Power transmission line reliability prediction method based on multi-mode cognitive network

Abstract

The invention discloses a power transmission line reliability prediction method based on a multi-modal cognitive network, and the method comprises the steps: firstly obtaining multi-source heterogeneous data including equipment ledgers, operation and maintenance logs, historical faults, meteorological information and geographic information, and extracting equipment-class, fault-class and environment-class entities and attributes thereof; and a unified three-layer meta-model is constructed through semantic similarity fusion. And then the association relationship and strength among various entities are identified, and a multi-modal cognitive network is constructed. On the basis, real-time state and degradation information of the equipment are extracted, and a dynamic failure rate prediction model driven by health indexes is established; environmental factors are further fused, a line-level overall reliability prediction model is constructed, and comprehensive assessment of the basic fault risk and the environmental risk of the power transmission line is achieved. The method has the advantages of high data fusion degree, high modeling precision and fine prediction granularity, and provides effective support for accurate operation and maintenance and risk early warning of power grid equipment.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent operation and maintenance of transmission lines, and particularly relates to a method for predicting the reliability of transmission lines based on a multi-modal cognitive network. Background Technique

[0002] As a core component of the power system, transmission lines are mainly composed of components such as towers, conductors, insulators, crossarms, jumper wires, and fittings. They are characterized by a wide distribution range, a large number of equipment, and a complex operating environment. According to statistics, about 75% of the unplanned power outages in the power grid are caused by transmission line failures, and the failure causes are mostly the accumulation of small and micro defects and the synergistic effect of external environmental risks. For example, the progressive accumulation of dirt on the surface of insulators may cause flashover, and the coupling effect of tower corrosion and strong wind loads may lead to structural instability. Sudden environmental events such as lightning strikes and icing will further increase the risk of cascading failures. However, the existing operation and maintenance technologies have the following bottlenecks: The multi-source data generated during the operation and maintenance process, such as equipment ledgers, inspection logs, and fault reports, show a high degree of heterogeneity, with structured data, semi-structured data, and unstructured text coexisting, and significant differences in space-time scales. Traditional methods rely on manual experience to screen key information, making it difficult to achieve the correlation analysis of equipment status, environmental risks, and historical faults, resulting in low data value density and weak decision-making support capabilities.

[0003] Existing reliability assessments are mostly based on periodic offline detections or single-sensor data, using static thresholds to determine the health status of equipment, lacking the ability to dynamically model the degradation trend of equipment, and even less able to quantify the time-varying impact of the external environment on the failure rate. The power grid renovation and maintenance strategies usually take the entire line as a unit and rely on the qualitative judgment of manual inspections, which are likely to cause two types of problems: over-maintenance: unnecessary replacement of equipment with good health status, resulting in waste of resources; maintenance lag: ignoring the real-time risks of hidden defects or environment-sensitive equipment, leading to the expansion of faults. Existing methods for quantifying environmental risks mostly rely on historical statistics and are not dynamically associated with the real-time status and topological structure of equipment, resulting in coarse risk warning granularity and poor timeliness.

[0004] In the prior art, Chinese Patent CN117131783A discloses a risk prediction model, method, and system for transmission lines based on multi-modal learning. The present invention proposes a method for constructing a risk prediction model for transmission lines based on multi-modal learning. First, a lightning intensity analysis module, a distance warning module, and a transmission line risk analysis module are respectively constructed. The lightning intensity analysis module predicts the lightning intensity at the next time point, the distance warning module is used to predict the lightning strike distance at the next time point, and the transmission line risk analysis module combines the lightning intensity, distance warning level, and transmission line performance at the next time point to predict the risk level of the transmission line being struck by lightning.

[0005] However, the technical solution still has the following deficiencies: First, although the method improves the prediction performance through multi-modal features, the lightning intensity analysis is only based on time series modeling, and fails to fully consider the spatial distribution characteristics of lightning activities and their dynamic evolution laws of propagation to adjacent areas. It lacks the description of potential spatial correlation risks between multiple lines and is difficult to meet the prediction requirements of lightning cross-region impacts in complex terrain areas. Second, the risk assessment module is implemented by a clustering method. Although it can achieve rapid classification, the clustering model is sensitive to data distribution and initial parameters and has weak generalization ability. Especially when facing multi-source heterogeneous data or extreme environmental changes, prediction deviations are likely to occur. In addition, each sub-module adopts a phased training strategy, and there is a lack of an end-to-end collaborative optimization mechanism between modules, resulting in problems of accuracy loss and error accumulation in the overall risk prediction.

[0006] Therefore, how to construct a reliability prediction model that integrates multi-source heterogeneous data and dynamically correlates device-environment-fault elements, realizes the leap from "local static assessment" to "global dynamic prediction", accurately identifies weak links in the line and generates risk-adaptive operation and maintenance strategies has become a key problem in improving the power supply reliability and operation and maintenance economy of the power grid. Summary of the Invention

[0007] The purpose of the present invention is to solve the above problems by providing a transmission line reliability prediction method that constructs a dynamic prediction model for the operation reliability of transmission lines by introducing a multi-modal cognitive network, calculates the overall reliability of the lines, identifies weak links in the line operation, and improves the digital operation and maintenance management level and reliability of transmission lines.

