Detection model generation method, power transmission line state determination method and related device

Abstract

The invention discloses a detection model generation method, a power transmission line state determination method and a related device, and the method comprises the steps: obtaining the maintenance record data of a power transmission line in a target time period, and enabling the target time period to represent a time period before a current moment; based on feature data of the detection points in the maintenance record data, a time-varying state network is determined, and the time-varying state network comprises the relation between the detection points in the power transmission line and the condition that the state of the power transmission line changes along with time; and determining a target detection model based on the time-varying state network and the candidate state corresponding to the power transmission line, the target detection model being used for detecting the state of the power transmission line.

Description

Technical Field

[0001] The present application relates to the technical field of data processing, and more specifically, to a method for generating a detection model, a method for determining the state of a transmission line, and related devices. Background Art

[0002] In the power system, as a key facility for power transmission, the stable operation of a transmission line is directly related to the reliability and safety of the entire power system. With the continuous growth of power demand, the scale and complexity of transmission lines have also been increasing. Currently, the method for evaluating the state of a transmission line based on on-line monitoring data has problems such as poor accuracy in evaluating the state of the transmission line and limited timeliness when facing a complex and changeable operating environment and line. Summary of the Invention

[0003] In view of this, the present application provides the following technical solutions:

[0004] A method for generating a detection model, comprising:

[0005] Obtaining maintenance record data of a transmission line within a target time period, where the target time period represents the time period before the current moment;

[0006] Based on the characteristic data of the detection points in the maintenance record data, determining a time-varying state network, where the time-varying state network includes the relationships between the various detection points in the transmission line and the situation of the state of the transmission line changing over time;

[0007] Based on the time-varying state network and the candidate states corresponding to the transmission line, determining a target detection model, where the target detection model is used to detect the state of the transmission line.

[0008] Optionally, the determining a time-varying state network based on the characteristic data of the detection points in the maintenance record data includes:

[0009] Based on the characteristic data of the detection points in the maintenance record data, generating a target vector for each detection point, where the target vector represents the operating state of the transmission line at the location where the detection point is located;

[0010] Based on the target vector of the detection point, determining a state attenuation trend parameter of the detection point, where the state attenuation trend parameter represents the change trend of the state corresponding to the detection point over time;

[0011] According to the relationships between the various detection points, determining a relationship model between the detection points, where the relationship model represents the mutual influence between the detection point states and the information on the change of the mutual influence over time;

[0012] Determine a time-varying state network based on the state decay trend parameters of the detection points and the relationship model between the detection points.

[0013] Optionally, the determining the time-varying state network based on the state decay trend parameters of the detection points and the relationship model between the detection points includes:

[0014] Determine the dynamic decay probability distribution of the detection points based on the state decay trend parameters of the detection points;

[0015] Determine a time-varying state network based on the dynamic decay probability distribution of the detection points and the relationship model between the detection points; wherein, the nodes of the time-varying state network represent detection points, and the edges between the nodes of the time-varying state network represent the state relationships between the detection points corresponding to the relationship model; the weights of the edges of the time-varying state network represent the state transition probabilities.

[0016] Optionally, the method further includes:

[0017] Update the time-varying state network when the maintenance record data is updated.

[0018] Optionally, the determining the target detection model based on the time-varying state network and the candidate states corresponding to the transmission line includes:

[0019] Determine the candidate states corresponding to the transmission line according to the operation characteristics of the transmission line and the evaluation requirement information;

[0020] Determine the time-varying state network as the observation sequence, and the dimension of the observation sequence matches the number of detection points included in the time-varying state network;

[0021] Determine a state transition probability matrix based on the observation sequence, and the state transition probability matrix represents the transition probability from the current candidate state to the next candidate state;

[0022] Determine the observation probability matrix from the candidate states to the observation sequence, and the observation probability matrix represents the relationship between the candidate states and the observed values corresponding to the observation sequence;

[0023] Determine the target detection model based on the state transition probability matrix and the observation probability matrix.

[0024] Optionally, the determining the target detection model based on the state transition probability matrix and the observation probability matrix includes:

[0025] Determine a dynamic weight matrix according to the real-time change of the time-varying state network and the fault data within the target time period, and the dynamic weight matrix is used to adjust the state transition probability matrix and the observation probability matrix;

[0026] Generate a preliminary model based on the state transition probability matrix, the observation probability matrix, and the dynamic weight matrix;

[0027] Optimize the model parameters of the preliminary model to obtain an object detection model.

