Converter station fault diagnosis method based on fault area identification and evidence information fusion

Abstract

The invention relates to the technical field of power system fault diagnosis, in particular to a converter station fault diagnosis method based on fault area identification and evidence information fusion, which comprises the following steps of: preprocessing sequence event record (SER) data of a converter station, constructing a characteristic event database, and calculating the confidence of each protection action so as to preliminarily define a fault range; the fuzzy Petri network model is utilized to process switching signal data in the debugging operation process of the converter station, and a MYCIN confidence reasoning method is combined to obtain switching evidence; a BP neural network model is constructed based on electrical data in the debugging operation process of the converter station, the electrical data are analyzed through deep learning, and electrical quantity evidence is extracted. And fusing the two types of evidences by adopting a D-S evidence theory to obtain an accurate fault diagnosis result. By comprehensively considering the two types of feature information, the fault point is accurately found out, the fault diagnosis precision is improved, rapid fault positioning is achieved, and a powerful guarantee is provided for safe operation of the converter station.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon fiber composite material preforming equipment, and in particular to a converter station fault diagnosis method based on fault area identification and evidence information fusion. Background Art

[0002] High-voltage direct current transmission system has become an important technology to solve the problem of long-distance and large-capacity power transmission due to its unique advantages such as flexible control, relatively independent transmission and receiving ends, low line loss, and easy grid interconnection. In recent years, with the vigorous development of new energy (such as solar energy, wind energy, etc.) and the change of energy distribution structure, high-voltage direct current transmission technology has been widely used in long-distance and large-capacity power transmission projects, and many high-voltage and ultra-high-voltage direct current transmission projects have been built and put into operation in my country. These projects have played an important role in promoting the development and utilization of new energy and promoting the implementation of "West-to-East Power Transmission".

[0003] DC transmission systems usually undertake tasks such as long-distance large-capacity power transmission or grid interconnection. The equipment in the DC converter station is complex and contains a large number of power electronic devices. Once a system failure occurs, it will seriously threaten the stable operation of the system and the safety of personnel and equipment. Therefore, a comprehensive and efficient DC converter station fault diagnosis method is urgently needed.

[0004] In the existing research on DC fault diagnosis, neural networks, expert systems, support vector machines and other methods are mainly used. However, most of these methods only analyze the single information source of electrical quantity recording in the fault recording, ignoring the DC transmission system sequential event record information (SER) and the switch quantity recording information reflecting the operation of the control protection device, which are also rich in fault information. Since different faults in the DC protection zone may have similar electrical quantity characteristics, relying solely on electrical quantity recording information can often only distinguish some faults with obvious differences in electrical quantity characteristics or identify the approximate area where the fault occurs, and cannot accurately locate the fault point in a certain DC protection zone, resulting in certain limitations of existing fault diagnosis methods when dealing with complex faults. Summary of the invention

[0005] The present invention provides a converter station fault diagnosis method based on fault area identification and evidence information fusion, which can effectively solve the problems in the background technology.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A converter station fault diagnosis method based on fault area identification and evidence information fusion comprises the following steps:

[0008] Building a characteristic event library through the preprocessed converter station SER data, and identifying the converter station fault area based on the confidence results of different faults in the SER data when the fault occurs;

[0009] By analyzing the fuzzy knowledge rule base of the switch quantity recording data components of different faults during the commissioning and operation of the converter station, a fuzzy Petri net for converter station switch quantity fault diagnosis is built based on the fuzzy knowledge rule base, and the fuzzy Petri net is used to iterate the switch quantity data of the converter station to obtain switch quantity fault diagnosis evidence information;

[0010] The wavelet energy spectra of different faults during the commissioning and operation of the converter station obtained by wavelet decomposition method are used to train the BP neural network. The electrical quantity recording data when diagnosing the fault after the training BP neural network is used to obtain the electrical quantity fault diagnosis evidence information.

[0011] The switching quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information are integrated by using the fusion rule of DS evidence theory to locate the fault point of the converter station.

[0012] Furthermore, the construction of a characteristic event library by using the pre-processed converter station SER data comprises the following steps:

[0013] Obtain the original SER data of the converter station, filter the core dimensions and exclude the interference dimensions;

[0014] Clean the retained SER data, remove duplicate records, fill in missing values, and unify the data format;

[0015] Count the distribution of system alarm information and event levels of different faults to form a preliminary set of characteristic events;

[0016] Associating the characteristic event set with the protection name tag to construct a characteristic event-protection name mapping table;

[0017] The data information of the characteristic event-protection name mapping table is integrated to form the characteristic event library.