[0008] The purpose of the present invention can be achieved through the following technical solutions: The present invention provides a transmission line reliability prediction method based on a multi-modal cognitive network, including the following steps: Obtain multi-source heterogeneous data of the transmission line, including equipment inventory data, operation and maintenance logs, historical fault records, meteorological data, and geographical information data; Based on the multi-source heterogeneous data, extract the initial entities and their attribute information of the equipment category, fault category, and environment category; Perform similarity calculation and screening on the extracted entities of each category, merge entities with similar semantics, construct a three-layer meta-model framework including the equipment category, fault category, and environment category, and calculate the association strength between entities; Based on the three-layer meta-model framework and the association strength between entities, construct a multi-modal cognitive network; Use the multi-modal cognitive network to extract the real-time state parameters and degradation information of the equipment, and construct a device-level dynamic failure rate prediction model based on health indicators; Integrate the device-level dynamic failure rate model with environmental factors, calculate the line-level basic fault risk and environmental risk, and construct an overall reliability prediction model for transmission lines; Based on the overall reliability prediction model of the transmission line, realize the prediction of the reliability and risk of the transmission line.

[0009] Furthermore, based on the multi-source heterogeneous data, extract the initial entities and their attribute information of the device class, fault class, and environmental class, specifically including: Extract device class entity information from the device ledger data and construct a device class entity set , where represents the th type of device. The device class entities include devices such as conductors and poles. Each entity is set with attributes including the device service time , health status index , topological structure information , shape parameter β, scale parameter η, and the starting time of normal device operation τ 0 , where the health status index is the real-time detection value of the health indicators such as the real-time current, voltage, and temperature of each device ; Extract fault class entity information from historical fault records and operation and maintenance logs and construct a fault class entity set , where represents the j th type of fault. The fault class entities include fault types such as lightning strikes, icing, and equipment aging. The set attributes include the historical fault frequency and the device information associated with the fault ; Extract environmental class entity information from meteorological data and geographical information data and construct an environmental class entity set , where represents the k th type of environmental factor. The environmental class entities include environmental factors such as slope, wind speed, lightning intensity, and land use type. The set attributes include the real-time monitoring value of each environmental factor , t the number of environmental events occurring within the period.

[0010] Furthermore, calculate the similarity and screen the extracted entities of each type, merge the entities with similar semantics, and construct a three-layer meta-model framework including the device class, fault class, and environmental class, specifically including: Use the FastText model to vectorize the text descriptions of the device class, fault class, and environmental class entities to generate entity semantic vectors , d = 300 is the dimension of the entity semantic vector, which is used to enhance the semantic representation ability; Calculate the similarity of entity semantic vectors between entities of the same category based on the Jaccard similarity coefficient. The calculation formula of the Jaccard similarity coefficient is: In the formula: A and B respectively represent the semantic vectors of the entities to be matched; Set the similarity threshold , when Jaccard(A, B) ≥ , determine that A and B are semantically similar entities and merge them into a unified entity node; Classify the merged device, fault, and environment entities according to their entity attributes and categories to construct a three-layer meta-model framework including device, fault, and environment categories.

[0011] Furthermore, calculating the association strength between entities specifically includes: Based on the sets of device, fault, and environment entities, calculate the association strength between the three types of entities. The calculation formula of the association strength is: Among them, represents the association strength among the device entity , the fault entity and the environment entity , respectively represent the cumulative occurrence times of the device entity , the fault entity and the environment entity identified and recorded in the multi-source heterogeneous data, respectively represent the number of times the device entity and the fault entity , the device entity and the environment entity , the environment entity and the fault entity appear together in the same data in the multi-source heterogeneous data; represents the number of times the device entity and the fault entity and the environment entity appear together in the same data in the multi-source heterogeneous data.

[0012] Furthermore, constructing a multi-modal cognitive network based on the three-layer meta-model framework and the association strength between entities specifically includes: Map the entity classes of equipment, faults, and environment in the three-layer meta-model framework to nodes in the multi-modal cognitive network, and the node attributes are entity attributes; Construct the edges between nodes according to the calculated association relationships and strengths between entities, and the edge attributes are the association strengths between the node entities; Use the Neo4j graph database for the structured storage of the multi-modal cognitive network, and the relationships between nodes and edges are organized in the form of a graph structure.

[0013] Furthermore, use the multi-modal cognitive network to extract the real-time status parameters and degradation information of the equipment, and construct a device-level dynamic failure rate prediction model based on health indicators, specifically including: Quantify the impacts of equipment aging and external environmental factors on the failure rate through the improved Weibull-Markov chain algorithm to obtain the equipment aging correction coefficient and the external environment correction coefficient; According to the real-time detection values of the equipment , use the health status indicators to evaluate the operating health status of the equipment. The calculation formula of the health index is: where is the detection value of the i th indicator at time t , and the indicators include health indicators such as the real-time current, voltage, and temperature of the equipment; is the weight of the i th health indicator, which is determined by the analytic hierarchy process; N is the number of real-time detection values, , are the threshold ranges of the real-time monitoring values; Based on the entity classes of equipment, faults, and environment and their attribute information extracted from the multi-modal cognitive network, combined with the equipment aging correction coefficient, the external environment correction coefficient, and the health index, construct a device-level time-varying failure rate prediction model.

[0014] Furthermore, the formula for the equipment aging correction coefficient is: where is the equipment aging correction coefficient of the equipment entity at the current time t, , are the shape parameter and scale parameter of the equipment entity respectively, is the current time, is the equipment entity starting time of normal operation; The calculation formula for the external environment correction coefficient is as follows: Wherein, represents the external environment correction coefficient of the equipment entity ; represents the number of occurrences of the k-th type of environmental entity associated with the equipment entity ; is the total number of occurrences of all environmental entities related to the equipment entity ; is the total number of environmental entities considered, represents the influence weight of the environmental entity on the equipment, which is determined according to expert knowledge.

