Intelligent decision-making method for operation and inspection of reactive power compensation equipment and terminal

Abstract

The invention discloses a reactive power compensation equipment operation inspection intelligent decision-making method and a terminal, and the method comprises the steps: carrying out the word segmentation of operation inspection data of reactive power compensation equipment, obtaining a word list, obtaining a word vector space according to the word list, extracting a feature vector from the word vector space, carrying out the matching of the feature vector with a preset knowledge graph, obtaining a matching result, and carrying out the operation inspection of the reactive power compensation equipment; if the matching result comprises an abnormal information node, generating an operation inspection decision based on the abnormal information node, preprocessing operation inspection data of the reactive compensation equipment through word segmentation processing and word vector space conversion, converting text data into numerical vectors which can be understood by a machine, and reserving a semantic relationship between words, so that subsequent analysis and matching are more convenient, and the accuracy of the operation inspection data is improved. Whether the operation inspection data is abnormal or not is determined in a mode of matching the feature vector with the preset knowledge graph, and the operation inspection decision is automatically generated based on the matched abnormal information node under the condition that the operation inspection data is determined to be abnormal, so that the operation inspection efficiency of the reactive compensation equipment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of operation and inspection of reactive power compensation equipment, and in particular to an intelligent decision-making method and terminal for operation and inspection of reactive power compensation equipment. Background Art

[0002] Reactive power compensation equipment is an important part of the power system. Comprehensively grasping and accurately evaluating the operating status of reactive power compensation equipment is the primary problem and difficulty to ensure the safe operation of the equipment, and it is also the main technical bottleneck of the intelligentization of power equipment. As one of the core equipment of power equipment, the operating status of reactive power compensation equipment will directly affect the power supply reliability of the power grid. Reactive power compensation refers to the regulation and compensation of reactive power in the power system to maintain the stable operation of the system. In the AC circuit, power can be divided into active power and reactive power. Active power is used to do work, while reactive power is used to maintain voltage stability. The existence of reactive power will cause a deviation between the current waveform and the voltage waveform in the system, affecting the stability of the system. Reactive power compensation equipment is a device used to regulate and compensate reactive power in the power system. It can regulate the reactive power in the power system through components such as capacitors, inductors, and static power converters, thereby improving the power factor of the system, improving the power quality, and reducing the reactive loss of the system to ensure the stable operation of the system.

[0003] With the continuous expansion of the scale of power systems and the rapid development of smart grids, reactive power compensation equipment plays an increasingly important role in ensuring the safe and stable operation of power grids. However, for reactive power compensation equipment, the traditional manual operation and inspection methods have problems such as low efficiency, strong subjectivity in decision-making, and difficulty in dealing with complex situations. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide an intelligent decision-making method and terminal for operation and inspection of reactive compensation equipment, which can improve the operation and inspection efficiency of reactive compensation equipment.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A reactive power compensation equipment operation and inspection intelligent decision-making method, comprising the steps of:

[0007] Obtain operation and inspection data of reactive power compensation equipment;

[0008] Segmenting the inspection data to obtain a word list, and obtaining a word vector space according to the word list;

[0009] Extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result, wherein the preset knowledge graph includes abnormal information nodes and normal information nodes;

[0010] If the matching result includes the abnormal information node, an operation and inspection decision is generated based on the abnormal information node.

[0011] In order to solve the above technical problems, another technical solution adopted by the present invention is:

[0012] A reactive power compensation equipment operation and inspection intelligent decision terminal comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the following steps when executing the computer program:

[0013] Obtain operation and inspection data of reactive power compensation equipment;

[0014] Segmenting the inspection data to obtain a word list, and obtaining a word vector space according to the word list;

[0015] Extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result, wherein the preset knowledge graph includes abnormal information nodes and normal information nodes;

[0016] If the matching result includes the abnormal information node, an operation and inspection decision is generated based on the abnormal information node.