[0018] Furthermore, the identifying of the converter station fault area by analyzing the confidence results of different faults in the SER data at the time of the fault based on the characteristic event library comprises the following steps:

[0019] Obtaining SER data at the time of failure, traversing the pre-processed SER on-site event list and matching it with the rules in the characteristic event library;

[0020] The support for the protection action in the converter area under different faults is counted, the maximum support count is taken as the reference value, and the support count is normalized to obtain the confidence of the protection action, which is specifically obtained through the following formula:

[0021]

[0022] Among them, Conf i Indicates the confidence of the protection action, Support i represents the support count of the protection action, α represents the scaling factor, which is 0.95;

[0023] Identify protection actions with high confidence and determine the fault area of ​​the converter station.

[0024] Furthermore, the fuzzy knowledge rule base is constructed by analyzing the switch quantity recording data of different faults during the commissioning operation of the converter station, including the following steps:

[0025] Extracting protection action conditions of different faults based on the switch quantity recording data of different faults during the commissioning operation of the converter station;

[0026] Based on the extracted protection action conditions, the association between different faults and protection actions is analyzed, and fuzzy knowledge rules are established;

[0027] The established fuzzy knowledge rules are organized into a library to obtain the fuzzy knowledge rule library.

[0028] Furthermore, the use of the fuzzy Petri net to iterate the inference of the converter station switch quantity data to obtain switch quantity fault diagnosis evidence information includes the following steps:

[0029] According to the fuzzy knowledge rule base and the established fuzzy knowledge rules, the fuzzy Petri net for converter station switch fault diagnosis is constructed;

[0030] Obtaining fault symptom information of fault site switch quantity signals of different faults and inputting the information into the fuzzy Petri net;

[0031] By using the MYCIN confidence method, reasoning iteration is performed in the fuzzy Petri net to obtain the switch fault diagnosis evidence information.

[0032] Furthermore, the wavelet energy spectra of different faults during the commissioning operation of the converter station obtained by the wavelet decomposition method are used to train the BP neural network, comprising the following steps:

[0033] The db8 wavelet is used to perform wavelet 8-layer decomposition on the electrical quantity recording data of different faults during the commissioning and operation of the converter station, and the wavelet coefficients at different scales are obtained.

[0034] The wavelet energy at each scale is calculated according to the Parseval principle, and the wavelet energy spectrum data set of different faults is obtained;

[0035]

[0036] Among them, E aJ is the low-frequency energy of the Jth scale, E dJ is the high-frequency energy of the jth scale;

[0037] According to the characteristics of the wavelet energy spectrum data set, a BP neural network is constructed to determine the specific number of input layers, hidden layers, and output layers. The sigmoid function is used as the activation function of the hidden layer, and the linear function is used as the activation function of the output layer.

[0038] The wavelet energy spectra of different faults are used as input features for BP neural network training, and the BP neural network is trained to obtain the trained BP neural network.

[0039] Furthermore, the integration of the switch quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information using the fusion rule of DS evidence theory includes the following steps:

[0040] Proposition A is defined to represent the possible fault points or fault types of the converter station, and the mass function satisfies the following conditions:

[0041]

[0042] in, is a proposition, m is the quality function of the proposition, and m(A) represents the quality of proposition A, that is, its importance;

[0043] The belief function Bel(A) and the likelihood function Pl(A) are defined to represent the maximum and minimum probabilities of the occurrence of proposition A, respectively. The specific expressions are as follows:

[0044]

[0045] Define two independent pieces of evidence information whose basic probabilities are assigned as m1 and m2. Use the fusion rule of DS evidence theory to obtain a new probability distribution, which indicates the support degree of the fused evidence for proposition A, that is, the probability of the occurrence of a fault point or fault type:

[0046]

[0047] Among them, m(A) is defined as the orthogonal sum of m1 and m2, which represents the probability of proposition A occurring after the fusion of evidence information.

[0048] A converter station fault diagnosis device based on fault area identification and evidence information fusion, the device comprising:

[0049] The fault area identification module is used to monitor the operating status data of key components of the converter station in real time, match and analyze the preset fault feature model with the real-time data, and preliminarily divide and identify potential fault areas;

[0050] The evidence information collection module is used to comprehensively collect the operation evidence information of each area in the converter station and convert this information into digital signals for subsequent processing;

[0051] The evidence information fusion module uses data fusion technology to integrate evidence information to form a comprehensive evidence set;

[0052] The fault diagnosis reasoning module uses logical reasoning and pattern recognition methods to accurately judge the fault type, cause and severity based on the fused evidence information, combined with the expert knowledge base and the fault case library, and generates a fault diagnosis report.

[0053] The early warning and decision support module triggers the early warning mechanism according to the fault diagnosis results, sends fault alarms to operation and maintenance personnel in a timely manner, provides fault handling suggestions or automatically executes preset emergency response strategies;

[0054] The human-computer interaction interface provides an operation interface and information display, which facilitates viewing of real-time monitoring data, fault diagnosis results and historical records, and supports manual intervention and parameter adjustment.

[0055] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion are implemented.

