Digital generation method, system and medium for substation maintenance task risk management

Abstract

The present invention discloses a digital generation method, system and medium for risk management of substation maintenance tasks, and belongs to the technical field of digital generation of risk management measures. The present invention is based on a text-enhanced risk factor causal relationship extraction method to obtain the causal relationship of the evolution of substation operation and maintenance task accidents, and construct a risk accident knowledge map for operation and maintenance tasks; according to the maintenance plan, key safety measures information is obtained, and a work ticket language rule base is established using a correlation analysis method; historical work tickets are obtained to establish a historical work ticket knowledge base; based on the risk accident knowledge map, the work ticket language rule base and the historical work ticket knowledge base, safety measures are automatically generated using a text similarity TF-IDF model. The present invention utilizes the risk accident knowledge map of operation and maintenance tasks, combined with the work ticket language rule base and the historical work ticket knowledge base, to automatically generate risk pre-control points for maintenance tasks, thereby improving the pertinence and scientific nature of the formulation of safety measures management and control plans.

Description

Technical Field

[0001] The present invention relates to the technical field of digital generation of risk management and control measures, and specifically to a digital generation method, system and medium for risk management and control of substation operation and maintenance tasks. Background Art

[0002] In order to further strengthen the risk control of production site operations, State Grid Corporation of China has formulated key measures and professional implementation rules for risk control of production site operations. In combination with the construction of three-level production control centers at the headquarters, provincial companies, and ultra-high voltage (municipal) companies, a "five-level, five-control" risk control system for production site operations has been established to continuously improve the safety level of production site operations. However, in the field operation control, relying on the operation risk classification table and the maintenance process risk library, the control of the entire operation process and the control of key links of high and medium risks often leads to risk classification errors, inadequate implementation of responsibilities and measures, lack of detailed requirements at all levels of professional management, and insufficient penetration of grassroots units. At the same time, there is a lack of prior personnel experience value analysis and violation record warnings for the uneven work leaders. Based on personal experience and maintenance procedures, they judge which components or connection terminals to disconnect, thereby manually generating safety measures tickets, with low reliability and automation levels. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art.

[0004] The present invention provides the following technical solution: a digital generation method for substation maintenance task risk management and control, comprising:

[0005] Construct a risk accident knowledge graph based on the causal relationship of the evolution of substation operation and maintenance tasks;

[0006] Establish a work ticket language rule library based on the operation and maintenance plan;

[0007] Obtain historical work tickets and establish a historical work ticket knowledge base;

[0008] Based on the risk accident knowledge graph, work ticket language rule base and historical work ticket knowledge base, safety measures are automatically generated using the text similarity TF-IDF calculation method.

[0009] Based on the text-enhanced risk factor causal relationship extraction method, the causal relationship of the accident evolution of substation operation and maintenance tasks is obtained.

[0010] The risk factor causal relationship extraction methods based on text enhancement include:

[0011] Use relevant texts in the power sector to expand and annotate the substation safety management text corpus;

[0012] The substation safety management text dataset expanded by the annotated corpus is subjected to text enhancement and encoding processing, and divided into a training set and a validation set;

[0013] Inputting the training set into the BiLSTM-CRF deep learning model for training to obtain a risk factor causal relationship extraction model;

[0014] The trained risk factor causal relationship extraction model is tested using the validation set, and the risk factor causal relationship extraction model is adjusted and optimized by comparing the gap between the labels predicted by the risk factor causal relationship extraction model and the actual labels to obtain the optimal risk factor causal relationship extraction model;

[0015] The optimal risk factor causal relationship extraction model is used to extract causal relationship nodes in safety management texts and obtain the causal relationship of accident evolution.

[0016] The BiLSTM-CRF deep learning model includes:

[0017] The feature text input layer expands each sentence segment input in each batch of the training set to the same length, and uses " <pad>"Tag completion, for words not in the dictionary, use" <unk>” label instead;

[0018] The feature text embedding layer maps all feature texts into a randomly initialized feature space to obtain the initial feature vector;

[0019] The bidirectional LSTM feature extraction layer is used to capture the contextual information of the security management text and extract high-dimensional features of the security management text;

[0020] The CRF causal node output layer adopts the CRF method as the loss function, uses the CRF output layer for back propagation, performs normalization at the sequence level, and realizes gradient descent.