[0017] The beneficial effects of the present invention are: performing word segmentation on the operation and inspection data of the reactive compensation equipment to obtain a word list, and obtaining a word vector space based on the word list, extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result. If the matching result includes an abnormal information node, an operation and inspection decision is generated based on the abnormal information node. The operation and inspection data of the reactive compensation equipment is preprocessed through word segmentation and conversion of the word vector space, and the text data is converted into a machine-understandable numerical vector, and the semantic relationship between words is retained, which is more convenient for subsequent analysis and matching. Whether the operation and inspection data is abnormal is determined by matching the feature vector with the preset knowledge graph. When an abnormality is determined, an operation and inspection decision is automatically generated based on the matched abnormal information node, thereby improving the operation and inspection efficiency of the reactive compensation equipment. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a method for intelligent decision-making in operation and inspection of reactive power compensation equipment according to an embodiment of the present invention;

[0019] Figure 2 The present invention is a schematic diagram of the structure of a reactive compensation equipment operation and maintenance intelligent decision-making terminal according to an embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to explain the technical content, achieved objectives and effects of the present invention in detail, the following is an explanation in combination with the implementation modes and the accompanying drawings.

[0021] Please refer to Figure 1 , a reactive power compensation equipment operation and inspection intelligent decision-making method, comprising the steps of:

[0022] Obtain operation and inspection data of reactive power compensation equipment;

[0023] Segmenting the inspection data to obtain a word list, and obtaining a word vector space according to the word list;

[0024] Extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result, wherein the preset knowledge graph includes abnormal information nodes and normal information nodes;

[0025] If the matching result includes the abnormal information node, an operation and inspection decision is generated based on the abnormal information node.

[0026] From the above description, it can be seen that the beneficial effects of the present invention are: the operation and inspection data of the reactive compensation equipment are segmented to obtain a word list, and a word vector space is obtained based on the word list, a feature vector is extracted from the word vector space, and the feature vector is matched with a preset knowledge graph to obtain a matching result. If the matching result includes an abnormal information node, an operation and inspection decision is generated based on the abnormal information node, and the operation and inspection data of the reactive compensation equipment are pre-processed through word segmentation processing and conversion of the word vector space, and the text data is converted into a machine-understandable numerical vector, and the semantic relationship between words is retained, which is more convenient for subsequent analysis and matching. The operation and inspection data is determined to be abnormal by matching the feature vector with the preset knowledge graph. When an abnormality is determined, an operation and inspection decision is automatically generated based on the matched abnormal information node, thereby improving the operation and inspection efficiency of the reactive compensation equipment.

[0027] Furthermore, the segmenting of the operation and inspection data to obtain a word list includes:

[0028] Use the precise mode to perform Jieba word segmentation on the inspection data to obtain a word list.

[0029] From the above description, we can know that Jieba word segmentation has three word segmentation modes, namely precise mode, full mode and search engine mode. The operation and inspection data of reactive compensation equipment usually contains a large number of professional terms and technical indicators. Using the precise mode to perform Jieba word segmentation on the operation and inspection data can identify all possible words as much as possible, including some shorter words and professional terms, so as to improve the accuracy and professionalism of word segmentation.

[0030] Furthermore, obtaining a word vector space according to the word list includes:

[0031] Mapping all words in the word list to indexes to generate a dictionary;

[0032] Input each word in the dictionary into the Word2Vec model and output the vector representation of the word;

[0033] Generate a word vector space based on the vector representation of all words.

[0034] From the above description, it can be seen that using the Word2Vec model to process the word list obtained by word segmentation can make full use of the word frequency information and the semantic relationship of words, and can provide better feature representation and classification effects for a large amount of technical documents and text data involved in the intelligent decision-making of reactive compensation equipment.

[0035] Furthermore, extracting a feature vector from the word vector space includes:

[0036] The TF-IDF algorithm is used to extract feature vectors from the word vector space.