[0015] Furthermore, the equipment-level time-varying failure rate prediction model is: Wherein, is the dynamic failure rate of the equipment entity at the current time t, is the reference failure rate, that is, the initial failure rate without degradation or environmental influence, is the set of failure type entities, is the set of environmental type entities, , are the association strengths between the equipment entity and the failure entity , and between the equipment entity and the environmental entity respectively, is the number of occurrences of the failure entity t within the time t is the total number of occurrences of all failure entities within the time is the real-time monitoring value of the environmental entity at the time t.

[0016] Furthermore, fusing the equipment-level dynamic failure rate model with environmental factors, calculating the line-level basic fault risk and environmental risk, and constructing the overall reliability prediction model of the transmission line specifically include: Based on the topological association relationship between equipment entities in the multi-modal cognitive network, according to the criticality weight of the equipment in the topological structure, aggregating the dynamic failure rate of each equipment entity into the line-level basic fault risk , and the calculation formula is: Wherein, is the device entity at the current moment t is the dynamic failure rate of is the device entity is the critical weight in the line structure, determined based on the device redundancy and load distribution ratio parameters. m is the total number of device entities in the transmission line. is the line-level basic fault risk at the current moment t; Utilize the association strength between the device and the environmental entity and the weight factor of each environmental factor to construct a device-level environmental risk model: Among them, is the device-level environmental risk of the device entity ; represents the weight factor of the environmental entity on the device, determined according to expert knowledge. represents the device entity and the environmental entity ; represents the environmental entity is the real-time monitoring value at the current moment t. p is the total number of environmental entities; Fuse the line-level basic fault risk and the device-level environmental risk to construct an overall reliability prediction model for the transmission line. Its comprehensive risk expression is: Among them, is the overall comprehensive risk of the transmission line at the moment t . m is the total number of device entities.

[0017] Furthermore, the weight factor of each environmental factor is determined by the entropy weight method. The formula is: Among them, is the information entropy of the environmental entity ; is the environmental entity at the moment is the real-time monitoring value. M is the number of samples, that is, the total number of time segments or monitoring points for statistical analysis. is the sum of all sample values of the environmental entity .

[0018] Compared with the prior art, the present invention has the following advantages: (1) The present invention introduces a multi-modal cognitive network, integrates multi-source heterogeneous data such as fault records and equipment ledgers to construct a multi-modal cognitive network for transmission lines, analyzes the fault principles and mutual correlations of each component, and then constructs a prediction model for the failure rate of transmission line equipment and a dynamic prediction model for operation reliability based on the multi-modal cognitive network, conducts comprehensive analysis on the whole, and provides support for the safe and stable operation of the power grid; it can be widely applied to the daily operation and maintenance management of transmission lines.

[0019] (2) The present invention realizes the unified modeling of multi-source heterogeneous data in transmission lines by constructing a knowledge graph structure that integrates three layers of multi-modal entities: equipment type, fault type, and environment type. Based on structured equipment ledger data, unstructured inspection texts, and image information, this graph uses entity recognition and relationship extraction methods to establish semantic connections such as hypernymy, association, and causality between entities, thus breaking the pattern of traditional operation and maintenance data silos and constructing a cognitive network for transmission lines with semantic relevance and temporal logic. This design effectively solves the technical bottlenecks in traditional methods where it is difficult to integrate different data modalities and difficult to conduct correlation analysis between equipment status and environmental risks, providing a unified data foundation and semantic support for subsequent risk modeling.

[0020] (3) The present invention proposes an improved Weibull-Markov chain combined modeling method to dynamically model the evolution process of the health state of key equipment in transmission lines. Compared with traditional static threshold determination methods, this method integrates the statistical distribution law of equipment life and the time series characteristics of state migration, can reflect the gradual degradation process of equipment from health to failure in real time, and introduces environmental variables such as temperature, humidity, ice coating, and lightning strike frequency as adjustment factors for state transition, thereby dynamically adjusting the fault probability estimation. This modeling technology effectively improves the early warning ability for potential equipment failures and avoids the problem of missed detection of hidden defects by the periodic detection mechanism.

[0021] (4) The present invention constructs a comprehensive risk assessment method for transmission lines based on the graph network topology structure, weights and fuses the equipment health state nodes, fault event nodes, and environmental risk nodes to form a risk propagation graph with dual spatial and semantic connection attributes. During the process of quantifying line risks, the system comprehensively calculates the risk level of the entire line and the critical fault propagation path according to the state values, connection strengths, and influence path weights of each node. This technical solution effectively overcomes the limitations of traditional methods that use equipment or lines as evaluation units and lack multi-factor fusion, improves the granularity and accuracy of risk assessment, and realizes the "point-line-plane" linkage intelligent early warning.

[0022] (5) By introducing a graph-driven fault deduction mechanism, the present invention can automatically trace the potential causes and related environmental factors after identifying the abnormal state of the device, and realize the automatic construction of the fault causal chain. Combining with the results of dynamic risk assessment, the system can predict the trend of risk events that may occur within a certain future time window and generate intervention suggestions. This mechanism avoids the lag and subjectivity of traditional manual review, and enhances the forward-looking and scientific nature of the operation and maintenance strategy.