[0021] The method for constructing the risk accident knowledge graph includes:

[0022] Preprocessing the causal relationship text obtained by the risk factor causal relationship extraction method;

[0023] Add text features and bigram features to the risk factor feature vector of the preprocessed causal text, and perform dimensionality reduction on the risk factor feature vector;

[0024] After generalizing the reduced-dimensional risk factor feature vectors using the AGNES algorithm, the same causal nodes in the risk factor causal relationships are extracted, and multiple causal relationship chains are cross-linked through common abstract risk factors to form a risk accident knowledge graph. The risk factor transfer probability in the substation operation and maintenance tasks is calculated and used as the causal transfer probability in the risk accident knowledge graph. The form of the risk accident knowledge graph is:

[0025] G = VE;

[0026] Among them, V is a set of points. In the knowledge graph G, each node represents a risk factor; E is a set of edges, representing the evolutionary causal relationship between risk factors.

[0027] The calculation formula for the risk factor transfer probability is:

[0028]

[0029] Among them, p(E j |E i ) represents the risk factor transfer probability, count(E i ,E j ) represents risk factor E i When it occurs, risk factor E j Frequency of occurrence; count(E i ,E k ) represents risk factor E i The total number of all possible risk factors E that may occur when k .

[0030] The method for establishing the work ticket language rule base includes:

[0031] Work tickets and safety tickets are generated from the operation and maintenance plan, and the association analysis method is used to establish the semantic description knowledge base and language rule base related to work tickets and safety tickets.

[0032] The automatic generation of safety measures includes: when a certain maintenance task is issued, the risk accident knowledge graph and the historical work ticket knowledge base are called in real time, and the text similarity TF-IDF calculation method is used to compare the information related to this task with the information in the risk accident knowledge graph and the historical work ticket knowledge base one by one, and the similarity between historical data and this task is calculated respectively. The top three historical work tasks with the highest familiarity are selected, and the work ticket and safety measure ticket are automatically generated through the work ticket language rule base, and the text familiarity is calculated again with the historical work ticket knowledge base. If the value is higher than the threshold α, it is considered that the safety measures content of the two work tasks is common, and the final generated content of the work ticket and safety measure ticket for this task is determined.

[0033] The text similarity TF-IDF calculation method is:

[0034] The TF-IDF method is used to calculate the TF-IDF scores of all words extracted from the work ticket text. The top m words with the highest scores are retained as the features of the work ticket. The work ticket text is converted into a weighted spatial vector, and the similarity of the text is measured by calculating the similarity between different vectors.

[0035] The TF-IDF method calculation formula is:

[0036]

[0037] Among them, w a,b Refers to the feature word a in text d b The weight in the text d is determined by the feature word a in the text d b The frequency of occurrence in tf a,b and inverse text frequency df a Calculated, N represents the text d b The total number of times the feature word appears.

[0038] The feature vector v(d b )for:

[0039] v(d b )=(W 1,b ,W 2,b ,…,W m,b );

[0040] Among them, W 1,b Indicates the first feature word in text d b The weight size in W 2,b Indicates that the second feature word is in text d b The weight size in W m,b Indicates the mth feature word in text d b The weight in , m represents the number of feature words.

[0041] The digital generation system for substation maintenance task risk management and control includes:

[0042] The knowledge graph module is used to build a risk accident knowledge graph based on the causal relationship of accident evolution in substation operation and maintenance tasks;

[0043] The rule base module is used to establish a work ticket language rule base based on the operation and maintenance plan;

[0044] Knowledge base module, obtains historical work tickets and establishes a historical work ticket knowledge base;

[0045] The generation module is used to automatically generate safety measures based on the risk accident knowledge graph, the work ticket language rule base and the historical work ticket knowledge base using the text similarity TF-IDF calculation method.

[0046] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 12.