[0028] Optionally, the optimizing the model parameters of the preliminary model to obtain an object detection model includes:

[0029] Iteratively optimize the model parameters of the preliminary model based on the test data corresponding to the transmission line to obtain an object detection model, where the test data includes the feature data of the detection points of the transmission line and the state information of the transmission line.

[0030] A method for determining the state of a transmission line includes:

[0031] Obtain the monitoring data of the detection points corresponding to the transmission line;

[0032] Process the monitoring data based on the object detection model to obtain the target state of the transmission line;

[0033] The generation process of the target processing model includes: obtaining the maintenance record data of the transmission line in the target time period, where the target time period represents the time period before the current moment; determining a time-varying state network based on the feature data of the detection points in the maintenance record data, where the time-varying state network includes the relationship between the detection points in the transmission line and the change of the state of the transmission line over time; determining an object detection model based on the time-varying state network and the candidate states corresponding to the transmission line.

[0034] A device for generating a detection model includes:

[0035] A first acquisition unit, configured to obtain the maintenance record data of the transmission line in the target time period, where the target time period represents the time period before the current moment;

[0036] A first determination unit, configured to determine a time-varying state network based on the feature data of the detection points in the maintenance record data, where the time-varying state network includes the relationship between the detection points in the transmission line and the change of the state of the transmission line over time;

[0037] A second determination unit, configured to determine an object detection model based on the time-varying state network and the candidate states corresponding to the transmission line, where the object detection model is used to detect the state of the transmission line.

[0038] A device for determining the state of a transmission line includes:

[0039] A second acquisition unit, configured to acquire monitoring data of a corresponding detection point of a transmission line;

[0040] A processing unit, configured to process the monitoring data based on a target detection model to obtain a target state of the transmission line;

[0041] The generation process of the target processing model includes: acquiring maintenance record data of the transmission line within a target time period, where the target time period represents the time period before the current moment; determining a time-varying state network based on the feature data of the detection points in the maintenance record data, where the time-varying state network includes the relationships between the detection points in the transmission line and the change of the state of the transmission line over time; determining a target detection model based on the time-varying state network and the candidate states corresponding to the transmission line. Description of the Drawings

[0042] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only the embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0043] Figure 1 It is a schematic flowchart of a method for generating a detection model provided by an embodiment of the present application;

[0044] Figure 2 It is a schematic flowchart of a method for determining a target detection model provided by an embodiment of the present application;

[0045] Figure 3 It is a schematic flowchart of a method for determining the state of a transmission line provided by an embodiment of the present application;

[0046] Figure 4 It is a schematic structural diagram of a device for generating a detection model provided by an embodiment of the present application;

[0047] Figure 5 It is a schematic structural diagram of a device for determining the state of a transmission line provided by an embodiment of the present application. Detailed Description of the Embodiments

[0048] The following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present application.

[0049] The terms "first", "second", etc. in this application are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising", "having", and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may include steps or units not listed.

[0050] An embodiment of this application provides a method for generating a detection model. This method is used to generate a target detection model, which can be applied to the scenario of detecting the state of a transmission line. For example, the target detection model can be applied to the scenario of detecting the state of a transmission line based on the Internet of Things. The Internet of Things technology enables interconnection and interoperability between devices. By deploying various sensors along the transmission line, the operating data of the line can be collected in real time, and then these data can be transmitted to a processing device, so that the processing device can provide comprehensive and accurate information for the evaluation of the transmission line state based on the target detection model, thereby obtaining a more accurate state of the transmission line.

[0051] See Figure 1 , which is a schematic flowchart of a method for generating a detection model provided by an embodiment of this application. This method may include the following steps:

[0052] S101. Obtain the maintenance record data of the transmission line during the target time period.

[0053] The transmission line is an important part of the power system for transmitting electric energy. It connects power plants and substations, as well as between substations, and conveys electric energy from the power generation end to the power consumption end. The transmission line can be divided into different types according to different classification conditions. The type of the corresponding transmission line can be obtained according to the actual application scenario. In the embodiment of this application, the type of the transmission line is not limited.

[0054] The target time period guarantees the time period before the current moment. In the embodiments of the present application, the maintenance record data of the transmission line can be collected in real time, and this maintenance record data is used as the basis for generating the target maintenance model. Correspondingly, the maintenance record data can include the record data of the corresponding detection points in the transmission line. For example, the maintenance record data can include the maintenance information, fault situation information, fault type information, and solution information of the corresponding detection points, etc. In order to be able to apply accurate maintenance record data, after the maintenance record data is collected by the sensors corresponding to the transmission line in the Internet of Things, the collected maintenance record data can be preprocessed, and then more accurate maintenance record data can be obtained. Among them, data preprocessing can include data cleaning (removing duplicate and incorrect data), data standardization (unifying data formats and dimensions), etc., and converting it into a standard data form that can be directly used in subsequent steps to provide reliable data support for subsequent analysis.