[0037] From the above description, it can be seen that using the TF-IDF algorithm to extract feature vectors from the word vector space is simple, effective, highly interpretable and easy to implement.

[0038] Furthermore, the extracting a feature vector from the word vector space using the TF-IDF algorithm includes:

[0039] Calculate the word frequency and inverse document frequency represented by each word vector in the word vector space;

[0040] Obtaining a weight for each word vector representation according to the word frequency and the inverse document frequency;

[0041] All word vector representations are sorted in descending order according to the weights to obtain sorted word vector representations, and a preset number of word vector representations are selected from the sorted word vector representations in turn as feature vectors.

[0042] From the above description, we can see that by calculating the word frequency and inverse document frequency to measure the importance of features, the original data is converted into a more compact and representative feature vector, thereby effectively reducing the dimension of the data and improving computing efficiency and data processing capabilities.

[0043] Further, the calculating the word frequency and inverse document frequency represented by each word vector in the word vector space includes:

[0044]

[0045] In the formula, TF(t,d) represents the frequency of the word vector representing t, and f t,d Indicates that the word vector represents the number of times t appears in the data set d, represents the total number of occurrences of the word vector representation in the data set d, IDF(t,D) represents the inverse document frequency of the word vector representation t, N represents the total number of documents in the document set D, and |{d∈D:t∈d}| represents the number of documents containing the word vector representation t.

[0046] From the above description, we can know that by calculating the word frequency and inverse document frequency represented by each word vector in the word vector space, a high TF value means that the feature appears frequently in the current document, while a high IDF value means that the feature appears less frequently in all documents. By calculating these two values, we can highlight the important features that are discriminative in a specific document, effectively filter out common but meaningless features, and ensure the reliability of subsequent decisions.

[0047] Furthermore, the weight of each word vector representation obtained according to the word frequency and the inverse document frequency includes:

[0048] TF-IDF(t,d,D)=TF(t,d)×IDF(t,D);

[0049] Where TF-IDF(t,d,D) represents the weight of the word vector representation t.

[0050] From the above description, we can see that the weight of each word vector representation is obtained according to the word frequency and inverse document frequency, so as to reflect its importance in the document.

[0051] Furthermore, the matching of the feature vector with a preset knowledge graph to obtain a matching result includes:

[0052] Calculating the cosine similarity between each of the feature vectors and the abnormal information node and the normal information node;

[0053] If the cosine similarity is greater than a preset value, it is determined that the feature vector and the node corresponding to the cosine similarity match, and a matching result is generated according to the matched feature vector and the node, wherein the node includes the abnormal information node or the normal information node.

[0054] From the above description, it can be seen that the cosine similarity between each feature vector and the abnormal information node and the normal information node is calculated. If the cosine similarity is greater than the preset value, the feature vector and node corresponding to the cosine similarity are determined to match. In this way, it can be determined whether there are abnormal data in the operation and inspection data or they are all normal data, thereby judging whether the reactive compensation equipment has an abnormality, so as to make intelligent decisions later.

[0055] Further, generating an operation and inspection decision based on the abnormal information node includes:

[0056] Determine a decision rule corresponding to the abnormal information node from a preset decision rule library;

[0057] An operation and inspection decision is generated according to the decision rule.

[0058] It can be seen from the above description that the decision rule corresponding to the abnormal information node is determined from the preset decision rule library, and the operation and inspection decision is automatically generated according to the decision rule, thereby improving the efficiency of the operation and inspection.

[0059] Please refer to Figure 2 Another embodiment of the present invention provides a reactive compensation equipment operation and inspection intelligent decision-making terminal, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step in the above-mentioned reactive compensation equipment operation and inspection intelligent decision-making method is implemented.

[0060] The above-mentioned reactive power compensation equipment operation and inspection intelligent decision-making method and terminal of the present invention can be applied to the reactive power compensation equipment operation and inspection scenario, and the following is described through specific implementation methods:

[0061] Please refer to Figure 1 , Embodiment 1 of the present invention is:

[0062] A reactive power compensation equipment operation and inspection intelligent decision-making method, comprising the steps of:

[0063] S1. Obtain the operation and inspection data of reactive power compensation equipment.