[0047] During operation, the present invention utilizes the risk accident knowledge graph of operation and maintenance tasks, combines the work ticket language rule base and the historical work ticket knowledge base, and automatically generates risk pre-control points for maintenance tasks, thereby improving the pertinence and scientific nature of the formulation of safety measures and control plans.

[0048] Compared with the prior art, the present invention has the following beneficial effects:

[0049] 1. Based on the expansion of the substation safety management text corpus using relevant texts in the power field, this invention combines feature enhancement and word splicing methods with bidirectional long short-term memory networks and conditional random field algorithms to obtain more text semantic expressions and improve the feature extraction capability of the causal relationship extraction method. By extracting causal relationship nodes in the safety management text, the causal relationship of the accident evolution is obtained. This invention solves the problem of small sample learning in substation accident texts and reduces the adverse impact of insufficient effective text on deep learning methods.

[0050] 2. This invention incorporates text features and binary phrase features into the feature vector representation. Using the AGNES algorithm, it generalizes the concrete risk factors in the safety management text obtained through the risk factor causal relationship extraction method to obtain the general risk evolution causal laws and transfer probabilities in substation operation and maintenance tasks, thereby constructing a risk accident knowledge graph. This effectively improves the summarization and induction of the risk accident knowledge graph, mines more semantic information, addresses the limitations of conventional text generalization methods on text expression, reduces the impact of Chinese word segmentation errors on accident generalization results, and is more suitable for the complex and changeable language expression characteristics of safety management texts.

[0051] 3. This invention uses a risk and accident knowledge graph and a historical work ticket knowledge base as its data foundation, employing a text familiarity calculation method, association relationships, and a rule base to automatically generate work tickets and safety precautions. The automatically generated work tickets and safety precautions are then compared again with the historical work ticket knowledge base for text familiarity calculation to determine the final generated content for the current task's work ticket and safety precautions. This invention improves the efficiency and accuracy of the automatic generation of work tickets and safety precautions. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0053] Figure 1 This is a flowchart of the steps of the digital generation method of risk control measures for substation operation and maintenance tasks of the present invention;

[0054] Figure 2 This is a flow chart of the steps for constructing a risk accident knowledge graph for operation and maintenance tasks of the present invention;

[0055] Figure 3 This is a flowchart of the steps of the risk factor causal relationship extraction method based on text enhancement of the present invention;

[0056] Figure 4 This is a schematic diagram of the BiLSTM-CRF deep learning model structure in this embodiment;

[0057] Figure 5 This is a design diagram of the inference engine of this embodiment;

[0058] Figure 6 It is a flow chart of the automatic generation of safety measures of the present invention;

[0059] Figure 7 This is a technical flow chart of the security measures solution optimization technology of this embodiment. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. The described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] like Figure 1-7 As shown, the present invention provides the following technical solution: a digital generation method for substation maintenance task risk management and control, the steps of which include:

[0062] S1. Based on the text-enhanced risk factor causal relationship extraction method, the causal relationship of the accident evolution of substation operation and maintenance tasks is obtained, and the risk accident knowledge graph of operation and maintenance tasks is constructed;

[0063] Among them, the risk accident knowledge graph construction method for operation and maintenance tasks includes:

[0064] S11. Preprocessing the causal relationship text obtained by the risk factor causal relationship extraction method using conventional processing means to convert the messy text data into structured data that can be analyzed;

[0065] S12: adding text features and bigram features to the risk factor feature vector of the pre-processed causal relationship text, and performing dimensionality reduction on the risk factor feature vector.

[0066] S13. After generalizing the reduced-dimensional risk factor feature vector using the AGNES algorithm, the same causal nodes in the risk factor causal relationship are extracted, and multiple causal relationship chains are cross-linked through one or some common abstract risk factors to form a risk accident knowledge graph. The risk factor transfer probability in the substation operation and maintenance tasks is calculated, and the risk factor transfer probability is used as the causal transfer probability in the risk accident knowledge graph.