[0055] S102. Determine the time-varying state network based on the characteristic data of the detection points in the maintenance record data.

[0056] The maintenance record data can include the record data of multiple detection points. Among them, the detection point represents the position point for line state maintenance in the transmission line, or can also represent the line point that has been repaired in the transmission line, etc. In the process of generating the detection model, the record data of each detection point can be applied, or the record data of the detection points with specific conditions can be applied. For example, the record data of the detection points that meet specific position conditions, or the record data of the representative detection points on the transmission line, etc. Determine the characteristic data of the detection points according to the maintenance record data of these detection points. The characteristic data can include maintenance characteristics, environmental factor characteristics of the detection point location, and electrical factor characteristics corresponding to the detection point, etc. Among them, the maintenance characteristics can include the maintenance data of the detection point; the environmental factor characteristics can include the environmental information of the data recording the detection point, the environmental information where the transmission line is located, or the environmental information corresponding to the location of the detection point, etc.; the electrical factor characteristics can include characteristics such as current and voltage.

[0057] Then, based on the characteristic data of the detection points, a time-varying state network is determined. The time-varying state network includes the relationships between the various detection points in the transmission line and the situation of the state of the transmission line changing over time. Among them, due to the different positions of the detection points on the transmission line, the relationships between the detection points can characterize the associations and related impacts of the operating states at different positions of the transmission line, and these relationships are not fixed and will change over time. In order to accurately describe the characteristics of each detection point, the time-varying state network also includes the situation of the state of the transmission line changing over time. For example, at the first moment, the transmission line is in a stable state, and at the second moment, the transmission line is in an abnormal state, etc. In this way, more accurate key data can be provided for the detection model when establishing the target detection model subsequently.

[0058] In the embodiment of the present application, the time-varying state network is a dynamic network model for describing the complex relationships of the state of the transmission line changing over time. Correspondingly, the time-varying state network can be a node network structure with the detection points as nodes, or a time series graph structure, etc. The corresponding time-varying state network structure can be selected according to the actual application scenario requirements and the quantity characteristics of the detection points.

[0059] S103. Determine a target detection model based on the time-varying state network and the candidate states corresponding to the transmission line.

[0060] The candidate states of the transmission line characterize the states that cannot be directly observed inside the transmission line but can affect the power transmission of the line. For example, the candidate states can include the normal state, abnormal state, fault state, etc. of the transmission line. The time-varying state network can be used as an observation sequence to determine the probability distribution between the candidate states and the observation sequence, and the probability from the observation sequence to the candidate states can also be determined. Thus, a target detection model can be established based on these dynamic probabilities. The target detection model can establish a state relationship network for various states of the transmission line, thereby balancing and optimizing multiple states. In the embodiment of the present application, the target detection model can be a state transition model, such as a state transition model generated based on the hidden Markov model structure.

[0061] In this way, based on the situation of the state changing over time in the time-varying state network, constructing the target detection model can avoid the influence of the relationship between the fixed state network and the candidate states, and ensure the balance of the state transition over time. Thus, the state of the transmission line can be predicted in real time based on the target detection model, improving the accuracy and timeliness of the transmission line state prediction.

[0062] Next, the generation method of the detection model in the embodiment of the present application will be described in combination with an actual application scenario.

[0063] To determine the target detection model, it is necessary to use the time-varying state network as the key data basis for generating the target detection model. The characteristic data of the detection points can be modeled according to the time series analysis method. Among them, the time series data can be the data of the characteristic data of the detection points changing with time. For example, if the characteristic data includes current data, the time series can be the change data of the current with time, etc. In this way, the time-varying state network can be constructed by analyzing the trends, periodicities, and randomness in the time series data. For example, the state relationship between detection points can also be modeled based on the method of graph neural network to obtain the corresponding time-varying state network. In the embodiments of the present application, the structure of the time-varying state network is not specifically limited.

[0064] In an implementation manner of the embodiments of the present application, based on the characteristic data of the detection points in the maintenance record data, the process of determining the time-varying state network may include the following steps:

[0065] S201. Generate a target vector for each detection point based on the characteristic data of the detection points in the maintenance record data.