[0023] (6) The present invention supports the self-learning and knowledge update of the system through a graph-level knowledge evolution mechanism. When new device types, new fault forms, or new environmental interference factors are connected, the system can automatically update the graph structure and state evaluation model through incremental learning, maintaining the timeliness and robustness of the knowledge base, solving the stubborn problem of the existing system of "static rules - fixed model - lagged response", and improving the response ability and adaptability to emergencies. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flowchart of the method of the present invention; Figure 2 is a lightning strike distribution map of transmission towers with different slopes; Figure 3 is a lightning strike distribution map of transmission towers with different aspects; Figure 4 is a lightning strike distribution map of transmission towers with different heights; Figure 5 is a lightning strike distribution map of transmission towers located in different land types; Figure 6 is a lightning strike distribution map of transmission towers with different river network densities. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] Embodiment 1: This embodiment relates to a method for dynamically predicting the reliability of a transmission line based on a multi-modal cognitive network. As Figure 1 shown, the method includes the following steps: Step 1: Collect multi-source heterogeneous data of the transmission line and construct a multi-modal cognitive network of the transmission line.

[0027] Taking the actual data of a local power transmission network as an example, a multi-modal cognitive network for power transmission lines is constructed. This multi-modal cognitive network covers 122 lines, including 10,350 nodes and 15,678 relationships.

[0028] Obtain multi-source heterogeneous data of power transmission lines, including equipment inventory data, operation and maintenance logs, historical fault records, meteorological data, and geographic information data; Based on the multi-source heterogeneous data, extract the initial entities and their attribute information of equipment category, fault category, and environment category, specifically including: Extract equipment category entity information from equipment inventory data and construct an equipment category entity set , where represents the category of equipment. The equipment category entities include equipment such as conductors and towers. Each entity is set with attributes including equipment service time , health status indicator , topological structure information , shape parameter β, scale parameter η, and the starting time of normal equipment operation τ 0 , where the health status indicator is the real-time detection value of health indicators such as real-time current, voltage, and temperature of each equipment ; Extract fault category entity information from historical fault records and operation and maintenance logs and construct a fault category entity set , where represents the j category of faults. The fault category entities include fault types such as lightning strikes, icing, and equipment aging. The set attributes include historical fault frequency and equipment information associated with the fault ; Extract environment category entity information from meteorological data and geographic information data and construct an environment category entity set , where represents the k category of environmental factors. The environment category entities include environmental factors such as slope, wind speed, lightning intensity, and land use type. The set attributes include the real-time monitoring values of each environmental factor , t the number of environmental events occurring within a period.

[0029] Calculate the similarity and screen the extracted entities of each category, merge entities with similar semantics, construct a three-layer meta-model framework including equipment category, fault category, and environment category, and identify the association relationships and association strengths between entities, specifically including: Vectorize the text descriptions of device, fault, and environment entities using the FastText model to generate entity semantic vectors , where d = 300 is the dimension of the entity semantic vector, used to enhance the semantic representation ability; Calculate the similarity of the semantic vectors between various entities based on the Jaccard similarity coefficient. The calculation formula of the Jaccard similarity coefficient is: In the formula: A 、 B respectively represent the semantic vectors of the entities to be matched; Set the similarity threshold , when Jaccard(A, B) ≥ , determine that A and B are semantically similar entities and merge them into a unified entity node; Structurally divide the merged device, fault, and environment entities according to their entity attributes and categories to construct a three-layer meta-model framework including device, fault, and environment categories.

[0030] Construct a multimodal cognitive network based on the three-layer meta-model framework and the association relationships and strengths between entities; Based on the sets of device, fault, and environment entities, identify the association relationships between entities and calculate the association strengths between the three types of entities. The calculation formula of the association strength is: Among them, represents the device entity , the fault entity and the environment entity The association strength among the three, respectively represent the device entity , the fault entity and the environment entity The cumulative number of occurrences identified and recorded in multi-source heterogeneous data, respectively represent the device entity and the fault entity , the device entity and the environment entity , the environment entity and the fault entity The number of times they co-occur in the same data in multi-source heterogeneous data; represents the device entity and the fault entity and the environment entity The number of times they co-occur in the same data in multi-source heterogeneous data.

[0031] Map the device class, fault class, and environment class entities in the three-layer meta-model framework to nodes in the multi-modal cognitive network, and the node attributes are the entity attributes; Construct the edges between the nodes according to the calculated association relationships and association strengths between the entities, and the attributes of the edges are the association strengths between the node entities; Use the Neo4j graph database for the structured storage of the multi-modal cognitive network, and organize the relationships between the nodes and edges in the form of a graph structure.

[0032] Step 1 mainly solves the problems of difficult fusion, inconsistent semantics, and insufficient association modeling among multi-source heterogeneous data of transmission lines. When facing multi-source data such as equipment ledgers, fault records, operation and maintenance logs, meteorological information, and geographical data, traditional methods are often difficult to manage uniformly due to inconsistent data structures and diverse description methods, resulting in fragmented information, serious redundancy, and affecting the accuracy and efficiency of subsequent analysis and fault diagnosis. Through the method of the present invention, first systematically extract the device class, fault class, and environment class entities and their attribute information from multi-source data, combine the FastText model to vectorize the entity text semantics, and use the Jaccard similarity coefficient for similarity calculation and entity merging, thus solving the redundancy and ambiguity problems caused by inconsistent semantic expressions and improving the accuracy of entity recognition and fusion. Further, propose an association strength calculation method based on the co-occurrence frequency of entities, quantify the coupling relationship between the three types of entities of equipment, faults, and the environment, effectively establish a multi-factor influence chain, and realize the association modeling and in-depth mining of complex fault scenarios. Finally, map the three types of entities to nodes in the multi-modal cognitive network, and construct the strong association relationships between the entities as graph structure edges to achieve structured and semantic knowledge expression. The entire network is efficiently stored and queried in the Neo4j graph database, with good scalability and visualization capabilities, effectively supporting applications such as knowledge reasoning, anomaly recognition, and risk warning in the intelligent operation and maintenance system, and improving the cognitive ability and operation intelligence level of the system.