[0067] The calculation formula for the risk factor transfer probability is:

[0068]

[0069] Among them, p(E j |E i ) represents the risk factor transfer probability, count(E i ,E j ) represents risk factor E i When it occurs, risk factor E j Frequency of occurrence; count(E i ,E k ) represents risk factor E i The total number of all possible risk factors E that may occur when k .

[0070] The form of risk accident knowledge graph:

[0071] G = VE;

[0072] Where V is a set of points. In the knowledge graph G, each node represents a risk factor; E is a set of edges, representing the evolutionary causal relationship between risk factors.

[0073] Among them, the risk factor causal relationship extraction method based on text enhancement ( Figure 3 )include:

[0074] S111. Expand and annotate substation safety management texts using relevant texts in the power sector;

[0075] S112. Performing text enhancement and encoding processing on the substation safety management text dataset expanded from the annotated corpus, and dividing it into a training set and a validation set;

[0076] S113, inputting the training set into the BiLSTM-CRF deep learning model for training to obtain a risk factor causal relationship extraction model; BiLSTM-CRF deep learning model ( Figure 4 )include:

[0077] The feature text input layer expands each sentence segment input in each batch of the training set to the same length, and uses " <pad>"Tag completion, for words not in the dictionary, use" <unk>” label instead;

[0078] The feature text embedding layer maps all feature texts into a randomly initialized feature space to obtain the initial feature vector;

[0079] The bidirectional LSTM feature extraction layer consists of a two-layer LSTM network structure: a forward LSTM and a backward LSTM. It captures the contextual information of the security management text and extracts high-dimensional features of the security management text.

[0080] The CRF causal node output layer adopts the CRF method as the loss function, uses the CRF output layer for back propagation, performs normalization at the sequence level, and realizes gradient descent.

[0081] S114. Testing the trained risk factor causal relationship extraction model using the validation set, adjusting and optimizing the risk factor causal relationship extraction model by comparing the gap between the labels predicted by the risk factor causal relationship extraction model and the actual labels, to obtain an optimal risk factor causal relationship extraction model.

[0082] S115. Use the optimal risk factor causal relationship extraction model to extract causal relationship nodes in the safety management text to obtain the causal relationship of accident evolution.

[0083] Using the feature enhancement + sentence splicing method to enhance text and input more semantic feature information into the deep neural network can improve the effect of causal relationship extraction tasks.

[0084] S2. Based on the O&M plan, key safety measures information is obtained and a work ticket language rule base is established using correlation analysis. Specifically, work tickets and safety measures tickets are generated from typical O&M plans. The project details for both tickets are standardized and defined. Key safety measures information includes safety measures, work content, operational semantics, and location. This key safety measures information is extracted and analyzed for correlation. A knowledge base and language rule base for semantic descriptions of work tickets and safety measures tickets are established. Based on the characteristics of the work ticket generation rules, a corresponding inference engine module is built. This inference engine module utilizes custom SWRL rules. First, the on-site O&M task is matched against the knowledge base rules. If the similarity exceeds a preset threshold A, the rule is selected. If the similarity of multiple rules exceeds a preset threshold B, a conflict resolution mechanism is used to eliminate each rule one by one, retaining only the one with the highest similarity. This step is repeated until all attributes have been calculated.

[0085] Taking busbar maintenance as an example for analysis, the design of the loop rule base is shown in Tables 4-1 to 4-4:

[0086] Table 4-1: Power outage rules for outgoing circuit equipment

[0087]

[0088] Table 4-2: Circuit breaker stop rules

[0089]

[0090] Table 4-3: Rules for stopping line-side disconnectors

[0091]

[0092] Table 4-4: Rules for closing the grounding switch

[0093]

[0094] Taking busbar maintenance as an example for analysis, the inference engine maintenance module can be mainly divided into outgoing line circuit, main transformer equipment, busbar mother switch, reactive compensation equipment, mutual inductor, and its inference engine design can be seen in Figure 5 .

