Method and system for defect data mining and feature extraction of direct current protection device

Abstract

The invention discloses a method and a system for defect data mining and feature extraction of a direct current protection device, and belongs to the technical field of direct current protection devices. The method comprises the following steps: on the basis of a regular expression, preprocessing defect text data to extract key words and useless words in the defect text data; based on a word segmentation method of a prefix dictionary, carrying out secondary text word segmentation processing on the preprocessed defect text data to obtain a word segmentation lexicon; calculating the weight of the defect text data corresponding to each segmented word according to the word segmentation result of the segmented word lexicon, taking the defect text data with the weight higher than the lowest threshold as feature data, and labeling the feature data; and based on the TF-IDF-CNN model, according to the defect text data of the target DC protection device, extracting defect feature data of the target DC protection device. According to the method, the features of the defect text data of the direct current protection device can be efficiently extracted.

Description

Technical Field

[0001] The present invention relates to the technical field of direct current protection devices, and more specifically, to a method and system for direct current protection device defect data mining and feature extraction. Background Art

[0002] The distribution of electric energy and electric load in my country is extremely unbalanced, and it is necessary to build long-distance and large-capacity DC power transmission projects. With the continuous acceleration of DC project construction, the equipment scale of DC protection systems continues to expand. As the core part of DC power transmission, the DC protection system has become an important factor affecting the safe and stable operation of the power grid. At present, it is difficult to accurately and quickly determine the cause and location of the defect and other defect information after a defect occurs in the DC protection equipment. It is heavily dependent on the level of manual experience and the technical service support of the equipment manufacturer, which makes it difficult to discover and quickly deal with the defects of the DC protection equipment in the first time, which brings hidden dangers to the safe and stable operation of the power grid. The DC protection system has a complex structure, involves many links, and has complex correlation and coupling of the health status of the equipment. At present, there is a lack of technical means specifically for the analysis and evaluation of the health status of DC protection equipment. Daily operation and maintenance management is difficult, and it is difficult to discover and deal with equipment defects in time and evaluate the comprehensive operation status of the equipment.

[0003] With the continuous construction and operation of DC transmission projects, a certain amount of defect data related to DC protection devices has been accumulated, including defect details, defect handling status, faulty equipment type, occurrence time, faulty equipment model, etc. The diagnosis and treatment of defects in DC protection system equipment often rely on expert experience and the manufacturer's technical service support. It is difficult to be discovered and handled in the first time. Defect analysis and processing is time-consuming and labor-intensive, and the handling experience cannot be shared, which brings hidden dangers to the safe and stable operation of the power grid. The DC protection system contains a large number of components and the internal logic is extremely complex. Traditional data mining methods are difficult to fully extract the semantic features of the DC protection device defect text. Summary of the invention

[0004] In view of the above problems, the present invention proposes a method for DC protection device defect data mining and feature extraction, comprising:

[0005] Screening out defective text data from multi-source information data, and preprocessing the defective text data based on regular expressions to extract key words and useless words from the defective text data;

[0006] Based on the prefix dictionary word segmentation method, the preprocessed defect text data is subjected to secondary text segmentation to obtain a word segmentation vocabulary;

[0007] According to the segmentation results of the segmentation word library, the weight of the defect text data corresponding to each segmentation word is calculated, the defect text data with a weight higher than a minimum threshold is used as feature data, and the feature data is labeled;

[0008] The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index;

[0009] Based on the TF-IDF-CNN model, defect feature data of the target DC protection device is extracted according to the defect information data of the target DC protection device.

[0010] Optionally, and based on a regular expression, the defect text data is preprocessed, including:

[0011] Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words;

[0012] Based on the key words, a stop word library is constructed;

[0013] Among them, after extracting the key words, the words irrelevant to defect diagnosis in the key words are deleted.

[0014] Optional, key words, including: equipment model and converter station name.

[0015] Optionally, a prefix dictionary-based word segmentation method is used to perform secondary text segmentation on the preprocessed defect text data, including:

[0016] Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary;

[0017] The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position;

[0018] The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

[0019] Optionally, the weight of the defect text data corresponding to each word segmentation is calculated, including:

[0020] Calculate the frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the frequency and importance, calculate the weight. The calculation formula is as follows:

[0021] TF-IDF i=TF ij ×IDF i

[0022] Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

[0023] Optionally, labeling the feature data includes:

[0024] The defect phenomena and defect causes in the feature data are labeled.