[0064] The operation and inspection data include operation data and historical inspection and maintenance data. In an optional implementation, the operation data of the reactive power compensation equipment is obtained from smart meters and sensors, and the historical inspection and maintenance data is obtained from an operation and maintenance management system or paper records.

[0065] In an optional implementation, after S1 and before S2, it also includes: cleaning the inspection data to convert data from different sources into a unified format to obtain cleaned data; filling missing values ​​in the cleaned data to obtain filled data; removing abnormal values ​​from the filled data to obtain removed data; removing duplicate data from the removed data to obtain final inspection data. Among them, interpolation or mean filling is used to fill missing values ​​in the cleaned data to obtain filled data; statistical methods or machine learning algorithms are used to detect abnormal values ​​in the filled data and remove abnormal values, such as current mutations, invalid readings, etc.

[0066] S2, segmenting the inspection data to obtain a word list, and obtaining a word vector space according to the word list, specifically including S21-S24:

[0067] S21. Use the precise mode to perform Jieba word segmentation on the operation and inspection data to obtain a word list.

[0068] Specifically, the Jieba word segmentation library is imported and the operation inspection data is loaded, and the jieba.cut method is called to segment the operation inspection data to obtain a word list.

[0069] For example, suppose there are three pieces of operation and inspection data: "2023-10-26 10:00:00 The equipment is running normally, the voltage is stable, the current fluctuation is small, and the power factor is 0.95.", "2023-11-15 14:30:00 The equipment fails, the current is abnormal, and the temperature rises to 80 degrees Celsius, requiring immediate maintenance.", "2024-01-05 09:45:00 The equipment is running stably, the voltage is normal, the power factor is 0.98, and no abnormality is found." After word segmentation, the word lists obtained are: W 1={2023-10-26, 10:00:00, equipment, operation, normal, voltage, stable, current, fluctuation, small, power factor, is, 0.95}; W2={2023-11-15, 14:30:00, equipment, fault, current, abnormal, temperature, increased, to, 80, degrees Celsius, immediate, maintenance}; W3={2024-01-05, 09:45:00, equipment, operation, stable, voltage, normal, power factor, is, 0.98, no, abnormal, situation, found}.

[0070] S22. Map all words in the word list to indexes to generate a dictionary.

[0071] Specifically, different words among all the words in the word list are mapped to different indexes, and the same word is mapped to the same index, so as to generate a dictionary.

[0072] Continuing with the above example, the generated dictionary is as follows:

[0073] {'2023-10-26':1,'10:00:00':2,'Equipment':3,'Operation':4,'Normal':5,'Voltage':6,'Stable':7,'Current':8,'Fluctuation':9,'Small':10,'Power Factor':11,'For':12,'0.95':13,'2023-11-15':14,'14:30:00 ':15,'occurred':16,'fault':17,'abnormal':18,'temperature':19,'increased':20,'to':21,'80':22,'degrees Celsius':23,'immediately':24,'maintenance':25,'2024-01-05':26,'09:45:00':27,'not':28,'found':29,'situation':30}.

[0074] S23. Input each word in the dictionary into the Word2Vec model and output the vector representation of the word.

[0075] In an optional implementation, each word in the dictionary is input into a Skip-gram model, and a vector representation of the word is output.

[0076] Among them, the training process for the Skip-gram model is as follows:

[0077] (1) For each target word, select several surrounding words as context words to form a training sample. For example, for the target word "voltage", its context words can be "stable" and "current", forming training samples ("voltage", "stable") and ("voltage", "current").

[0078] (2) The structure of the Skip-gram model consists of two layers, the input layer and the output layer. The input layer consists of the vector representation of the target words. The vector representation of each target word can be obtained through One-hot encoding or other word embedding methods. The output layer is a softmax layer, which outputs the probability of each context word in the dictionary.

