, the present application provides a computer device, comprising: a memory and a processor; The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions, which realize the steps of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory when executed by the processor.

[0017] In a fourth aspect, the present application provides a computer readable storage medium storing computer executable instructions, which realize the steps of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory when executed by the processor.

[0018] The present application has the following beneficial effects: The present application overcomes the limitation that a single data source leads to inaccurate fault judgment, introduces electrical quantity information into power grid fault diagnosis, extracts fault features using the redundancy and diversity of multi-source fault information, and comprehensively diagnoses faults of both electrical quantities and switch quantities, which can effectively improve the accuracy and fault tolerance of fault diagnosis, and has good practical value and practical significance for power grid fault diagnosis. BRIEF DESCRIPTION OF DRAWINGS

[0019] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0020] Figure 1is the overall flow chart of the power grid fault identification method based on fuzzy theory and improved D-S evidence theory provided by the application.

[0021] Figure 2 is the three-stage protection time sequence constraint relationship diagram of the power grid fault identification method based on fuzzy theory and improved D-S evidence theory provided by the application. DETAILED DESCRIPTION

[0022] In order to make the above-mentioned objects, features and advantages of the present application more apparent and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are 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 should belong to the protection scope of the present application.

[0023] Embodiment 1, refer to Figure 1 As the first embodiment of the present application, the embodiment provides a power grid fault identification method based on fuzzy theory and improved D-S evidence theory, comprising: S1: obtaining power grid original data including fault voltage, current information and action state information of protection relays and circuit breakers; S2: performing noise reduction processing on the obtained fault voltage and current information, training a support vector machine classification model optimized by a fruit fly algorithm using historical fault sample data, and classifying the real-time fault electrical quantity features after noise reduction to obtain electrical quantity fault diagnosis results; S3: using time sequence constraint relationship to screen and mine the action state information of protection relays and circuit breakers, automatically identifying and correcting the error fault information in the switch quantity through reverse and forward time sequence reasoning; S4: performing matrix iteration operation on the corrected switch quantity information to obtain switch quantity fault probability; S5: fusing the electrical quantity fault diagnosis results and the switch quantity fault probability to realize accurate fault discrimination.

[0024] It should be noted that through steps S1-S5, a complete and efficient power grid fault identification system is constructed. From the acquisition of original data, to the processing and analysis of electrical quantity and switch quantity information respectively, and then to the final fusion of the two to realize accurate fault discrimination, the deficiencies of traditional fault identification methods in accuracy, real-time performance and robustness are effectively overcome, which can quickly and accurately locate faults in complex power grid operating environment, especially in the face of complex fault modes under harsh conditions such as high-altitude complex mountainous areas, providing a strong guarantee for the stable operation and timely repair of power grids, and significantly improving the efficiency and reliability of power grid fault diagnosis.

[0025] Embodiment 2, refer to Figures 1-2 For an embodiment of the present application, a power grid fault identification method based on fuzzy theory and improved D-S evidence theory is provided based on the previous embodiment, comprising: In the embodiment, the power grid raw data obtained in the above step S1 includes fault voltage, current information, and action state information of protection relays and circuit breakers, comprising: The data source is obtained through devices such as SCADA (Supervisory Control And Data Acquisition, data acquisition and monitoring control system) and EMS (Energy Management System, energy management system).

[0026] In the embodiment, the fault voltage and current information obtained in the above step S2 are subjected to noise reduction processing, the support vector machine classification model optimized by the fruit fly algorithm is trained using historical fault sample data, and the real-time fault electrical quantity features after noise reduction are classified to obtain electrical quantity fault diagnosis results, comprising: The obtained fault voltage, current, protection and circuit breaker action state information are subjected to VMD (Variational Mode Decomposition) algorithm noise reduction to remove signal burr phenomenon; Specifically, the maximum decomposition layer number K value is set, the iteration number counter n=0 is initialized 、 、 For 、 、 . Wherein, is the frequency domain representation of the first modal component at the first iteration, is the center frequency of the first modal component at the first iteration, is the frequency domain representation of the Lagrange multiplier at the first iteration; constantly update 、 and , denoted as: Wherein, is the frequency domain representation of the original signal, is a quadratic penalty factor in the VMD algorithm, used to balance the smoothness of the signal and the accuracy of the decomposition; is the update step size of the Lagrange multiplier.