[0025] Optionally, the extracted defect characteristic data of the target DC protection device includes:

[0026] Defect characteristics of equipment, components, defect information and treatment measures.

[0027] On the other hand, the present invention also proposes a system for DC protection device defect data mining and feature extraction, comprising:

[0028] A data mining unit, used to obtain multi-source information data of DC protection device defects, filter out defect text data in the multi-source information data, and pre-process the defect text data based on regular expressions to extract key words and useless words in the defect text data;

[0029] A data processing unit is used to perform secondary text segmentation processing on the preprocessed defect text data based on a segmentation method of a prefix dictionary to obtain a segmentation vocabulary, calculate the weight of the defect text data corresponding to each segmentation according to the segmentation result of the segmentation vocabulary, take the defect text data with a weight higher than a minimum threshold as feature data, and label the feature data;

[0030] The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index;

[0031] The feature extraction unit is used to extract defect feature data of the target DC protection device based on the defect information data of the target DC protection device based on the TF-IDF-CNN model.

[0032] Optionally, and based on a regular expression, the defect text data is preprocessed, including:

[0033] Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words;

[0034] Based on the key words, a stop word library is constructed;

[0035] Among them, after extracting the key words, the words irrelevant to defect diagnosis in the key words are deleted.

[0036] Optional, key words, including: equipment model and converter station name.

[0037] Optionally, a prefix dictionary-based word segmentation method is used to perform secondary text segmentation on the preprocessed defect text data, including:

[0038] Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary;

[0039] The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position;

[0040] The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

[0041] Optionally, the weight of the defect text data corresponding to each word segmentation is calculated, including:

[0042] Calculate the frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the frequency and importance, calculate the weight. The calculation formula is as follows:

[0043] TF-IDF i =TF ij ×IDF i

[0044] Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

[0045] Optionally, labeling the feature data includes:

[0046] The defect phenomena and defect causes in the feature data are labeled.

[0047] Optionally, the extracted defect characteristic data of the target DC protection device includes:

[0048] Defect characteristics of equipment, components, defect information and treatment measures.

[0049] In yet another aspect, the present invention further provides a computing device, comprising: one or more processors;

[0050] a processor for executing one or more programs;

[0051] When the one or more programs are executed by the one or more processors, the above-described method is implemented.

[0052] In yet another aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed, the method described above is implemented.

[0053] Compared with the prior art, the present invention has the following beneficial effects:

[0054] The present invention provides a method for DC protection device defect data mining and feature extraction, comprising: screening out defect text data from multi-source information data, and preprocessing the defect text data based on regular expressions to extract key words and useless words from the defect text data; performing secondary text segmentation processing on the preprocessed defect text data based on a prefix dictionary segmentation method to obtain a segmentation vocabulary; calculating the weight of the defect text data corresponding to each segmentation according to the segmentation result of the segmentation vocabulary, taking the defect text data with a weight higher than a minimum threshold as feature data, and labeling the feature data; the feature data of the label is used to train a TF-IDF-CNN model, and the TF-IDF-CNN model reaches the optimal evaluation index; based on the TF-IDF-CNN model, extracting the defect feature data of the target DC protection device according to the defect information data of the target DC protection device. The present invention can efficiently extract the features of the defect text data of the DC protection device. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 is a flow chart of the method of the present invention;

[0056] Figure 2 A flowchart of an embodiment of the method of the present invention;

[0057] Figure 3 It is a principle diagram of an embodiment of the method of the present invention;

[0058] Figure 4 This is a feature extraction result diagram of an embodiment of the method of the present invention;

[0059] Figure 5It is a structural diagram of the system of the present invention. DETAILED DESCRIPTION

[0060] Now, exemplary embodiments of the present invention are described with reference to the accompanying drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. These embodiments are provided to disclose the present invention in detail and completely and to fully convey the scope of the present invention to those skilled in the art. The terms used in the exemplary embodiments shown in the accompanying drawings are not intended to limit the present invention. In the accompanying drawings, the same units / elements are marked with the same reference numerals.

[0061] Unless otherwise specified, the terms (including technical terms) used herein have the commonly understood meanings to those skilled in the art. In addition, it is understood that the terms defined in commonly used dictionaries should be understood to have the same meanings as those in the context of the relevant fields, and should not be understood as idealized or overly formal meanings.