[0095] S3. Obtain historical work tickets and establish a historical work ticket knowledge base;

[0096] S4, based on the risk accident knowledge graph, work ticket language rule base and historical work ticket knowledge base, the text similarity TF-IDF model is used to automatically generate safety measures, specifically ( Figure 6 ):

[0097] When a maintenance task is issued, the risk accident knowledge graph and the historical work ticket knowledge base are called in real time, and the text similarity TF-IDF calculation method is used to compare the information related to this task with the information in the risk accident knowledge graph and the historical work ticket knowledge base one by one. The similarity between the historical data and the current task is calculated respectively, and the top three historical work tasks with the highest familiarity are selected. The work ticket language rule base is used to automatically generate the work ticket and the safety measure ticket, and the text familiarity is calculated again with the historical work ticket knowledge base. If the value is higher than the threshold α=90%, it is considered that the safety measures content of the two work tasks are common. These key content information constitute the final generated content of the work ticket and safety measure ticket for this task.

[0098] The text similarity TF-IDF calculation method is:

[0099] The TF-IDF method is used to calculate the TF-IDF scores of all words extracted from the work ticket text. The top m words with the highest scores are retained as the features of the work ticket. The work ticket text is converted into a weighted spatial vector. The similarity between different vectors is calculated to measure the similarity of the text. The TF-IDF method calculation formula is:

[0100]

[0101] Among them, w a,b Refers to the feature word a in text d b The weight in the text d is determined by the feature word a in the text d b The frequency of occurrence in tf a,b and inverse text frequency df a Calculated, N represents the text d b The total number of times the feature word appears.

[0102] The feature vector v(d b )for:

[0103] v(d b )=(W 1,b ,W 2,b ,…,W m,b );

[0104] Among them, W 1,b Indicates the first feature word in text d b The weight size in W 2,b Indicates that the second feature word is in text d b The weight size in W m,b Indicates the mth feature word in text d b The weight in m represents the number of feature words. According to the changes in the characteristic attributes of the maintenance task or the correction information of human feedback, the unreasonable contents of the initial safety measures plan generated by the above-mentioned text similarity TF-IDF model are optimized in a targeted manner. The disposal measures content in the optimized safety measures plan has a certain correlation with the characteristic attributes input later, so that the safety measures optimization has a certain scalability and adaptability, and according to the general format of the production rules, these attributes and the corresponding specific disposal measures are combined to achieve the optimization of the safety measures plan. Among them, the safety measures plan optimization process is as follows: Figure 7 shown.

[0105] The digital generation system for substation operation and maintenance task risk management and control includes:

[0106] The knowledge graph module is used to build a risk accident knowledge graph based on the causal relationship of accident evolution in substation operation and maintenance tasks;

[0107] The rule base module is used to establish a work ticket language rule base based on the operation and maintenance plan;

[0108] Knowledge base module, obtains historical work tickets and establishes a historical work ticket knowledge base;

[0109] The generation module is used to automatically generate safety measures based on the risk accident knowledge graph, the work ticket language rule base and the historical work ticket knowledge base using the text similarity TF-IDF calculation method.

[0110] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 12.

[0111] The present invention utilizes the risk accident knowledge graph of operation and maintenance tasks, combines the work ticket language rule base and the historical work ticket knowledge base, and automatically generates risk pre-control points for maintenance tasks, with a high degree of automation and high reliability.

[0112] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprise," "include," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0113] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.< / unk> < / pad> < / unk> < / pad>