[0027] determine whether the iteration stopping condition is met, if it is met, the iteration is stopped, and the final modal component and the center frequency is obtained; if it is not met, the iteration is continued, and the iteration stopping condition is expressed as: wherein, represents the preset convergence error.

[0028] In another possible implementation, when the acquired fault voltage and current information is denoised, a wavelet threshold denoising method can be used. First, a suitable wavelet basis function and decomposition layer number are selected to perform wavelet decomposition on the fault voltage and current information to obtain wavelet coefficients at different scales. Then, the wavelet coefficients are processed according to a set threshold rule (such as a soft threshold or a hard threshold) to remove the wavelet coefficients corresponding to the noise. Finally, the processed wavelet coefficients are reconstructed to obtain the denoised fault voltage and current information.

[0029] In another possible implementation, when the acquired fault voltage and current information is denoised, an empirical mode decomposition (EMD) denoising method can also be used to decompose the fault voltage and current information into a plurality of intrinsic mode functions (IMFs). By calculating the correlation coefficient or energy distribution of each IMF, it is determined which IMFs mainly contain noise components, and these IMFs are removed. Finally, the remaining IMFs are reconstructed to realize the denoising of the fault voltage and current information.

[0030] The support vector machine classification model optimized by the fruit fly algorithm is trained by using historical fault sample data. Specifically, a sample set composed of historical fault sample data is read in, a fruit fly population with a number of n is determined, and an initial position thereof is randomly generated; The direction and distance of the random search of the fruit fly individual are determined; The concentration of the position of the fruit fly is estimated by the distance between the position and the origin; The sample parameters are combined into the support vector machine model, the absolute value of the error is taken as the taste judgment function, and the taste concentration of the position of the fruit fly is calculated; The fruit fly with the highest taste concentration in the fruit fly population is found, and its optimal concentration value and position coordinates are recorded, and then all fruit flies fly to this position; Multiple iterations are performed until the optimal taste concentration is obtained, and the parameter training result is output.

[0031] The trained support vector machine classification model is used to classify the noise-reduced real-time fault electrical quantity features to obtain electrical quantity fault diagnosis results.

[0032] Specifically, assuming a power grid fault data sequence containing N sampling points is X(t), calculate the number of sampling points in the left and right parts of the sequence from point i in X(t). and The mean of these two points is then used , Describe the standard deviation using... , The following describes the calculation of the merging deviation between the two sequences at point i, using the following formula: Where represents the merging deviation between the two sequences at point i, making Minimization, which means reducing the error in fault data acquisition to the lowest possible level, is the basis for fault feature extraction. The calculation formula is as follows: In the formula and Let these represent the feedforward weight vector and the feedback weight vector of the j-th fault signal, respectively. and These represent the feedback signal and the output signal of the fault signal, respectively.

[0033] The power grid faults are classified based on the feature extraction results, and the calculation formula is as follows: In the formula, N represents the total amount of sample data.

[0034] Based on the multi-level progressive structure of fault classification and different levels of fault type identification methods, the classification results are matched with the fault types to achieve fault diagnosis. The results are expressed as follows: in, The fault diagnosis results are represented by matching the classification results with the fault types.

[0035] In another possible implementation, when training a support vector machine classification model optimized by the fruit fly algorithm using historical fault sample data, a combination of random initialization and cluster initialization can be used to determine the initial positions of the fruit fly population. First, a subset of fruit fly initial positions are randomly generated. Then, cluster analysis is performed on the historical fault sample data, and the cluster centers are used as the initial positions for another subset of fruit flies. During the iteration process, an adaptive step size adjustment mechanism is introduced to dynamically adjust the step size based on the convergence of the fruit fly population, thereby accelerating the convergence speed.

[0036] In another possible implementation, when training the support vector machine classification model optimized by the fruit fly algorithm using historical fault sample data, the crossover and mutation operations of the genetic algorithm can also be introduced in the iteration process of the fruit fly algorithm. When the fruit fly population is iterated to a certain number of generations, a part of the fruit flies are randomly selected for crossover operation to exchange their part of position information; at the same time, a part of the fruit flies are subjected to mutation operation to randomly change a certain dimension of their position information. In this way, the diversity of the fruit fly population can be increased, the algorithm can be prevented from falling into local optimum, and the training effect of the support vector machine classification model can be improved.