[0079] (3) The vector representation of the target word is input into the Skip-gram model. The model calculates the probability distribution of context words in the dictionary through a neural network, uses the cross-entropy loss function to measure the difference between the probability distribution predicted by the model and the true label, and updates the model parameters through the back-propagation algorithm to minimize the loss function, thus obtaining the trained Skip-gram model.

[0080] Each word in the dictionary is input into the trained Skip-gram model, and each word will get a vector representation, for example: v_{voltage} = [0.1, 0.2, 0.3, ..., 0.9]; v_{current} = [0.2, 0.3, 0.4, ..., 0.9]. In actual use, the corpus needs to be trained to obtain the specific vector representation of the vocabulary.

[0081] S24. Generate a word vector space based on the vector representations of all words.

[0082] Combining Jieba word segmentation with Word2Vec for data preprocessing can provide a faster word segmentation speed while ensuring the accuracy of word segmentation, which is suitable for large-scale text processing.

[0083] S3, extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result, wherein the preset knowledge graph includes abnormal information nodes and normal information nodes, specifically including S31-S33:

[0084] S31, extracting feature vectors from the word vector space using the TF-IDF algorithm, specifically including S311-S313:

[0085] S311, calculating the word frequency and inverse document frequency represented by each word vector in the word vector space, specifically:

[0086]

[0087] In the formula, TF(t,d) represents the frequency of the word vector representing t, and f t,d Indicates that the word vector represents the number of times t appears in the data set d, represents the total number of occurrences of the word vector representation in the data set d, IDF(t,D) represents the inverse document frequency of the word vector representation t, N represents the total number of documents in the document set D, and |{d∈D:t∈d}| represents the number of documents containing the word vector representation t.

[0088] Among them, the data set d refers to the original inspection data, the document set D refers to the written materials that record and describe the data set, the word frequency refers to the frequency of a word appearing in the document, and the inverse document frequency measures the rarity of a data in the entire document set.

[0089] S312: Obtain the weight of each word vector representation according to the word frequency and the inverse document frequency, specifically:

[0090] TF-IDF(t,d,D)=TF(t,d)×IDF(t,D);

[0091] Where TF-IDF(t,d,D) represents the weight of the word vector representation t.

[0092] S313. Sort all word vector representations in descending order of the weights to obtain sorted word vector representations, and select a preset number of word vector representations from the sorted word vector representations in turn as feature vectors.

[0093] S32: Calculate the cosine similarity between each of the feature vectors and the abnormal information node and the normal information node.

[0094] In the knowledge graph, entities and relationships can be represented as vectors. The formula for calculating cosine similarity is:

[0095]

[0096] In the formula, CS(A,B) represents the cosine similarity between vector A and vector B, A·B represents the dot product of vector A and vector B, n represents the total number of vectors, A i Represents the i-th vector A, B irepresents the i-th vector B, ||A|| represents the modulus of vector A, and ||B|| represents the modulus of vector B.

[0097] S33. If the cosine similarity is greater than a preset value, determine that the feature vector and the node corresponding to the cosine similarity match, and generate a matching result according to the matched feature vector and the node, wherein the node includes the abnormal information node or the normal information node.

[0098] Among them, the preset knowledge graph is a knowledge graph based on the Neo4j database, which is suitable for scenarios that need to process a large amount of complex relational data, and has good scalability and performance; the nodes in the preset knowledge graph describe the normal and abnormal conditions of the reactive compensation equipment, the abnormal information node is the node that describes the abnormal condition of the equipment, and the normal information node is the node that describes the normal condition of the equipment. By matching the feature vector with the nodes in the preset knowledge graph, it can be judged whether the operation and inspection data is normal or abnormal.

[0099] For example, assuming that the feature vector includes high voltage, device overheating, and frequency fluctuation, the nodes in the preset knowledge graph include node 1: voltage abnormality, node 2: device overtemperature, and node 3: frequency instability.