[0033] Step 2: Use the constructed multi-modal cognitive network to construct a time-varying failure rate prediction model for transmission equipment and obtain the prediction results. The specific steps include: Through the improved Weibull-Markov chain algorithm, quantify the dynamic influence weights of key factors such as equipment aging, lightning strikes, and strong winds, and correct the dynamic weights. The calculation formula for the equipment aging correction coefficient is: Among them, is the equipment entity The equipment aging correction coefficient at the current moment t, , are respectively the shape parameter and scale parameter of the equipment entity , is the current time, is the device entity starting time of normal operation; The calculation formula for the external environment correction coefficient is: where, represents the external environment correction coefficient of the device entity , represents the number of occurrences of the k-th type of environmental entity associated with the device entity , is the total number of occurrences of all environmental entities related to the device entity , is the total number of environmental entities considered, represents the influence weight of the environmental entity on the device, determined according to expert knowledge.

[0034] The obtained weight situation is shown in Table 1. In the table respectively represent transmission lines, fittings, insulators, iron parts, tower poles, guy wires, respectively represent the weights of device aging, lightning strike, strong wind, external foreign objects, tree obstacles and other fault influencing factors.

[0035] Table 1 Fault influencing factor weights (2) Extract device health indicators through a multi-modal cognitive network, determine indicator weights using the analytic hierarchy process, and calculate the health index, specifically: The calculation formula for the device health index (HI) is: where, is the detection value of the i th item of indicator at time t , and the indicators include health indicators such as the real-time current, voltage, and temperature of the device; is the weight of the i th health indicator, determined through the analytic hierarchy process; N is the number of real-time detection values, , are the threshold ranges of the real-time monitoring values; Compare the predicted value of the device failure rate based on the fault influencing factors with the historical average failure rate, as shown in Table 2.

[0036] Table 2 Comparison of different device failure rates As can be seen from the table, there are significant differences in the failure rates of some equipment. For example, for transmission lines, it may be due to lightning strikes and weather conditions such as wind and rain that affect the real-time failure rate of the equipment, while the historical average failure rate is difficult to reflect such differences.

[0037] (3)Based on the historical failure data and real-time status parameters of equipment in the multi-modal cognitive network, combined with the dynamic weight correction coefficient and health index, construct a time-varying failure rate prediction model for equipment: Among them, is the dynamic failure rate of the equipment entity at the current time t ; is the reference failure rate, that is, the initial failure rate without degradation or environmental influence, is the set of failure type entities, is the set of environmental entities, , are the association strengths between the equipment entity and the failure entity , and between the equipment entity and the environmental entity respectively, is the number of occurrences of the failure entity t within the time , is the total number of occurrences of all failure entities within the time t , is the real-time monitoring value of the environmental entity at time t.

[0038] Input the query data of the multi-modal cognitive network into the health index evaluation model to calculate the health index and real-time failure rate of each transmission equipment, as shown in Table 3. Compared with the failure rate of the failure factor correction model and the historical average failure rate, the improved prediction model of the failure rate based on the health index can better reflect the differences between different transmission equipment.

[0039] Table 3 Real-time Failure Rate Prediction Model of Transmission Equipment Based on HI Step 2 mainly solves the problem that the traditional failure rate evaluation model of transmission equipment is difficult to effectively describe the equipment state change, external environmental interference and the superposition effect of failure factors. Previous methods usually estimate the equipment failure rate based on static parameters or empirical models, ignoring the time-varying characteristics of the equipment operation state and the complex coupling relationship among multiple influencing factors, with limited prediction accuracy and difficult to meet the requirements of dynamic perception and accurate assessment of the equipment health state in the actual power transmission system.

[0040] Through the constructed multi-modal cognitive network, this invention organizes three types of entities, namely equipment, faults, and environment, and their association strengths into a structured knowledge graph, providing a data basis and semantic support for subsequent dynamic modeling. On this basis, an improved Weibull-Markov chain algorithm is used to model the equipment aging process, and the aging correction coefficient is calculated using the shape parameter and scale parameter of the equipment to accurately quantify the impact of the equipment operation time on its health status. At the same time, by combining the environmental factors extracted from multi-source heterogeneous data, an external environment correction coefficient model is constructed, and expert knowledge is introduced to set the environmental weight, thereby quantitatively expressing the contributions of factors such as lightning strikes, strong winds, and tree obstacles to equipment risks.

[0041] By integrating equipment operation parameters, such as real-time current, voltage, and temperature, a health index model is constructed, and the analytic hierarchy process is introduced to assign reasonable weights to each index, making the health assessment process scientific and interpretable. Further, by integrating historical fault data, health index, and dynamic correction coefficient, a time-varying failure rate prediction model at the equipment level is constructed, which can dynamically reflect the health level and potential fault risks of the equipment at different time points.

[0042] Compared with the traditional average model, this method can achieve a more fine-grained and personalized prediction effect, accurately distinguish different types and states of transmission equipment, and effectively support the optimal allocation of operation and maintenance resources and the early triggering of risk warnings. Its significant advantage lies in integrating static structural features and dynamic operation information, comprehensively considering historical and real-time factors, and fully reflecting the timeliness, comprehensiveness, and intelligence level of the health status assessment of transmission equipment.

[0043] Step 3: Use the obtained multi-modal cognitive network to predict the lightning strike risk of transmission lines and obtain the prediction results. The specific steps are as follows: (1) Through the analysis of the multi-modal cognitive network, the slope, aspect, land use type, river network density, and tower height are selected as the main influencing factors for the transmission line to be struck by lightning.