[0056] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion.

[0057] The beneficial effects of the present invention are:

[0058] 1. The present invention proposes a converter station fault diagnosis method based on fault area identification and evidence information fusion, which integrates the existing technologies and improves them on this basis to assist in realizing intelligent and accurate diagnosis of converter station faults.

[0059] 2. The present invention proposes a method for iteratively reasoning about the switch quantity data of the converter station based on fuzzy Petri nets to obtain evidence information, uses a graphical method to intuitively describe the changes in the system switch quantity when the converter station fails, and uses the established Petri net to iterate the sign information at the fault site to obtain the corresponding evidence information source.

[0060] 3. The present invention proposes to train the energy of the converter station electrical quantity after wavelet decomposition based on BP neural network to obtain the evidence information of the electrical quantity. It creatively proposes to use the energy of the recorded wave at various scales after wavelet transformation to extract features and determine the fault type through the wavelet energy spectrum characteristics.

[0061] 4. The present invention proposes to apply the DS evidence fusion method to fuse the switch quantity evidence information and electrical quantity evidence information of the converter station, thereby greatly improving the completeness of the information and the reliability and accuracy of the diagnosis results. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0063] Figure 1 A multi-information fusion diagnosis flow chart of a converter station layered by layer in a converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0064] Figure 2 Partial SER events of converter stations in the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0065] Figure 3 It is a converter station partial SER feature event library of the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0066] Figure 4 It is the on-site SER event analysis result of the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0067] Figure 5 The Petri net construction process of the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0068] Figure 6 A fuzzy Petri net model for converter station switch quantity diagnosis of a converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0069] Figure 7 It is a wavelet energy spectrum distribution diagram of the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention;

[0070] Figure 8 This is a BP neural network structure diagram of the converter station fault diagnosis method based on fault area identification and evidence information fusion of the present invention. DETAILED DESCRIPTION

[0071] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0072] It should be noted that when an element is referred to as being "fixed to" another element, it may be directly on the other element or there may be a central element. When an element is considered to be "connected to" another element, it may be directly connected to the other element or there may be a central element at the same time. The terms "vertical", "horizontal", "left", "right" and similar expressions used herein are for illustrative purposes only and do not represent the only implementation method.

[0073] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0074] Among the existing DC fault diagnosis methods, most of them use neural networks, expert systems, support vector machines and other methods, which can only help identify and solve some fault problems to a certain extent; their diagnostic principles mostly only analyze the single information source of electrical quantity recording in the fault recording, while ignoring the DC transmission system sequential event record information (SER) and the switch quantity recording information that reflects the action of the control and protection device, which are also rich in fault information.

[0075] Since different faults within a DC protection zone may have similar electrical quantity characteristics, relying solely on electrical quantity recording information can often only distinguish some faults with obvious differences in electrical quantity characteristics or identify the approximate area where the fault occurs, but cannot accurately locate the fault point within a certain DC protection zone; this leads to certain limitations in existing fault diagnosis methods when dealing with complex faults.

[0076] In view of the limitations of existing fault diagnosis methods and the close relationship between protection partitions and fault diagnosis, the present invention proposes a converter station fault diagnosis strategy that integrates fault area identification and evidence information fusion. A two-step strategy is adopted to diagnose converter station faults: first, the switch signal and electrical data are processed separately, and the respective fault characteristics are extracted using fuzzy Petri nets and BP neural networks; secondly, advanced evidence fusion technology is used to comprehensively consider the two types of feature information, so as to accurately find the fault point.

[0077] See Figure 1As shown, the present invention discloses a converter station fault diagnosis method based on fault area identification and evidence information fusion, comprising the following steps:

[0078] Firstly, a characteristic event library is constructed through the preprocessed converter station SER data. Based on the characteristic event library, the confidence results of different faults in the SER data at the time of fault are analyzed to identify the fault area of ​​the converter station. The SER data of the converter station is subjected to dimensionality reduction processing, characteristic events are counted, and a characteristic event library is established to provide a basis for preliminary fault section positioning.

[0079] Secondly, by analyzing the fuzzy knowledge rule base of the switch quantity recording data components of different faults during the commissioning and operation of the converter station, a fuzzy Petri net for converter station switch quantity fault diagnosis is built based on the fuzzy knowledge rule base, and the fuzzy Petri net is used to iterate the inference of the converter station switch quantity data to obtain the switch quantity fault diagnosis evidence information;

[0080] At the same time, the wavelet energy spectra of different faults during the commissioning and operation of the converter station obtained by wavelet decomposition method are used to train the BP neural network. The electrical quantity recording data when diagnosing the fault after the training BP neural network is used to obtain the electrical quantity fault diagnosis evidence information.