[0037] In the embodiment, the screening and mining of the action state information of the protection relay and the circuit breaker in step S3 by using the time sequence constraint relationship includes: It should be noted that the screening and mining of the action state information of the protection relay and the circuit breaker by using the time sequence constraint relationship based on the weighted fuzzy Petri net includes that the library can represent a fault element, a protection, a circuit breaker and a virtual library, a token represents a state of the library, and a transition represents an event. According to the principle of relay protection, when a fault occurs in the power grid, the electrical quantity changes, the protection device performs setting calculation on the detected electrical quantity, and if the action condition is met, an action instruction is sent, and the corresponding circuit breaker trips immediately after receiving the action instruction to remove the fault element. The occurrence of the above events is distributed within a certain time range and is mutually coordinated and constrained in time, and there is a certain time sequence constraint relationship. Therefore, reasonable use of the fault alarm time sequence information can effectively improve the fault diagnosis performance.

[0038] According to the principle of relay protection, in three-stage protection, the cooperation of the protections at all levels and the action of the circuit breakers meet a certain time delay relationship. As shown in FIG. 1, 1 represents main protection action, 3 represents near backup protection action, and 5 represents remote backup action. 2, 4 and 6 represent the corresponding circuit breakers. Figure 2

[0039] The time distance constraints from the fault occurrence time to the action time of the main protection, the near backup protection and the remote backup protection are defined as follows: ​​​​​​​​ wherein, is the time point of fault occurrence, is the time point of main protection action, is the time point of near backup protection action, is the time point of far backup protection action, is the exact time distance from the time point of fault occurrence to the time point of main protection action, is the uncertainty of the time distance from the time point of fault occurrence to the time point of main protection action, is the exact time distance from the time point of fault occurrence to the time point of near backup protection action, is the uncertainty of the time distance from the time point of fault occurrence to the time point of near backup protection action, is the exact time distance from the time point of fault occurrence to the time point of far backup protection action, is the uncertainty of the time distance from the time point of fault occurrence to the time point of far backup protection action, the time interval unit is ms, and the subscript c is fault occurrence, the subscript m is main protection, the subscript p is near backup protection, and the subscript s is far backup protection.

[0040] The delay interval of each level of protection and the corresponding circuit breaker action is defined as: wherein, the subscript r represents each level of protection triggering the circuit breaker, and the subscript b represents the circuit breaker corresponding to each level of protection, i.e. is the time point of each level of protection triggering the circuit breaker, is the time point of the circuit breaker corresponding to each level of protection, is the exact time distance from the time point of each level of protection to the time point of the corresponding circuit breaker, is the uncertainty of the time distance from the time point of each level of protection to the time point of the corresponding circuit breaker.

[0041] According to the principle of relay protection, there are inherent action times of fault occurrence, each level of protection device, and the corresponding circuit breaker, the alarm information time points generated by the three are distributed within a certain range, and are mutually coordinated and constrained in time. The time sequence constraint relationship includes unary time point constraint and binary time distance constraint.

[0042] Specifically, the unary time point constraint is: the time point of event occurrence is defined as t, and due to many uncertain factors in actual system operation, t and the uncertainty are used to jointly define, i.e., the time interval T(t) of event occurrence time point is .

[0043] The binary time distance constraint is: the time point t i of event i and the time point tj The exact length between the time points t ij , i.e. The uncertainty of the time length is denoted as , then the time distance between the time points t i and t j is .

[0044] If event a can trigger event b, event a is called the cause event of event b, and event b is called the result event of event a.

[0045] Further, the forward and reverse temporal reasoning operations are defined as follows: Forward temporal reasoning: Given the time point constraint of the cause event a, and the binary time distance constraint between event a and the result event b, the forward reasoning obtains event b and its corresponding time interval, and the unary time point constraint of event b is expressed as: Reverse temporal reasoning: Given the time point constraint of the result event b, and the binary time distance constraint between event b and the cause event a, the reverse reasoning obtains event a and its corresponding time interval, and the unary time point constraint of event a is expressed as: For each suspicious fault element, the forward and reverse temporal reasoning analysis is performed on the corresponding switch quantity temporal information to obtain the unary time point constraint of each place in the Petri net general model, and then the operator is used to compare and check the actual obtained temporal information and the unary time point constraint of each place, so as to automatically filter the false fault alarm information. The operator is expressed as: Suppose that an event occurs, is the time point of the event, and T( ) is the unary time point constraint of the event, where α is the function expression of the temporal correlation characteristic in section 3.3.3. If the time point of the event satisfies the unary time point constraint T( ), the initial confidence of the event is assigned a value. If the time point of the event does not satisfy the unary time point constraint T( ), the initial confidence of the event is assigned a lower value.