Claims

1. A digital generation method for substation maintenance task risk management and control, characterized by: include: Construct a risk accident knowledge graph based on the causal relationship of the evolution of substation operation and maintenance task accidents; Establish a work ticket language rule library based on the operation and maintenance plan; According to the operation and maintenance plan, key safety measure information is obtained, and a work ticket language rule base is established using the correlation analysis method. Specifically, work tickets and safety measure tickets are generated from typical operation and maintenance plans, and the project details filled in the two tickets are standardized. Key safety measure information is extracted and analyzed for correlation. A semantic description knowledge base and language rule base related to work tickets and safety measure tickets are established. Based on the characteristics of the work ticket generation rules, a corresponding inference engine module is established. The inference engine module is constructed using customized SWRL rules. First, the on-site operation and maintenance tasks are matched with the rules in the work ticket language rule base. If the similarity is higher than the preset threshold A, it is selected. When the similarity of multiple rules is higher than the preset threshold B, a conflict elimination mechanism is used to eliminate each rule one by one, retaining only the one with the highest similarity. Repeat this step until all attributes have completed the corresponding calculation; Obtain historical work tickets and establish a historical work ticket knowledge base; Based on the risk accident knowledge graph, the work ticket language rule base and the historical work ticket knowledge base, the text similarity TF-IDF calculation method is used to automatically generate safety measures. The automatic generation of safety measures includes: when a certain maintenance task is issued, the risk accident knowledge graph and the historical work ticket knowledge base are called in real time, and the text similarity TF-IDF calculation method is used to compare the relevant information of this task with the information in the risk accident knowledge graph and the historical work ticket knowledge base one by one, and calculate the similarity between historical data and this task respectively, select the top three historical work tasks with the highest familiarity, and automatically generate work tickets and safety measures tickets through the work ticket language rule base, and perform text familiarity calculation again with the historical work ticket knowledge base. If the calculated text familiarity value is higher than the threshold α, it is considered that the safety measures content of the two work tasks are common, and the final generated content of the work ticket and safety measures ticket for this task is determined.

2. The digital generation method for substation maintenance task risk management and control according to claim 1 is characterized in that: Based on the text-enhanced risk factor causal relationship extraction method, the causal relationship of the accident evolution of substation operation and maintenance tasks is obtained.

3. The digital generation method for substation maintenance task risk management and control according to claim 2 is characterized in that: The risk factor causal relationship extraction methods based on text enhancement include: Use relevant texts in the power sector to expand and annotate the substation safety management text corpus; The substation safety management text dataset expanded by the annotated corpus is subjected to text enhancement and encoding processing, and divided into a training set and a validation set; Inputting the training set into the BiLSTM-CRF deep learning model for training to obtain a risk factor causal relationship extraction model; The trained risk factor causal relationship extraction model is tested using the validation set, and the risk factor causal relationship extraction model is adjusted and optimized by comparing the gap between the labels predicted by the risk factor causal relationship extraction model and the actual labels to obtain the optimal risk factor causal relationship extraction model; The optimal risk factor causal relationship extraction model is used to extract causal relationship nodes in safety management texts and obtain the causal relationship of accident evolution.

4. The digital generation method for substation maintenance task risk management and control according to claim 3 is characterized in that: The BiLSTM-CRF deep learning model includes: The feature text input layer expands each sentence segment input in each batch of the training set to the same length, and uses " <pad>"Tag completion, for words not in the dictionary, use" <unk> ” label instead;< / unk> < / pad> The feature text embedding layer maps all feature texts into a randomly initialized feature space to obtain the initial feature vector; The bidirectional LSTM feature extraction layer is used to capture the contextual information of the security management text and extract high-dimensional features of the security management text; The CRF causal node output layer adopts the CRF method as the loss function, uses the CRF output layer for back propagation, performs normalization at the sequence level, and realizes gradient descent.

5. The digital generation method for substation maintenance task risk management and control according to claim 2 is characterized in that: The method for constructing the risk accident knowledge graph includes: Preprocessing the causal relationship text obtained by using the risk factor causal relationship extraction method; Add text features and bigram features to the risk factor feature vector of the preprocessed causal text, and perform dimensionality reduction on the risk factor feature vector; After generalizing the reduced-dimensional risk factor feature vectors using the AGNES algorithm, the same causal nodes in the risk factor causal relationship are extracted, and multiple causal relationship chains are cross-linked through common abstract risk factors to form a risk accident knowledge graph. The risk factor transfer probability in substation operation and maintenance tasks is calculated and used as the causal transfer probability in the risk accident knowledge graph.

6. The digital generation method for substation maintenance task risk management and control according to claim 5 is characterized in that: The form of the risk accident knowledge graph is: G = VE; Among them, V is a set of points. In the knowledge graph G, each node represents a risk factor; E is a set of edges, representing the evolutionary causal relationship between risk factors.