[0081] Finally, for the two types of evidence information, switch quantity and electrical quantity, the fusion rules of DS evidence theory are used to integrate the switch quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information to improve the diagnosis accuracy; through evidence fusion, the fault point of the converter station is located to improve the accuracy and reliability of fault diagnosis.

[0082] Specifically, the present invention proposes a converter station fault diagnosis strategy that integrates fault area identification and evidence information fusion. First, the converter station sequential event record (SER) data is preprocessed to build a characteristic event database, and the confidence of each protection action is calculated to preliminarily define the fault range; then, the fuzzy Petri net model is used to process the switch signal data, and the MYCIN confidence reasoning method is combined to obtain switch evidence; at the same time, a BP neural network model is constructed to analyze electrical data through deep learning and extract electrical quantity evidence. Finally, the DS evidence theory is used to fuse the two types of evidence to obtain accurate fault diagnosis results.

[0083] Example 1

[0084] This embodiment realizes efficient use of converter station SER data through preprocessing and construction of a characteristic event library, thereby quickly locating the fault section. Specifically, the construction of a characteristic event library through preprocessed converter station SER data includes the following steps:

[0085] The original SER data of the converter station is obtained, the core dimensions are screened and the interference dimensions are excluded; the retained SER data is cleaned, duplicate records are removed, missing values ​​are filled (for example, using the previous and next values ​​or default values), and the data format is unified; the distribution of system alarm information and event levels of different faults is counted to form a preliminary set of characteristic events; the characteristic event set is associated with the protection name label, and a characteristic event-protection name mapping table is constructed to ensure that each characteristic event can correspond to a specific protection action; the data information of the characteristic event-protection name mapping table is integrated to form the characteristic event library, and each record in the library contains a characteristic event description, a corresponding protection name and a fault type label.

[0086] Specifically, export the sequential event record (SER) data from the converter station monitoring system to ensure that the data contains key feature dimensions such as time, host, system alarm, event level, alarm group, event list, etc.; analyze the importance of each dimension of data to fault diagnosis, and exclude interference dimensions, such as event level (if it has little impact on fault location) or alarm group (if its classification is too general), to reduce the amount of data and improve processing efficiency. Data cleaning includes removing irrelevant log information, such as daily inspection records, retaining key events related to the fault, and ensuring that the timestamps of all events are unified and in the correct format for subsequent analysis;

[0087] By collecting and counting the SER files corresponding to various types of faults, we can extract fault characteristic events and their corresponding protection name tags, associate the characteristic events obtained with the corresponding fault types, protection actions, etc., and build a characteristic event library of converter station fault signals. For example, we can count the combination of specific alarm information, and a specific combination of system alarms may indicate a certain type of converter fault.

[0088] Further, in this embodiment, the identifying of the converter station fault area by analyzing the confidence results of different faults in the SER data at the time of the fault based on the characteristic event library includes the following steps:

[0089] Obtain the SER data at the time of the fault, traverse the pre-processed SER field event list and match it with the rules in the characteristic event library; count the support of different faults for the protection action in the converter area, take the maximum support count as a reference value, normalize the support count to obtain the protection action confidence, identify the protection action with high confidence, and determine the fault area of ​​the converter station. Specifically, the following formula is used:

[0090]

[0091] Among them, Conf i Indicates the confidence of the protection action, Support i represents the support count of the protection action, and α represents the scaling factor, which is 0.95.

[0092] Specifically, for actual faults, first extract the field event list from the SER system and organize it in the preprocessed format; secondly, match the organized field event list with the records in the feature event library, and count the number of occurrences of each feature event, that is, the support count; according to the support count, use the normalization method to calculate the confidence of each protection action to quantify the strength of the association between the protection action and the fault. Finally, the initial identification of the fault area is carried out by analyzing the confidence calculated to identify the protection actions with higher confidence. These actions are usually closely related to the fault area. Further combined with the topological structure and protection configuration of the converter station, the confidence results are used to narrow the scope of fault investigation and preliminarily determine the fault area.

[0093] In summary, this implementation scheme uses text learning and statistical methods to extract characteristic events related to faults from a large amount of SER data, and quantifies the degree of correlation between protection actions and faults through confidence calculation, thereby quickly locating the fault area; thus improving the accuracy and efficiency of converter station fault diagnosis.

[0094] Example 2

[0095] This embodiment realizes efficient diagnosis of converter station switch faults by systematically analyzing protection action conditions, establishing a fuzzy knowledge rule base, building a fuzzy Petri net, and performing reasoning iteration. Specifically, by analyzing the switch recording data of different faults during the commissioning operation of the converter station to form a fuzzy knowledge rule base, the following steps are included:

[0096] Based on the switch quantity recording data of different faults during the commissioning operation of the converter station, the protection action conditions of different faults are extracted; based on the extracted protection action conditions, the relationship between different faults and protection actions is analyzed to establish fuzzy knowledge rules; the established fuzzy knowledge rules are organized into a library to obtain the fuzzy knowledge rule library.