[0046] ​It should be noted that the error alarm information can be identified by analyzing whether the fault alarm time sequence information meets the time sequence constraint condition through time sequence reasoning, thereby effectively improving the fault tolerance of fault diagnosis.

[0047] Based on the above principle, the specific operation steps are as follows: Determine the power failure area, and regard the elements in the power failure area as suspected fault elements; For each suspected fault element, the alarm information related to the element is divided to form a set of different elements; For each set, reverse time sequence reasoning is performed using the actually obtained alarm time sequence information to obtain a one-element time point constraint of fault occurrence; The one-element time point constraint of fault occurrence obtained by reverse time sequence reasoning of each set is merged to obtain a one-element time point total constraint T of fault occurrence; Forward time sequence reasoning is performed on T to obtain a one-element time point constraint of each library, and then The operator compares it with the actually obtained time sequence information to identify the error alarm information that does not meet the time sequence information.

[0048] In another possible implementation, when determining the power failure area and regarding the elements in the area as suspected fault elements, the power failure area can also be determined based on the power grid topology structure and real-time power flow data. First, a topology model of the power grid is constructed according to the topological connection relationship of the power grid. Then, in combination with the real-time power flow data, it is analyzed which lines have zero or lower than a set threshold power transmission, and the elements in the area connected by these lines are determined as the power failure area, and the elements in the area are regarded as suspected fault elements.

[0049] In another possible implementation, when determining the power failure area and regarding the elements in the area as suspected fault elements, the power failure area can also be determined using smart meter data and fault indicator information. The smart meter can monitor the power consumption of the user in real time, and when it is detected that a large number of users in a certain area are powered off, the area is marked. At the same time, the fault indicator can indicate whether a fault occurs on the line, and in combination with the position information of the fault indicator, the range of the power failure area is further narrowed. The elements in the finally determined power failure area are regarded as suspected fault elements.

[0050] In the embodiment, the matrix iterative operation on the corrected switch quantity information in step S4 to obtain the switch quantity fault probability includes: Determine the power failure area, and determine the suspected fault element set; For each suspected fault element, the alarm information is divided into an alarm information set corresponding to each element; A general model of TWFPN (Time-Weighted Fuzzy Petri Net) of the element is established, and the false alarm information not meeting the time sequence constraint relationship is removed through time sequence constraint checking; The initial confidence of the library is evaluated by fully mining the time sequence correlation characteristics. Through matrix operation and multiple iterations, the fault probability representation of the element, i.e. the switching value fault probability, is obtained.

[0051] In the embodiment, the fusion of the electrical quantity fault diagnosis result and the switching value fault probability in step S5 realizes the accurate discrimination of the fault, which includes: For the electrical quantity fault diagnosis result, for an identification framework containing n possible fault types The electrical quantity fault diagnosis result is represented as evidence Wherein is the number of electrical quantity diagnosis evidence.

[0052] For the switching value fault probability, for the same identification framework A series of evidence is given Wherein is the number of switching value diagnosis evidence.

[0053] All the evidence is integrated together, and the total number of evidence is The evidence set is Wherein, to is the electrical quantity evidence, to is the switching value evidence.

[0054] The average evidence is calculated and represented as: Wherein, represents the average evidence, , , … represents different evidence, each evidence may correspond to different types of fault diagnosis information, represents the total number of evidence.

[0055] The deviation of each evidence and the average evidence is calculated, and the deviation is measured by the improved Cityblock distance function; Specifically, for each evidence ( ), the Cityblock distance between the evidence and the average evidence is calculated as follows: Assuming that the focus element set is , then: Calculate the average of the deviation of each evidence from the average evidence : If the deviation of a certain evidence from the average evidence is greater than , then the evidence is considered abnormal evidence; Correct the data source: replace the abnormal evidence with the average evidence, and denote the corrected evidence set as ; Then, use the D-S combination formula to fuse the corrected data sources to obtain the final fusion result.