[0062] Embodiment 1:

[0063] The present invention proposes a method for DC protection device defect data mining and feature extraction, such as Figure 1 As shown, including:

[0064] Step 1: Filter out defective text data from multi-source information data, and pre-process the defective text data based on regular expressions to extract key words and useless words from the defective text data;

[0065] Step 2: Based on the prefix dictionary word segmentation method, the preprocessed defect text data is subjected to secondary text segmentation processing to obtain a word segmentation vocabulary;

[0066] Step 3: Calculate the weight of the defect text data corresponding to each segmentation word according to the segmentation results of the segmentation word library, take the defect text data with a weight higher than the minimum threshold as feature data, and label the feature data;

[0067] The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index;

[0068] Step 4: Based on the TF-IDF-CNN model, the defect feature data of the target DC protection device is extracted according to the defect information data of the target DC protection device.

[0069] Wherein, based on the regular expression, the defect text data is preprocessed, including:

[0070] Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words;

[0071] Based on the key words, a stop word library is constructed;

[0072] Among them, after extracting the key words, the words irrelevant to defect diagnosis in the key words are deleted.

[0073] Among them, key words include: equipment model and converter station name.

[0074] Among them, the word segmentation method based on the prefix dictionary performs secondary text segmentation processing on the preprocessed defect text data, including:

[0075] Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary;

[0076] The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position;

[0077] The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

[0078] Among them, the weight of the defect text data corresponding to each word segmentation is calculated, including:

[0079] Calculate the frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the frequency and importance, calculate the weight. The calculation formula is as follows:

[0080] TF-IDF i =TF ij ×IDF i

[0081] Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

[0082] The step of labeling the feature data includes:

[0083] The defect phenomena and defect causes in the feature data are labeled.

[0084] The extracted defect characteristic data of the target DC protection device includes:

[0085] Defect characteristics of equipment, components, defect information and treatment measures.

[0086] The present invention is further described below in conjunction with specific implementation cases:

[0087] The implementation steps are as follows Figure 2 As shown, the implementation principle is as follows Figure 3 As shown, specifically including:

[0088] S101, performing stop word preprocessing on defect text data of the DC protection device based on regular expressions;

[0089] S102, performing secondary text segmentation on the DC protection defect text data using a segmentation method based on a prefix dictionary;

[0090] S103, calculating the comprehensive weight according to the word segmentation result, and labeling the feature data;

[0091] S104, output the defect feature information extraction result to form a feature library such as defect type, defect cause, model, etc.

[0092] In specific implementation, in step S101, the defect text data of the DC protection device is processed based on the regular expression in the following manner:

[0093] S201. Define text matching rules through grammar and character combination, compare the set regular expression with the text to be matched, extract key words in the multi-source information data of DC protection defects, including equipment model, converter station name, etc., and remove words that are not very useful for defect diagnosis. The regular expression is shown in Table 1:

[0094] Table 1

[0095]

[0096] S202, performing regular expression processing on the DC protection defect data to construct a stop word list, as shown in Table 2.

[0097] Table 2

[0098]

[0099] In step S102, the secondary text segmentation of the DC protection defect text data is performed based on the prefix dictionary segmentation method in the following manner:

[0100] S301. Based on the professional dictionary of DC protection device defects, a discontinued word library of DC protection defect data is imported to construct a prefix dictionary.

[0101] S302, segmenting the defect data using a prefix dictionary, and constructing a directed acyclic graph according to the segmentation positions;

[0102] S303, calculating the probability of each adopted word of the directed acyclic graph appearing in the dictionary, the formula is:

[0103]

[0104] Where W i Represents the i-th adopted word.

[0105] S304, solve the cumulative probability. The cumulative probability of each word is equal to the original probability multiplied by the maximum left neighbor word probability. The path with the maximum cumulative probability is obtained as the optimal segmentation form of the defective text. The formula is:

[0106] P′(w i )=P(w i )×P(w i-1 )

[0107] In step S103, the comprehensive weight is calculated according to the word segmentation result in the following way:

[0108] S401, calculate the frequency of word i in text j, the formula is:

[0109]

[0110] Where TF ij Represents the frequency of word i in text j, num ij is the number of times i appears in text j, the denominator is the sum of the number of times all words appear in text j, and k is the number of different words in text j.

[0111] S402, calculate and measure the importance of word i in text j, the formula is:

[0112]

[0113] In the formula, IDF i represents the inverse document frequency of word i in text j, N Doc is the total number of documents in the database, N i is the number of documents in the database that contain word i. 1 is added to prevent N i The case of zero.