[0097] The switch quantity recording data during the commissioning and operation of the converter station records the protection actions under different fault conditions, which is the basis for building a fuzzy knowledge rule base and performing fault diagnosis. Specifically, the protection actions under each fault are extracted from the switch quantity recording data during the commissioning and operation of the converter station, including the time, type, and status of related switches of the protection actions, such as valve short-circuit protection actions and converter differential protection actions. By analyzing the correlation between different faults and protection actions, preliminary fuzzy knowledge rules are formed. These rules describe which protection actions may occur under different fault conditions and the logical relationship between them. The formed fuzzy knowledge rules are organized into a library, namely the fuzzy knowledge rule base. This library will serve as the basis for the subsequent construction of fuzzy Petri nets and reasoning iterations.

[0098] Furthermore, in this embodiment, the use of the fuzzy Petri net to iterate the inference of the converter station switch quantity data to obtain switch quantity fault diagnosis evidence information includes the following steps:

[0099] According to the fuzzy knowledge rule base and the established fuzzy knowledge rules, the fuzzy Petri net for converter station switch fault diagnosis is constructed; the fault symptom information of the switch signal at the fault site of different faults is obtained and input into the fuzzy Petri net; through the MYCIN confidence method, reasoning iteration is performed in the fuzzy Petri net to obtain the switch fault diagnosis evidence information.

[0100] In the fuzzy Petri net, faults, protection actions, and switch states are taken as nodes, and the logical relationships between them are taken as edges to form a complete fault diagnosis model. Fault symptom information is obtained from the switch quantity signals at the fault site, which may include the sequence of protection actions, time intervals, state changes of related switches, etc. The obtained fault symptom information is input into the constructed fuzzy Petri net, and the MYCIN confidence method is used for reasoning iterations, which can gradually approach the real fault state and obtain evidence information for switch quantity fault diagnosis, which includes the possible type of fault, the probability of fault occurrence, the credibility of related protection actions, etc.

[0101] Example 3

[0102] This embodiment realizes efficient diagnosis of electrical quantity faults in converter stations through wavelet decomposition, BP neural network training and fault diagnosis, wherein wavelet decomposition is used to extract characteristic information in electrical quantity signals, and these characteristics are learned and classified by BP neural network, so as to identify the fault type. Specifically, the wavelet energy spectrum of different faults during the commissioning operation of the converter station obtained by wavelet decomposition method is used to train the BP neural network, including the following steps:

[0103] The db8 wavelet is used to perform wavelet 8-layer decomposition on the electrical quantity recording data of different faults during the commissioning and operation of the converter station, and the wavelet coefficients at different scales are obtained; the wavelet energy at each scale is calculated according to the Parseval principle, and the wavelet energy spectrum data set of different faults is obtained;

[0104]

[0105] Among them, E aJ is the low-frequency energy of the Jth scale, E dJ is the high-frequency energy of the jth scale.

[0106] First, the recorded data is decomposed into wavelet coefficients at different scales. Then, for each scale, the low-frequency energy EaJ and high-frequency energy EdJ are calculated to form a wavelet energy spectrum. These energy spectra reflect the energy distribution of the recorded data at different scales. During the commissioning of the converter station, the electrical quantity under different fault conditions is recorded, and the above steps are repeated to collect the wavelet energy spectra under different faults. These energy spectra are used as input features for BP neural network training.

[0107] In the specific implementation process, according to the characteristics of the wavelet energy spectrum data set, a BP neural network is constructed to determine the specific number of input layers, hidden layers, and output layers. The number of input layer nodes corresponds to the number of wavelet energy spectrum features, and the number of output layer nodes corresponds to the number of fault types. The sigmoid function is used as the excitation function of the hidden layer, and the linear function is used as the excitation function of the output layer; the wavelet energy spectrum of different faults is used as the input feature of the BP neural network training, and the BP neural network is trained to obtain the trained BP neural network. During the training process, the network weights and biases are adjusted so that the network can accurately output the fault type.

[0108] In this embodiment, the electrical quantity recording data when diagnosing the fault is based on the BP neural network, and then the electrical quantity fault diagnosis evidence information is obtained; in the specific fault diagnosis and evidence information acquisition process, the fault symptom information is first obtained from the electrical quantity signal at the fault site, which may include abnormal changes in electrical quantities such as voltage and current. Secondly, the acquired fault symptom information is subjected to wavelet decomposition, and the wavelet energy spectrum is calculated. The calculated wavelet energy spectrum is input into the trained BP neural network for fault diagnosis. The network output is the evidence information for electrical quantity fault diagnosis, including the fault type and its probability.