[0100] After the node is vectorized, its cosine similarity with the feature vector is calculated. Assuming the preset value is 0.8, the matching results are shown in Table 1.

[0101] Table 1 Matching results

[0102]

[0103] S4. If the matching result includes the abnormal information node, generating an operation and inspection decision based on the abnormal information node, specifically including S41-S42:

[0104] S41. If the matching result includes the abnormal information node, determine a decision rule corresponding to the abnormal information node from a preset decision rule library.

[0105] Specifically, if the matching result includes the abnormal information node, the abnormal information node is parsed to obtain a parsing result, and a decision rule corresponding to the abnormal information node is determined from a preset decision rule library according to the parsing result.

[0106] The preset decision rule base contains various rules and strategies related to the operation and inspection of reactive power compensation equipment. These rules can be formulated based on experience, industry standards and expert knowledge. The preset decision rule base includes:

[0107] (1) Equipment operation status evaluation rules: Evaluate whether the current operation status of the equipment is normal.

[0108] (2) Equipment failure prediction rules: predict the possibility of equipment failure based on historical data and current status.

[0109] (3) Maintenance plan formulation rules: Develop maintenance plans based on equipment operating status and prediction results.

[0110] (4) Environmental adaptability adjustment rules: adjust operating parameters according to changes in the environment in which the equipment is located.

[0111] S42: Generate an operation and inspection decision according to the decision rule.

[0112] In an optional implementation manner, after S42, the following steps are further included:

[0113] The operation and inspection decision is verified by using simulation test or small-scale trial operation method to obtain verification results; the operation and inspection decision is adjusted and optimized according to the verification results to obtain the final operation and inspection decision, thereby ensuring the rationality, feasibility, effectiveness and reliability of the decision.

[0114] For example, after parsing the matching results, we get:

[0115] Equipment operating status: temperature is higher than the normal range and load fluctuations are large.

[0116] Historical fault records: Multiple failures due to high temperature.

[0117] Environmental information: The current load fluctuates frequently.

[0118] The preset decision rule library includes:

[0119] Rule 1: When the temperature is higher than the threshold and the load fluctuates greatly, emergency cooling is required.

[0120] Rule 2: For equipment that has repeatedly failed due to high temperatures, the inspection frequency should be increased during high temperature periods.

[0121] Rule 3: When load fluctuations occur frequently, adjust equipment operating parameters to adapt to environmental changes.

[0122] Using rules 1 and 2, we generate emergency cooling decisions and decisions to increase inspection frequency, including operation steps and operation time. Then we verify the effect of cooling through simulation tests to ensure the feasibility of the inspection plan. We optimize the decision based on the simulation test results and actual conditions to ensure its effectiveness.

[0123] Please refer to Figure 2 , Embodiment 2 of the present invention is:

[0124] A reactive compensation equipment operation and inspection intelligent decision-making terminal comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, each step in the reactive compensation equipment operation and inspection intelligent decision-making method in Embodiment 1 is implemented.

[0125] In summary, the present invention provides an intelligent decision-making method and terminal for operation and inspection of reactive compensation equipment, which performs word segmentation on the operation and inspection data of the reactive compensation equipment to obtain a word list, and obtains a word vector space based on the word list, extracts a feature vector from the word vector space, and matches the feature vector with a preset knowledge graph to obtain a matching result. If the matching result includes an abnormal information node, an operation and inspection decision is generated based on the abnormal information node, and the operation and inspection data of the reactive compensation equipment are pre-processed by word segmentation processing and conversion of the word vector space, and the text data is converted into a numerical vector that can be understood by a machine, and the semantic relationship between words is retained, which is more convenient for subsequent analysis and matching, and the feature vector is used to generate an operation and inspection decision based on the abnormal information node. The method of matching with the preset knowledge graph is used to determine whether the operation and inspection data is abnormal. If it is determined to be abnormal, the operation and inspection decision is automatically generated based on the matched abnormal information node, thereby improving the operation and inspection efficiency of the reactive compensation equipment; in addition, the Word2Vec model is used to process the word list obtained by word segmentation, which can make full use of the word frequency information and the semantic relationship of the words, and can provide better feature representation and classification effects for the large amount of technical documents and text data involved in the intelligent decision-making of reactive compensation equipment; and the decision rules corresponding to the abnormal information nodes are determined from the preset decision rule library, and the operation and inspection decisions are automatically generated according to the decision rules, thereby improving the efficiency of the operation and inspection.