[0114] S403. Calculate the TF-IDF comprehensive weight. The formula is:

[0115] TF-IDF i =TF ij ×IDF i

[0116] S404. Label the feature data whose comprehensive weight is higher than the minimum threshold, including labels such as defect phenomenon and defect cause. Use the label data and evaluation indicators to train the TF-IDF-CNN convolutional neural network model, adjust the model parameters, and make the model reach the optimal evaluation indicator.

[0117] Step S104, outputting the defect feature information extraction result is implemented by the following method:

[0118] By training the TF-IDF-CNN model using DC protection defect text data, we can extract defect features such as equipment, components, defect information, and disposal measures. Taking the defect data in the DC converter station as an example, the specific example is as follows: Figure 4 As shown in the figure, the defect description of a defective device is such as: "AC filter control protection AFP3B host crashes, and restart fails." After defect data word segmentation preprocessing, it becomes: "AC filter / control protection / AFP3B / host / crash / , / restart / failed." Data feature extraction is performed on it to obtain that the defective device is AC filter protection, the component is the host, the model is AFP3B, the defect phenomenon is crash, and the disposal measure is restart.

[0119] The present invention effectively segments the DC protection defect text by combining regular expressions and Jieba segmentation methods, constructs a defect text segmentation database, and effectively improves the segmentation efficiency and database quality.

[0120] The present invention can extract defect features such as equipment, components, defect information, and disposal measures from unstructured DC protection device defect text data through continuous learning, thereby achieving efficient extraction of DC protection device defect text data features.

[0121] Embodiment 2:

[0122] The present invention also proposes a system 200 for DC protection device defect data mining and feature extraction, such as Figure 5 As shown, including:

[0123] The data mining unit 201 is used to obtain multi-source information data of DC protection device defects, filter out defect text data in the multi-source information data, and pre-process the defect text data based on regular expressions to extract key words and useless words in the defect text data;

[0124] The data processing unit 202 is used to perform secondary text segmentation processing on the preprocessed defect text data based on the word segmentation method of the prefix dictionary to obtain a word segmentation vocabulary, calculate the weight of the defect text data corresponding to each word segmentation according to the word segmentation result of the word segmentation vocabulary, take the defect text data with a weight higher than a minimum threshold as feature data, and label the feature data;

[0125] The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index;

[0126] The feature extraction unit 203 is used to extract defect feature data of the target DC protection device according to the defect information data of the target DC protection device based on the TF-IDF-CNN model.

[0127] Wherein, based on the regular expression, the defect text data is preprocessed, including:

[0128] Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words;

[0129] Based on the key words, a stop word library is constructed;

[0130] Among them, after extracting the key words, the words irrelevant to defect diagnosis in the key words are deleted.

[0131] Among them, key words include: equipment model and converter station name.

[0132] Among them, the word segmentation method based on the prefix dictionary performs secondary text segmentation processing on the preprocessed defect text data, including:

[0133] Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary;

[0134] The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position;

[0135] The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

[0136] Among them, the weight of the defect text data corresponding to each word segmentation is calculated, including:

[0137] Calculate the frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the frequency and importance, calculate the weight. The calculation formula is as follows:

[0138] TF-IDF i =TF ij ×IDF i

[0139] Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

[0140] The step of labeling the feature data includes:

[0141] The defect phenomena and defect causes in the feature data are labeled.

[0142] The extracted defect characteristic data of the target DC protection device includes:

[0143] Defect characteristics of equipment, components, defect information and treatment measures.

[0144] The present invention can efficiently extract the features of the defect text data of the DC protection device.

[0145] Embodiment 3:

[0146] Based on the same inventive concept, the present invention also provides a computer device, which includes a processor and a memory, wherein the memory is used to store a computer program, the computer program includes program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is suitable for implementing one or more instructions, and is specifically suitable for loading and executing one or more instructions in the computer storage medium to implement the corresponding method flow or corresponding functions, so as to implement the steps of the method in the above embodiment.

[0147] Embodiment 4:

[0148] Based on the same inventive concept, the present invention also provides a storage medium, specifically a computer-readable storage medium (Memory), which is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both a built-in storage medium in a computer device and an extended storage medium supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the steps of the method in the above embodiment.

[0149] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes. The schemes in the embodiments of the present invention may be implemented in various computer languages, for example, object-oriented programming language Java and literal scripting language JavaScript, etc.