[0109] Example 4

[0110] This embodiment uses the DS evidence fusion method to effectively fuse the switch quantity evidence information and the electrical quantity evidence information of the converter station, thereby improving the accuracy and reliability of fault diagnosis. Specifically, the fusion rules of the DS evidence theory are used to integrate the switch quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information, including the following steps:

[0111] Proposition A is defined to represent the possible fault points or fault types of the converter station, and the mass function satisfies the following conditions:

[0112]

[0113] in, is a proposition, m is the quality function of proposition A, which indicates the importance or credibility of proposition A, based on the analysis results of switch quantity evidence information and electrical quantity evidence information, and m(A) indicates the quality of proposition A, i.e., its importance;

[0114] The trust function Bel(A) and the likelihood function Pl(A) are defined to represent the maximum and minimum probabilities (i.e., pessimistic and optimistic estimates) of the occurrence of proposition A, respectively. The specific expressions are as follows:

[0115]

[0116] Define two independent pieces of evidence information and assign them the basic probabilities m1 and m2. Use the fusion rule of DS evidence theory to obtain a new probability distribution. This new probability comprehensively considers the evidence information of both switch quantity and electrical quantity, and indicates the support degree of the fused evidence for proposition A, that is, the probability of the occurrence of a fault point or fault type:

[0117]

[0118] In the above formula, m(A) is defined as the orthogonal sum of m1 and m2, which represents the probability of proposition A occurring after the fusion of evidence information. In the specific implementation process, whether proposition A occurs is judged based on the probability distribution after fusion. If the probability distribution after fusion exceeds a certain set threshold, proposition A is considered to be true, that is, this type of fault occurs.

[0119] Example 5

[0120] In the specific implementation process, the application of text learning to construct the SER feature event library of converter station fault signals includes the following steps:

[0121] ① The original SER data of the converter station is preprocessed in six feature dimensions, including time, host, system alarm, event level, alarm group, and event list, to eliminate interference dimensions and greatly reduce the data volume. Some SER events after preprocessing are as follows: Figure 2 shown.

[0122] ② The SER files corresponding to various faults during the annual adjustment of a DC project converter station were counted, with a total of 4464 event information. After preprocessing, 590 SER feature events were obtained. The feature events and protection name tags were counted and classified to obtain the SER feature event library. The corresponding relationships of some feature events are as follows: Figure 3 shown.

[0123] ③ Taking the actual fault of a DC converter station as an example, the pre-processed SER field event list is traversed and matched one by one with the rules in the SER feature library. The support of each event for the protection action in the converter area is counted, and the maximum support count is taken as the reference value. The support count is normalized to obtain the protection action confidence, such as Figure 4 shown.

[0124] ④ Analyze the confidence results to realize the mining of characteristic event groups of converter stations and preliminary identification of fault areas, thus narrowing the interval of pending faults.

[0125] In the specific implementation process, the application of fuzzy Petri net to iteratively reason about the converter station switch quantity data to obtain evidence information includes the following steps:

[0126] ① Analyze the protection action conditions under various faults during the commissioning and operation of the DC converter station, that is, the switch quantity recording data, and use this to establish its fuzzy knowledge rule base. The specific Petri net construction process is as follows: Figure 5 As shown, the place represents the protection action signal or fault status, the transition represents the rule triggering condition, and the arc represents the signal transmission direction.

[0127] ② Establish fuzzy knowledge and form rules based on the protection action signals under different fault conditions during the commissioning and operation of the converter station. The following initial library is selected first: P1: valve short-circuit protection action, P2: converter differential protection action, P3: commutation failure occurs, P4: commutation failure protection action, P5: valve overcurrent protection action, P6: F4, P7: F12, P8: F3, P9: F1. According to the main protection action corresponding to each fault point, the following fuzzy knowledge can be obtained:

[0128] S1:IF P2 and P3 THEN P6,CF=0.9;

[0129] S2:IF P2 and P3 THEN P7,CF=0.9;

[0130] S3:IF P1 and P3 and P5 THEN P8,CF=0.9;

[0131] S4:IF P3 and P4 THEN P9,CF=0.8.

[0132] ③Build the fuzzy Petri net for fault diagnosis of converter station switch quantity, as follows Figure 6 shown.

[0133] ④ Then, the fault symptom information is obtained according to the switch signal at the fault site, and the MYCIN confidence method is used for reasoning iteration to obtain the switch fault diagnosis evidence information, as shown in Table 1.

[0134] Table 1. Diagnosis results of converter station switch quantity evidence information

[0135]

[0136] In the specific implementation process, the BP neural network is used to train the energy of the converter station electrical quantity after wavelet decomposition, and the evidence information is obtained in the following steps:

[0137] ① Use db8 wavelet to perform 8-layer wavelet decomposition on the converter station recording to obtain wavelet coefficients at different scales. According to the Parseval principle, the wavelet energy E of the signal is f The relationship between the wavelet coefficients at each scale is used to calculate the wavelet energy spectrum at each scale, such as Figure 7 shown.