[0126] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent transformations made using the contents of the present invention's specification and drawings, or directly or indirectly applied in related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. An intelligent decision-making method for operation and inspection of reactive power compensation equipment, characterized in that: Includes steps: Obtain operation and inspection data of reactive power compensation equipment; Segmenting the inspection data to obtain a word list, and obtaining a word vector space according to the word list; Extracting a feature vector from the word vector space, and matching the feature vector with a preset knowledge graph to obtain a matching result, wherein the preset knowledge graph includes abnormal information nodes and normal information nodes; If the matching result includes the abnormal information node, an operation and inspection decision is generated based on the abnormal information node.

2. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 1, characterized in that: The word list obtained by segmenting the operation and inspection data includes: Use the precise mode to perform Jieba word segmentation on the inspection data to obtain a word list.

3. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 1, characterized in that: The obtaining of a word vector space according to the word list comprises: Mapping all words in the word list to indexes to generate a dictionary; Input each word in the dictionary into the Word2Vec model and output the vector representation of the word; Generate a word vector space based on the vector representation of all words.

4. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 1, characterized in that: The extracting a feature vector from the word vector space comprises: The TF-IDF algorithm is used to extract feature vectors from the word vector space.

5. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 4 is characterized in that: The extracting feature vector from the word vector space using the TF-IDF algorithm includes: Calculate the word frequency and inverse document frequency represented by each word vector in the word vector space; Obtaining a weight represented by each word vector according to the word frequency and the inverse document frequency; All word vector representations are sorted in descending order according to the weights to obtain sorted word vector representations, and a preset number of word vector representations are selected from the sorted word vector representations in turn as feature vectors.

6. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 5, characterized in that: The calculating of the word frequency and the inverse document frequency represented by each word vector in the word vector space includes: In the formula, TF(t,d) represents the frequency of the word vector representing t, and f t,d Indicates the number of times the word vector t appears in the data set d, represents the total number of occurrences of the word vector representation in the dataset d, IDF(t,D) represents the inverse document frequency of the word vector representation t, N represents the total number of documents in the document set D, and |{d∈D:t∈d}| represents the number of documents containing the word vector representation t.

7. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 6 is characterized in that: The weight of each word vector representation obtained according to the word frequency and the inverse document frequency includes: TF-IDF(t,d,D)=TF(t,d)×IDF(t,D); Where TF-IDF(t,d,D) represents the weight of the word vector representation t.

8. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 1, characterized in that: The matching result obtained by matching the feature vector with the preset knowledge graph includes: Calculating the cosine similarity between each of the feature vectors and the abnormal information node and the normal information node; If the cosine similarity is greater than a preset value, it is determined that the feature vector and the node corresponding to the cosine similarity match, and a matching result is generated according to the matched feature vector and the node, wherein the node includes the abnormal information node or the normal information node.

9. The method for intelligent decision-making on operation and inspection of reactive power compensation equipment according to claim 1, characterized in that: The generating of the operation and inspection decision based on the abnormal information node includes: Determine a decision rule corresponding to the abnormal information node from a preset decision rule library; An operation and inspection decision is generated according to the decision rule.

10. A reactive power compensation equipment operation and inspection intelligent decision terminal, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, each step of the intelligent decision-making method for operation and inspection of reactive compensation equipment described in any one of claims 1 to 9 is implemented.

Images