7. The digital generation method for substation maintenance task risk management and control according to claim 5 is characterized in that: The calculation formula for the risk factor transfer probability is: Among them, p(E j |E i ) represents the risk factor transfer probability, count(E i ,E j ) represents risk factor E i When it occurs, risk factor E j Frequency of occurrence; count(E i ,E k ) represents risk factor E i The total number of all possible risk factors E that may occur when k .

8. The digital generation method for substation maintenance task risk management and control according to claim 1 is characterized in that: The method for establishing the work ticket language rule base includes: Work tickets and safety tickets are generated from the operation and maintenance plan, and the association analysis method is used to establish the semantic description knowledge base and language rule base related to work tickets and safety tickets.

9. The digital generation method for substation maintenance task risk management and control according to claim 8 is characterized in that: The text similarity TF-IDF calculation method is: The TF-IDF method is used to calculate the TF-IDF scores of all words extracted from the work ticket text. The top m words with the highest scores are retained as the features of the work ticket. The work ticket text is converted into a weighted spatial vector, and the similarity of the text is measured by calculating the similarity between different vectors.

10. The digital generation method for substation maintenance task risk management and control according to claim 8, characterized in that: The TF-IDF method calculation formula is: Among them, w a,b Refers to the feature word a in text d b The weight in the text d is determined by the feature word a in the text d b The frequency of occurrence in tf a,b and inverse text frequency df a Calculated, N represents the text d b The total number of times the feature word appears.

11. The digital generation method for substation maintenance task risk management and control according to claim 9 is characterized in that: The feature vector v(d b )for: v(d b )=(W 1,b ,W 2,b ,…,W m,b ); Among them, W 1,b Indicates the first feature word in text d b The weight size in W 2,b Indicates that the second feature word is in text d b The weight size in W m,b Indicates the mth feature word in text d b The weight in , m represents the number of feature words.

12. A digital generation system for substation maintenance task risk management and control, characterized by: include: The knowledge graph module is used to build a risk accident knowledge graph based on the causal relationship of accident evolution in substation operation and maintenance tasks; The rule base module is used to establish a work ticket language rule base based on the operation and maintenance plan; obtain key safety measure information based on the operation and maintenance plan, and establish a work ticket language rule base using the correlation analysis method. Specifically, work tickets and safety measure tickets are generated from typical operation and maintenance plans, and the project details filled in the two tickets are standardized and defined. The key safety measure information is extracted and analyzed for correlation, and a semantic description knowledge base and language rule base related to work tickets and safety measure tickets are established. According to the characteristics of the work ticket generation rules, a corresponding inference engine module is established; the inference engine module is constructed using customized SWRL rules. First, the on-site operation and maintenance tasks are matched with the rules in the work ticket language rule base. If the similarity is higher than the preset threshold A, it is selected; when the similarity of multiple rules is higher than the preset threshold B, the conflict elimination mechanism is used to eliminate each rule one by one, and only the one with the highest similarity is retained; Repeat this step until all attributes have completed the corresponding calculation; Knowledge base module, obtains historical work tickets and establishes a historical work ticket knowledge base; A generation module is used to automatically generate safety measures based on the risk accident knowledge graph, the work ticket language rule base and the historical work ticket knowledge base using the text similarity TF-IDF calculation method. The automatic generation of safety measures includes: when a certain maintenance task is issued, the risk accident knowledge graph and the historical work ticket knowledge base are called in real time, and the text similarity TF-IDF calculation method is used to compare the relevant information of this task with the information in the risk accident knowledge graph and the historical work ticket knowledge base one by one, calculate the similarity between historical data and this task respectively, select the top three historical work tasks with the highest similarity, and automatically generate work tickets and safety measures tickets through the work ticket language rule base, and perform text similarity calculation again with the historical work ticket knowledge base. If the calculated text similarity value is higher than the threshold α, it is considered that the safety measures content of the two work tasks is common, and the final generated content of the work ticket and safety measures ticket for this task is determined.

13. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 11 is implemented.

Images