[0044] (2) Determine the influence factor weights through the entropy weight method to accurately judge the influence of specific variables on the overall evaluation. The formula for determining the influence factor weights by the entropy weight method is: Where is the information entropy of the environmental entity , is the real-time monitoring value of the environmental entity at time , M is the number of samples, that is, the total number of time segments or monitoring points used for statistical analysis, is the sum of all sample values of the environmental entity .

[0045] (3) Based on the association strength in the multimodal cognitive network , calculate the environmental risk. The formula for environmental risk is: Among them, is the device-level environmental risk of the device entity . represents the weight factor of the environmental entity on the device, which is determined according to expert knowledge. represents the association strength between the device entity and the environmental entity . represents the real-time monitoring value of the environmental entity at the current moment t. p is the total number of environmental entities; (4) Based on the device topology association relationship in the multimodal cognitive network, aggregate the device-level real-time failure rate into the line-level basic fault risk, and dynamically couple the line-level basic fault risk with the environmental risk to generate a line overall lightning strike risk warning model.

[0046] The formula for aggregating the device-level real-time failure rate into the line-level basic fault risk is: Among them, is the dynamic failure rate of the device entity at the current moment t. is the criticality weight of the device entity in the line structure, which is determined based on the device redundancy and load distribution ratio parameters. m is the total number of device entities in the transmission line. is the line-level basic fault risk at the current moment t; Dynamically couple the line-level basic fault risk with the environmental risk factor to construct a comprehensive lightning strike risk prediction model: Among them, is the overall comprehensive risk of the transmission line at time t . m is the total number of device entities.

[0047] (5) Analyze the risk influencing factors in the embodiment The correlation analysis between lightning strikes and the slope of the tower is as Figure 2 shown. As the slope increases, the proportion of being struck by lightning also gradually increases. Therefore, it can be concluded that there is a positive correlation between the slope of the tower and the probability of being struck by lightning.

[0048] The correlation analysis between lightning strikes and the aspect of the tower is as Figure 3As shown in the figure, slope aspect data is queried through a multi-modal cognitive network, and the proportion of lightning strikes on transmission towers in different slope aspects is calculated. The proportion of transmission towers with a slope aspect of southeast suffering from lightning strikes is significantly higher than that in other directions, about 0.051%; followed by the transmission towers with slope aspects of east and south, and the percentage of lightning trip of transmission towers is about 0.048%, ranking second; the flat ground with the lowest percentage of lightning trip of transmission towers is about 0.019%. Summer is a period with frequent lightning trip accidents in the power grid. In summer, the main winds in the southeast coastal area of China are easterly and southeasterly. Therefore, the mountainous and hilly areas due east or southeast are windward slopes, where airflows are easily lifted, resulting in strong convective weather and more prone to lightning hazards.

[0049] Analysis of the correlation between lightning strikes and the height of transmission towers is as Figure 4 shown. Through the equipment ledger network, the height data of transmission towers is queried, and the relationship between different transmission tower heights and lightning strikes on transmission towers is statistically calculated. As the height of the transmission tower increases, the probability of being struck by lightning also increases. This may be because the inductance of the tower body increases, resulting in an increase in lightning overvoltage, while the lightning protection measures for transmission lines are limited.

[0050] Analysis of the correlation between lightning strikes and land type is as Figure 5 shown. The percentage of lightning trip transmission towers for each land use type is calculated respectively. Transmission towers in the land use type of forest area have the highest probability of being struck by lightning, followed by wetlands, farmlands, grasslands, and the remaining land types have a lower probability of being struck by lightning. Because forest areas are usually located in high-altitude areas and have higher humidity, they are more prone to lightning strikes.

[0051] Analysis of the correlation between lightning strikes and river network density is as Figure 6 shown. By calculating the river network density within the grid, the overall river network density distribution map of the target area can be generated. The greater the river network density, the greater the probability of transmission towers being struck by lightning. On the one hand, a large river network density indicates that the proportion of river area in the region is larger and the water vapor conditions in the air are better, so the lightning strike probability is higher; on the other hand, the conductivity of rivers is much higher than that of flat ground, so lightning is more likely to occur.

[0052] (6) Evaluate the risk level of the embodiment to obtain the lightning strike risk prediction result of the transmission line, and compare it with the actual data to verify the effectiveness of this solution.

[0053] According to the foregoing analysis, in accordance with the degree of association with the probability of lightning strikes on the line, each influencing factor is sorted, in order of river network density, tower height, slope, slope aspect, and land type, and their numbers are respectively defined as 、 、 、 、 , and a lightning strike association matrix is established, as shown in Table 4.

[0054] Table 4 Association matrix of lightning strike influencing factors Through calculation, the maximum eigenvalue of the correlation matrix of the lightning strike impact factor is 5.05, and the consistency ratio is 0.019 < 0.10, indicating that the consistency of the matrix is reasonable. Its weight is determined by the entropy weight theory.

[0055] Finally, the final evaluation results are shown in Table 5. As can be seen from the table, the possibility that the transmission line is generally damaged by lightning is the highest, at 0.37. The possibilities of being damaged by lightning very highly and very lowly are relatively small, at 0.04 and 0.06 respectively, and the uncertainty is 0. Therefore, based on the overall evaluation results, the probability that the transmission line is damaged by lightning is general.