[0150] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0151] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0152] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0153] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0154] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A method for DC protection device defect data mining and feature extraction, characterized in that: The method comprises: Screening out defective text data from multi-source information data, and preprocessing the defective text data based on regular expressions to extract key words and useless words from the defective text data; Based on the prefix dictionary word segmentation method, the preprocessed defect text data is subjected to secondary text segmentation to obtain a word segmentation vocabulary; According to the segmentation results of the segmentation word library, the weight of the defect text data corresponding to each segmentation word is calculated, the defect text data with a weight higher than a minimum threshold is used as feature data, and the feature data is labeled; The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index; Based on the TF-IDF-CNN model, defect feature data of the target DC protection device is extracted according to the defect information data of the target DC protection device.

2. The method according to claim 1, characterized in that: The method of preprocessing the defect text data based on the regular expression includes: Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words; Based on the key words, a stop word library is constructed; Among them, after the key words are extracted, the words in the key words that are not related to defect diagnosis are deleted.

3. The method according to claim 2, characterized in that The key words include: equipment model and converter station name.

4. The method according to claim 1, characterized in that The word segmentation method based on the prefix dictionary performs secondary text segmentation processing on the preprocessed defect text data, including: Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary; The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position; The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

5. The method according to claim 1, characterized in that The step of calculating the weight of the defect text data corresponding to each word segmentation includes: Calculate the word frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the word frequency and importance, calculate the weight. The calculation formula is as follows: TF-IDF i =TF ij ×IDF i Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

6. The method according to claim 1, characterized in that The labeling of the feature data includes: The defect phenomena and defect causes in the characteristic data are labeled.

7. The method according to claim 1, characterized in that The extracted defect characteristic data of the target DC protection device includes: Defect characteristics of equipment, components, defect information and treatment measures.

8. A system for DC protection device defect data mining and feature extraction, characterized in that: The system comprises: A data mining unit, used to obtain multi-source information data of DC protection device defects, filter out defect text data in the multi-source information data, and pre-process the defect text data based on regular expressions to extract key words and useless words in the defect text data; A data processing unit is used to perform secondary text segmentation processing on the preprocessed defect text data based on a segmentation method of a prefix dictionary to obtain a segmentation vocabulary, calculate the weight of the defect text data corresponding to each segmentation according to the segmentation result of the segmentation vocabulary, take the defect text data with a weight higher than a minimum threshold as feature data, and label the feature data; The feature data of the label is used to train the TF-IDF-CNN model, and the TF-IDF-CNN model is made to achieve the optimal evaluation index; The feature extraction unit is used to extract defect feature data of the target DC protection device based on the defect information data of the target DC protection device based on the TF-IDF-CNN model.

9. The system according to claim 8, characterized in that The method of preprocessing the defect text data based on the regular expression includes: Define text matching rules through grammar and character combination, set regular expressions based on the matching rules, compare the regular expressions with the defective text data, and extract key words; Based on the key words, a stop word library is constructed; Among them, after the key words are extracted, the words in the key words that are not related to defect diagnosis are deleted.

10. The system according to claim 9, characterized in that The key words include: equipment model and converter station name.

11. The system according to claim 8, characterized in that The word segmentation method based on the prefix dictionary performs secondary text segmentation processing on the preprocessed defect text data, including: Import the DC protection device defect professional dictionary into the stop word library to construct a prefix dictionary; The word segmentation method based on prefix dictionary is used to segment the defect text data, and a directed acyclic graph is constructed according to the segmentation position; The probability of each adopted word of the directed acyclic graph appearing in the prefix dictionary is calculated, and the cumulative probability is solved. The path with the largest cumulative probability is used as the optimal segmentation form of the defective text data, and the defective text data is subjected to secondary text segmentation processing.

12. The system according to claim 8, characterized in that The step of calculating the weight of the defect text data corresponding to each word segmentation includes: Calculate the word frequency of each word in the defect text data and the importance of each word in the defect text data. Based on the word frequency and importance, calculate the weight. The calculation formula is as follows: TF-IDF i =TF ij ×IDF i Among them, TF-IDF i is the weight of the i-th word, TF ij is the frequency of the i-th word in the j-th defect text data, IDF i is the importance of the i-th word in the defective text data.

13. The system according to claim 8, characterized in that The labeling of the feature data includes: The defect phenomena and defect causes in the characteristic data are labeled.

14. The system according to claim 8, characterized in that The extracted defect characteristic data of the target DC protection device includes: Defect characteristics of equipment, components, defect information and treatment measures.

15. A computer device, characterized in that: include: one or more processors; a processor for executing one or more programs; When the one or more programs are executed by the one or more processors, the method according to any one of claims 1 to 7 is implemented.

16. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed, the method according to any one of claims 1 to 7 is implemented.

Images