[0056] Specifically, for any two corrected evidences and , the D-S combination formula is: wherein denotes the basic probability value (i.e., the confidence) assigned to a specific fault proposition A (such as "a certain line is faulty") by the new and unified evidence body generated by D-S fusion of the i th and j th evidence sources, respectively, are the propositions (fault type combinations) supported by the two evidences, is the conflict coefficient, reflecting the conflict degree of the two evidences.

[0057] Since the D-S combination satisfies the associative law, multiple evidences are synthesized iteratively: wherein denotes the final evidence body obtained after fusion of all the evidences, i.e., the final fusion result.

[0058] In another possible implementation, when the electrical quantity fault diagnosis result and the switching quantity fault probability are fused, a weighted average fusion method can also be used, and different weights are assigned to the electrical quantity fault diagnosis result and the switching quantity fault probability. The determination of the weights can be based on the accuracy evaluation of historical data or expert experience. The electrical quantity fault diagnosis result is multiplied by its corresponding weight, the switching quantity fault probability is multiplied by its corresponding weight, and then the two are added to obtain the fused fault discrimination result.

[0059] In another possible implementation, when the electrical quantity fault diagnosis result and the switching quantity fault probability are fused, the Bayesian fusion method can also be used, and a Bayesian network model is established according to the electrical quantity fault diagnosis result and the switching quantity fault probability. By using the Bayesian formula to calculate the posterior probability of various fault types under the given electrical quantity and switching quantity information according to the known prior probability and conditional probability, the fault type with the maximum posterior probability is taken as the fused fault discrimination result.

[0060] In this embodiment, the technical scheme of the power grid fault identification system based on the fuzzy theory and the improved D-S evidence theory is not described in detail, and the description of the technical scheme of the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory can be referred to.

[0061] This embodiment also provides a power grid fault identification system based on the fuzzy theory and the improved D-S evidence theory, which comprises: A data acquisition module is configured to acquire power grid original data including fault voltage and current information and action state information of protection relays and circuit breakers. An electrical quantity diagnosis module is configured to perform noise reduction processing on the acquired fault voltage and current information, train a support vector machine classification model optimized by a fruit fly algorithm using historical fault sample data, and classify real-time fault electrical quantity features after noise reduction to obtain an electrical quantity fault diagnosis result. A switching quantity correction module is configured to filter and mine the action state information of protection relays and circuit breakers using a time sequence constraint relationship, automatically identify and correct false fault information in the switching quantity through reverse and forward time sequence reasoning. A probability calculation module is configured to perform matrix iteration operation on the corrected switching quantity information to obtain a switching quantity fault probability. A result fusion module is configured to fuse the electrical quantity fault diagnosis result and the switching quantity fault probability to realize accurate discrimination of faults.

[0062] This embodiment also provides an electronic device suitable for the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory, which comprises: The memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to implement the power grid fault identification method based on the fuzzy theory and the improved D-S evidence theory.

[0063] The embodiment further provides a storage medium, which stores a computer program, and the computer program is executed by a processor to implement the power grid fault identification method based on fuzzy theory and improved D-S evidence theory as proposed in the above embodiment.

[0064] The storage medium proposed in the embodiment belongs to the same inventive concept as the power grid fault identification method based on fuzzy theory and improved D-S evidence theory proposed in the above embodiment, and the technical details not described in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0065] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all should be covered in the scope of the claims of the present application.

Claims

1. A power grid fault identification method based on fuzzy theory and improved DS evidence theory, characterized in that, include: Acquire raw power grid data, including fault voltage and current information, as well as the operating status information of protective relays and circuit breakers; The acquired fault voltage and current information is denoised. A support vector machine classification model optimized by the fruit fly algorithm is trained using historical fault sample data. The real-time fault electrical quantity features after denoising are classified to obtain the electrical quantity fault diagnosis results. By utilizing timing constraints, the operating status information of protection relays and circuit breakers is filtered and mined. Through reverse and forward timing reasoning, erroneous fault information in switch quantities is automatically identified and corrected. Perform matrix iteration operations on the corrected switch quantity information to obtain the switch quantity failure probability; By integrating the fault diagnosis results of electrical quantities and the fault probability of switching quantities, accurate fault identification can be achieved.

2. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 1, characterized in that, The process involves denoising the acquired fault voltage and current information, training a support vector machine classification model optimized by the fruit fly algorithm using historical fault sample data, and classifying the denoised real-time fault electrical quantity features to obtain electrical quantity fault diagnosis results, including: The fault voltage and current information are denoised by continuously updating the modal components, center frequency and Lagrange multipliers. The iteration is stopped based on the preset convergence error to obtain the denoising result. A support vector machine classification model optimized by the fruit fly algorithm is trained using historical fault sample data to determine the initial position of the fruit fly population. After multiple iterations, the fruit fly population searches for the optimal position, and the parameter training results are output.

3. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 2, characterized in that, The process of denoising the acquired fault voltage and current information, training a support vector machine classification model optimized by the fruit fly algorithm using historical fault sample data, and classifying the denoised real-time fault electrical quantity features to obtain electrical quantity fault diagnosis results also includes: The trained support vector machine classification model is used to classify the noise-reduced real-time fault electrical quantity features. The merging deviation between the left and right parts of the fault data sequence at a certain point is calculated and minimized. Fault features are extracted, and the power grid faults are classified according to the fault features. The classification results are matched with the fault types to obtain the electrical quantity fault diagnosis results.

4. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 3, characterized in that, The process of filtering and mining the operational status information of protection relays and circuit breakers using timing constraints, and automatically identifying and correcting erroneous fault information in switching quantities through reverse and forward timing reasoning, includes: The power outage area is identified, and the components within the area are considered as suspected faulty components. For each suspected faulty component, relevant alarm information is divided into sets. Reverse time sequence reasoning is performed on each set using actual alarm time sequence information to obtain a univariate time point constraint for the occurrence of the fault.

5. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 4, characterized in that, The method of using timing constraints to filter and mine the operating status information of protection relays and circuit breakers, and automatically identifying and correcting erroneous fault information in switching quantities through reverse and forward timing reasoning, also includes: The constraints of each set are merged to obtain the unary time point constraint of the fault occurrence; forward temporal reasoning is performed on the total constraint to obtain the unary time point constraint of each storage location; through specific operators, the actual acquired time-series information is compared with the unary time point constraint of each storage location to identify erroneous alarm information that does not meet the time-series information.

6. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 5, characterized in that, The method of fusing electrical quantity fault diagnosis results and switch quantity fault probabilities to achieve accurate fault identification includes: For electrical quantity fault diagnosis results and switch quantity fault probabilities, under the same possible fault type identification framework, they are represented as electrical quantity diagnosis evidence and switch quantity diagnosis evidence, respectively, and the two types of evidence are integrated into a total evidence set; Calculate the average evidence for the total evidence set, then calculate the deviation of each piece of evidence from the average evidence, and use an improved distance function to measure the deviation.

7. The power grid fault identification method based on fuzzy theory and improved DS evidence theory as described in claim 6, characterized in that, The method of fusing electrical quantity fault diagnosis results and switch quantity fault probabilities to achieve accurate fault identification also includes: Calculate the average deviation of each piece of evidence from the average evidence, and consider evidence with a deviation greater than the average as abnormal evidence; The data source is corrected by replacing abnormal evidence with average evidence to obtain a corrected evidence set. The DS synthesis formula is then used to fuse information from the corrected data source.

8. A power grid fault identification system based on fuzzy theory and improved DS evidence theory, using the method described in any one of claims 1 to 7, characterized in that, include: The data acquisition module is used to acquire raw power grid data, including fault voltage and current information, as well as the operating status information of protective relays and circuit breakers. The electrical quantity diagnosis module is used to denoise the acquired fault voltage and current information, train a support vector machine classification model optimized by the fruit fly algorithm using historical fault sample data, and classify the real-time fault electrical quantity features after denoising to obtain the electrical quantity fault diagnosis results. The switch quantity correction module is used to filter and mine the operating status information of protection relays and circuit breakers by using timing constraints, and automatically identify and correct erroneous fault information in the switch quantities through reverse and forward timing reasoning. The probability calculation module is used to perform matrix iteration operations on the corrected switch quantity information to obtain the switch quantity failure probability. The result fusion module is used to fuse electrical quantity fault diagnosis results and switch quantity fault probabilities to achieve accurate fault identification.

9. An electronic device, characterized in that, include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions, which, when executed by the processor, implement the steps of the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores computer-executable instructions that, when executed by a processor, implement the steps of the method according to any one of claims 1 to 7.

Images