[0056] Table 5 Lightning damage risk assessment results of transmission lines All transmission lines in the embodiment are counted and compared with the actual data. The statistical data are shown in Table 6. As can be seen from the table, the prediction results are basically consistent with the actual situation. As the lightning damage risk level increases, the number of lightning damages decreases significantly, and the number of lightning damages at different risk levels is also basically the same. Therefore, the method proposed in this solution can effectively predict the probability of transmission lines being struck by lightning.

[0057] Table 6 Comparison between lightning damage prediction and actual data In summary, by applying the dynamic prediction method for the reliability of transmission lines based on a multi-modal cognitive network proposed in the present invention, the weak links of the lines can be accurately identified, and the operation and maintenance decision-making efficiency and the power supply reliability of the power grid can be improved.

[0058] Example 2: This embodiment provides a transmission line reliability and risk prediction system based on a multi-modal cognitive network, including: A data acquisition module, which is used to obtain structured equipment ledger data, sensor monitoring data, as well as unstructured inspection images, text logs, and environmental monitoring data from transmission line-related equipment, so as to achieve comprehensive acquisition of multi-source heterogeneous data; A knowledge graph construction module, which is used to preprocess, entity identify, relationship extract, and fuse the collected data, construct a multi-modal cognitive graph including equipment entities, fault entities, and environmental entities, and form a graph structure with semantic associations and temporal logics; A semantic fusion and reasoning module, which is used to achieve semantic aggregation between different modal data based on the word vector model and graph neural network technology, mine potential relationships such as causality, common cause, and evolution between entities, and support knowledge reasoning based on rules or probabilities; The device status evaluation module is used to dynamically evaluate the health status of key devices based on the improved Weibull-Markov model, and to achieve quantitative prediction of the device degradation trend by combining the operation history and environmental factors; The risk prediction module is used to construct a risk propagation network by combining the entity status information, graph topology structure and historical evolution path in the cognitive map structure, and to achieve quantitative evaluation and prediction warning of potential risks; The auxiliary decision-making module is used to integrate the device status evaluation results and risk prediction results to form a line-level comprehensive reliability analysis report, and to provide maintenance scheduling suggestions, resource optimization configuration plans and operation and maintenance intervention priority rankings; The knowledge evolution module is used to realize the incremental update of the graph structure and evaluation model according to the newly accessed data, support the continuous evolution of the knowledge system, and improve the adaptability and intelligence level of the system in the face of new devices, new fault modes and emergencies.

[0059] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disks or optical discs that can store program codes.

[0060] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. A transmission line reliability prediction method based on a multimodal cognitive network, characterized in that, It includes the following steps: Obtain multi-source heterogeneous data of the transmission line, including equipment inventory data, operation and maintenance logs, historical fault records, meteorological data, and geographical information data; Based on the multi-source heterogeneous data, extract the initial entities and their attribute information of the equipment category, fault category, and environment category; Perform similarity calculation and screening on the extracted entities of various categories, merge entities with similar semantics, construct a three-layer meta-model framework including the equipment category, fault category, and environment category, and calculate the association strength between entities; Based on the three-layer meta-model framework and the association strength between entities, construct a multi-modal cognitive network; Use the multi-modal cognitive network to extract the real-time status parameters and degradation information of the equipment, and construct an equipment-level dynamic failure rate prediction model based on health indicators; Fuse the equipment-level dynamic failure rate model with environmental factors, calculate the line-level basic fault risk and environmental risk, and construct an overall reliability prediction model for the transmission line; Based on the overall reliability prediction model of the transmission line, realize the prediction of the reliability and risk of the transmission line.

2. The reliability prediction method for a transmission line based on a multi-modal cognitive network according to claim 1, wherein, The extracting the initial entities and their attribute information of the equipment category, fault category, and environment category based on the multi-source heterogeneous data specifically includes: Extract device - type entity information from the device ledger data to construct a set of device - type entities , where represents the th type of device. The device - type entities include devices such as conductors and poles. Each entity has set attributes including the service time of the device , health - status indicators , topological - structure information , shape parameter β, scale parameter η, and the starting time of normal device operation τ 0 . Among them, the health - status indicator is the real - time detection value of health indicators such as the real - time current, voltage, and temperature of each device ; Extract fault class entity information from historical fault records and operation and maintenance logs to construct a fault class entity set , where represents the j th type of fault. The fault class entities include fault types such as lightning strikes, icing, and equipment aging. The set attributes include historical fault frequency and equipment information associated with the fault ; Extract environmental entity information from meteorological data and geographical information data to construct an environmental entity set , where represents the k th type of environmental factor. The environmental entities include environmental factors such as slope, wind speed, lightning intensity, and land use type. The set attributes include the real-time monitoring values of each environmental factor , t the number of environmental events occurring within a period.

3. A transmission line reliability prediction method based on a multi-modal cognitive network according to claim 1, characterized in that, The performing similarity calculation and screening on the extracted entities of various categories, merging entities with similar semantics, and constructing a three-layer meta-model framework including the equipment category, fault category, and environment category specifically includes: Use the FastText model to vectorize the text descriptions of device, fault, and environment entities, generating entity semantic vectors , where d = 300 is the dimension of the entity semantic vector, used to enhance the semantic representation ability; Calculate the similarity of the entity semantic vectors between entities of the same category based on the Jaccard similarity coefficient. The calculation formula of the Jaccard similarity coefficient is: Wherein: A and B respectively represent the semantic vectors of the entities to be matched; Set the similarity threshold , when Jaccard(A, B) ≥ , it is determined that A and B are semantically similar entities and merged into a unified entity node; Divide the merged entities of the equipment category, fault category, and environment category according to the entity attributes and their respective categories, and construct a three-layer meta-model framework including the equipment category, fault category, and environment category.