[0138] ② According to the electrical quantity recording under different fault conditions during the commissioning and operation of the converter station, the wavelet energy spectra under different faults are collected.

[0139] ③ Construct a BP neural network and determine the specific number of input layers, hidden layers, and output layers. The hidden layer of the network uses the sigmoid function as the excitation function, and the output layer uses a linear function, such as Figure 8 shown.

[0140] ④Finally, the fault symptom information is obtained according to the electrical quantity signal at the fault site and diagnosed through the trained BP neural network to obtain the electrical quantity fault diagnosis evidence information, as shown in Table 2.

[0141] Table 2. Diagnosis results of BP neural network evidence information of converter station electrical quantity

[0142]

[0143] In the specific implementation process, the DS evidence fusion method is applied to fuse the switch quantity evidence information and electrical quantity evidence information of the converter station, and then the accurate identification of the fault point of the converter station is achieved based on the fusion results. The fusion results are shown in Table 3.

[0144] Table 3. Converter station switch quantity Petri and electrical quantity BP neural network evidence information fusion and diagnosis results

[0145]

[0146] The present embodiment provides a converter station fault diagnosis method based on fault area identification and evidence information fusion, which adopts a method of obtaining evidence information by iteratively reasoning about the converter station switch quantity data based on fuzzy Petri nets, and uses a graphical method to intuitively describe the changes in the system switch quantity when the converter station fails, and iterates the symptom information at the fault site based on the established Petri net to obtain the corresponding evidence information source. The present invention uses a BP neural network to train the energy after wavelet decomposition of the converter station electrical quantity to obtain evidence information. It creatively proposes to use the energy value spectrum of the electrical quantity recording at various scales after wavelet transformation to extract features, so as to accurately determine the fault type. The present invention adopts a DS evidence fusion method to fuse the converter station switch quantity evidence information and the electrical quantity evidence information, which greatly improves the completeness of the information and the reliability and accuracy of the diagnosis results.

[0147] In the present invention, a converter station fault diagnosis device based on fault area identification and evidence information fusion is provided, and the device comprises:

[0148] The fault area identification module is used to monitor the operating status data of key components of the converter station in real time, match and analyze the preset fault feature model with the real-time data, and preliminarily divide and identify potential fault areas;

[0149] The evidence information collection module is used to comprehensively collect the operation evidence information of each area in the converter station and convert this information into digital signals for subsequent processing;

[0150] The evidence information fusion module uses data fusion technology to integrate evidence information to form a comprehensive evidence set;

[0151] The fault diagnosis reasoning module uses logical reasoning and pattern recognition methods to accurately judge the fault type, cause and severity based on the fused evidence information, combined with the expert knowledge base and the fault case library, and generates a fault diagnosis report.

[0152] The early warning and decision support module triggers the early warning mechanism according to the fault diagnosis results, sends fault alarms to operation and maintenance personnel in a timely manner, provides fault handling suggestions or automatically executes preset emergency response strategies;

[0153] The human-computer interaction interface provides an operation interface and information display, which facilitates viewing of real-time monitoring data, fault diagnosis results and historical records, and supports manual intervention and parameter adjustment.

[0154] In the present invention, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion are implemented.

[0155] In the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion are implemented.

[0156] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0157] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0158] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0159] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A converter station fault diagnosis method based on fault area identification and evidence information fusion, characterized in that: The following steps are involved: Building a characteristic event library through the preprocessed converter station SER data, and identifying the converter station fault area based on the confidence results of different faults in the SER data when the fault occurs; By analyzing the fuzzy knowledge rule base of the switch quantity recording data components of different faults during the commissioning and operation of the converter station, a fuzzy Petri net for converter station switch quantity fault diagnosis is built based on the fuzzy knowledge rule base, and the fuzzy Petri net is used to iterate the switch quantity data of the converter station to obtain switch quantity fault diagnosis evidence information; The wavelet energy spectra of different faults during the commissioning and operation of the converter station obtained by wavelet decomposition method are used to train the BP neural network. The electrical quantity recording data when diagnosing the fault after the training BP neural network is used to obtain the electrical quantity fault diagnosis evidence information. The switching quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information are integrated by using the fusion rule of DS evidence theory to locate the fault point of the converter station.

2. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 1 is characterized in that: The method of constructing a characteristic event library through the pre-processed converter station SER data comprises the following steps: Obtain the original SER data of the converter station, filter the core dimensions and exclude the interference dimensions; Clean the retained SER data, remove duplicate records, fill in missing values, and unify the data format; Count the distribution of system alarm information and event levels of different faults to form a preliminary set of characteristic events; Associating the characteristic event set with the protection name tag to construct a characteristic event-protection name mapping table; The data information of the characteristic event-protection name mapping table is integrated to form the characteristic event library.

3. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 1 is characterized in that: The method of analyzing the confidence results of different faults in the SER data at the time of the fault based on the characteristic event library to identify the fault area of ​​the converter station includes the following steps: Obtaining SER data at the time of failure, traversing the pre-processed SER on-site event list and matching it with the rules in the characteristic event library; The support for the protection action in the converter area under different faults is counted, the maximum support count is taken as the reference value, and the support count is normalized to obtain the confidence of the protection action, which is specifically obtained through the following formula: Among them, Conf i Indicates the confidence of the protection action, Support i represents the support count of the protection action, α represents the scaling factor, which is 0.95; Identify protection actions with high confidence and determine the fault area of ​​the converter station.

4. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 1 is characterized in that: The fuzzy knowledge rule base is constructed by analyzing the switch quantity recording data of different faults during the commissioning operation of the converter station, including the following steps: Extracting protection action conditions of different faults based on the switch quantity recording data of different faults during the commissioning operation of the converter station; Based on the extracted protection action conditions, the association between different faults and protection actions is analyzed, and fuzzy knowledge rules are established; The established fuzzy knowledge rules are organized into a library to obtain the fuzzy knowledge rule library.

5. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 4 is characterized in that: The method of using the fuzzy Petri net to iterate the converter station switch quantity data to obtain switch quantity fault diagnosis evidence information includes the following steps: According to the fuzzy knowledge rule base and the established fuzzy knowledge rules, the fuzzy Petri net for converter station switch fault diagnosis is constructed; Obtaining fault symptom information of fault site switch quantity signals of different faults and inputting the information into the fuzzy Petri net; By using the MYCIN confidence method, reasoning iteration is performed in the fuzzy Petri net to obtain the switch fault diagnosis evidence information.

6. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 1 is characterized in that: The wavelet energy spectra of different faults during the commissioning and operation of the converter station obtained by the wavelet decomposition method are used to train the BP neural network, including the following steps: The db8 wavelet is used to perform wavelet 8-layer decomposition on the electrical quantity recording data of different faults during the commissioning and operation of the converter station, and the wavelet coefficients at different scales are obtained. The wavelet energy at each scale is calculated according to the Parseval principle, and the wavelet energy spectrum data set of different faults is obtained; According to the characteristics of the wavelet energy spectrum data set, a BP neural network is constructed to determine the specific number of input layers, hidden layers, and output layers. The sigmoid function is used as the activation function of the hidden layer, and the linear function is used as the activation function of the output layer. The wavelet energy spectra of different faults are used as input features for BP neural network training, and the BP neural network is trained to obtain the trained BP neural network.

7. The converter station fault diagnosis method based on fault area identification and evidence information fusion according to claim 1 is characterized in that: The method of integrating the switch quantity fault diagnosis evidence information and the electrical quantity fault diagnosis evidence information by using the fusion rule of DS evidence theory includes the following steps: Proposition A is defined to represent the possible fault points or fault types of the converter station, and the mass function satisfies the following conditions: in, is a proposition, m is the quality function of the proposition, and m(A) represents the quality of proposition A, that is, its importance; The belief function Bel(A) and the likelihood function Pl(A) are defined to represent the maximum and minimum probabilities of the occurrence of proposition A, respectively. The specific expressions are as follows: Define two independent pieces of evidence information and assign the basic probabilities to m1 and m2. Use the fusion rule of DS evidence theory to obtain a new probability distribution, which indicates the support degree of the fused evidence for proposition A, that is, the probability of the occurrence of a fault point or fault type: Among them, m(A) is defined as the orthogonal sum of m1 and m2, which represents the probability of proposition A occurring after the fusion of evidence information.

8. A converter station fault diagnosis device based on fault area identification and evidence information fusion, characterized in that: The device comprises: The fault area identification module is used to monitor the operating status data of key components of the converter station in real time, match and analyze the preset fault feature model with the real-time data, and preliminarily divide and identify potential fault areas; The evidence information collection module is used to comprehensively collect the operation evidence information of each area in the converter station and convert this information into digital signals for subsequent processing; The evidence information fusion module uses data fusion technology to integrate evidence information to form a comprehensive evidence set; The fault diagnosis reasoning module uses logical reasoning and pattern recognition methods to accurately judge the fault type, cause and severity based on the fused evidence information, combined with the expert knowledge base and fault case library, and generates a fault diagnosis report. The early warning and decision support module triggers the early warning mechanism according to the fault diagnosis results, sends fault alarms to operation and maintenance personnel in a timely manner, provides fault handling suggestions or automatically executes preset emergency response strategies; The human-computer interaction interface provides an operation interface and information display, which facilitates viewing of real-time monitoring data, fault diagnosis results and historical records, and supports manual intervention and parameter adjustment.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the converter station fault diagnosis method based on fault area identification and evidence information fusion according to any one of claims 1 to 7 are implemented.

Images