4. A method for predicting the reliability of a transmission line based on a multimodal cognitive network according to claim 1, characterized in that, The calculating the association strength between entities specifically includes: Based on the entity sets of the equipment category, fault category, and environment category, calculate the association strength between the three categories of entities. The calculation formula of the association strength is: Among them, represents the association strength among device entity , fault entity and environment entity ; respectively represents the cumulative occurrence times of device entity , fault entity and environment entity identified and recorded in multi-source heterogeneous data; respectively represents the co-occurrence times of device entity and fault entity , device entity and environment entity , environment entity and fault entity in the same data of multi-source heterogeneous data; represents the co-occurrence times of device entity and fault entity and environment entity in the same data of multi-source heterogeneous data.

5. A method for predicting the reliability of a transmission line based on a multi-modal cognitive network according to claim 1, characterized in that, The constructing a multi-modal cognitive network based on the three-layer meta-model framework and the association strength between entities specifically includes: Map the entities of the equipment category, fault category, and environment category in the three-layer meta-model framework to the nodes in the multi-modal cognitive network, and the node attributes are entity attributes; According to the calculated association relationship and association strength between entities, construct the edges between nodes, and the attributes of the edges are the association strength between the node entities; Use the Neo4j graph database for the structured storage of the multi-modal cognitive network, and organize the relationship between nodes and edges in the form of a graph structure.

6. The reliability prediction method for a transmission line based on a multi-modal cognitive network according to claim 1, wherein, The using the multi-modal cognitive network to extract the real-time status parameters and degradation information of the equipment, and constructing an equipment-level dynamic failure rate prediction model based on health indicators specifically includes: Through the improved Weibull-Markov chain algorithm, quantify the influence of equipment aging and external environmental factors on the failure rate, and obtain the equipment aging correction coefficient and external environment correction coefficient; According to the real-time detection value of the device , using the health status indicator to evaluate the operating health status of the device, the calculation formula of the health index is as follows: Among them, is the detection value of the i th indicator at time t . The indicators include health indicators such as the real-time current, voltage, and temperature of the device; is the weight of the i th health indicator, which is determined by the analytic hierarchy process; N is the number of real-time detection values, , are the threshold ranges of the real-time monitoring values; Based on the entities and their attribute information of the equipment category, fault category, and environment category extracted from the multi-modal cognitive network, combined with the equipment aging correction coefficient, external environment correction coefficient, and health index, construct an equipment-level time-varying failure rate prediction model.

7. A method for predicting the reliability of a transmission line based on a multi-modal cognitive network according to claim 6, characterized in that The formula for the equipment aging correction coefficient is: Among them, is the device entity is the device aging correction coefficient at the current moment t, and are respectively the shape parameter and scale parameter of the device entity ; is the current time, is the device entity is the starting time of normal operation; The calculation formula for the external environment correction coefficient is as follows: Among them, represents the external environment correction coefficient of the device entity ; represents the number of occurrences of the k-th type of environmental entity associated with the device entity ; is the total number of occurrences of all environmental entities related to the device entity ; is the total number of environmental entities considered ; represents the influence weight of the environmental entity on the device, which is determined according to expert knowledge.

8. A method for predicting the reliability of a transmission line based on a multi-modal cognitive network according to claim 6, characterized in that, The equipment-level time-varying failure rate prediction model is as follows: Among them, is the dynamic failure rate of the device entity at the current moment t , is the reference failure rate, that is, the initial failure rate without degradation or environmental impact, is the set of failure type entities, is the set of environmental type entities, , are respectively the device entity and the failure entity , the device entity and the environmental entity 's association strength, is the number of occurrences of the failure entity t within the time , is the total number of occurrences of all failure entities within the time t , is the real-time monitoring value of the environmental entity at time t.

9. A method for predicting the reliability of a transmission line based on a multi-modal cognitive network according to claim 1, characterized in that, Fusing the equipment-level dynamic failure rate model with environmental factors, calculating the line-level basic fault risk and environmental risk, and constructing the overall reliability prediction model for the transmission line, specifically including: Based on the topological association relationship among device entities in the multimodal cognitive network, according to the criticality weight of the device in the topological structure, the dynamic failure rate of each device entity is aggregated into the basic fault risk at the line level , and the calculation formula is as follows: Among them, is the device entity at the current moment t is the dynamic failure rate of is the device entity is the key weight in the line structure, determined based on the device redundancy and load distribution ratio parameters, and m is the total number of device entities in the transmission line. is the line-level basic fault risk at the current moment t; Using the association strength between the device and the environmental entity and the weight factor of each environmental factor , construct an environmental risk model at the device level: Among them, is the equipment-level environmental risk of the equipment entity . represents the weight factor of the environmental entity for the equipment, which is determined according to expert knowledge represents the equipment entity and the environmental entity The association strength between them represents the environmental entity The real-time monitoring value at the current moment t, where p is the total number of environmental entities; Integrate the line-level basic fault risk with the equipment-level environmental risk to construct an overall reliability prediction model for the transmission line. Its comprehensive risk expression is as follows: Among them, is the overall comprehensive risk of the power transmission line at time t , and m is the total number of equipment entities.

10. A method for predicting the reliability of a transmission line based on a multi-modal cognitive network according to claim 9, characterized in that, The weight factors of the respective environmental factors are determined by the entropy weight method, and the formula is as follows: Among them, is the information entropy of the environmental entity . is the real-time monitoring value of the environmental entity at time . M is the number of samples, that is, the total number of time segments or monitoring points used for statistical analysis. is the sum of all sample values of the environmental entity .

Images