[0066] S202. Determine the state decay trend parameter of the detection point based on the target vector of the detection point.

[0067] S203. Determine the relationship model between the detection points according to the relationship between each detection point.

[0068] S204. Determine the time-varying state network based on the state decay trend parameter of the detection point and the relationship model between the detection points.

[0069] In step S201, the maintenance record data is collected from several detection points on the transmission line, and data including environmental factor characteristics (such as temperature, humidity, wind force, etc.) and electrical factor characteristics (such as current, voltage, short circuit, etc.) is used as the characteristic data of the detection points. These characteristic data are converted into target vectors, which can also be called the state index vectors after line fusion. The target vector characterizes the operating state of the transmission line at the location of the detection point, and its comprehensive characteristics of environmental factors and electrical factors can describe the comprehensive state of the detection point. For example, the environmental factors and electrical factors of each detection point can be normalized to ensure that the data of different characteristics are in the same magnitude. The normalized characteristic data are combined into a vector as the target vector of the detection point. Thus, the target vector is used as the basis for subsequent determination of the time-varying state network.

[0070] In step S202, the target vector of the detection point is obtained, and the state decay trend parameter is determined. This state decay trend parameter characterizes the change trend of the state corresponding to the detection point over time. For example, by setting a decay function, the decay confidence of the target vector can be calculated, which reflects the change trend of the state over time. Different state levels (such as normal power transmission line, potential fault, fault, etc.) can be divided according to the decay confidence. Correspondingly, the state decay trend parameter of each detection point can include the decay confidence and the dynamic decay probability distribution, which reflects the dynamic change of the detection point state and provides a quantitative basis for the subsequent construction of the time-varying state network.

[0071] In step S203, the relationship model between detection points is determined. This relationship model characterizes the mutual influence between the states of detection points and the information of the mutual influence changing over time. In this step, based on the target thinking and the state decay trend parameter between detection points, a dynamic relationship model is constructed to describe the state relationship between detection points. This relationship model reflects the mutual influence and transfer trend between the states of detection points, dynamically changes over time, and is the core of the time-varying state network, used to describe the evolution process of the detection point state. Thus, the time-varying state network can be determined in step S204. This time-varying state network is a dynamic graph structure, where nodes represent detection points, edges represent the state relationship between detection points, and the weight of the edge reflects the probability of state transfer. The time-varying state network captures the dynamic evolution process of the detection point state and provides a basis for the subsequent state transfer analysis.

[0072] In the embodiment of the present application, when determining the time-varying state network based on the state decay trend parameter of the detection point and the relationship model between detection points, the time-varying state network can be determined according to the adopted time-varying state network structure, or the state decay trend parameter of the detection point can be analyzed to determine the data affected by time for state transfer, and then the time-varying state network is generated. Correspondingly, in an implementation manner of the embodiment of the present application, the process of determining the time-varying state network based on the state decay trend parameter of the detection point and the relationship model between detection points can include: determining the dynamic decay probability distribution of the detection point based on the state decay trend parameter of the detection point; determining the time-varying state network based on the dynamic decay probability distribution of the detection point and the relationship model between detection points.

[0073] Among them, the nodes of the time-varying state network represent the detection points, and the edges between the nodes of the time-varying state network represent the state relationships between the detection points corresponding to the relationship model; the weights of the edges of the time-varying state network represent the state transition probabilities. The dynamic decay probability distribution represents the possibility of the detection point state changing over time. For example, a decay function can be preset to calculate the decay confidence of the state vector of each detection point, and this decay confidence reflects the change trend of the detection point state over time. According to the decay confidence, the state decay trend parameters are divided into different state levels (such as normal, potential fault, fault, etc.). Then calculate the dynamic decay probability distribution of each detection point. Each detection point serves as a node of the time-varying state network. According to the relationship model between the detection points, construct the edges between the nodes, and the edges represent the state relationships between the detection points. According to the dynamic decay probability distribution, calculate the state transition probability between the nodes as the weight of the edge. The state transition probability reflects the transition law between the detection point states. Thus, the time-varying state network captures the dynamic evolution process of the detection point states and provides a basis for subsequent state transition analysis.

[0074] Since the time-varying state network represents the trend of the detection point state changing over time, it is necessary to dynamically update the time-varying state network to improve its application accuracy. In one implementation, when the maintenance record data is updated, the time-varying state network is updated. With the acquisition of new maintenance record data, the structure and weights of the time-varying state network are updated in real time so that the time-varying state network can reflect the latest line state changes of the transmission line.

[0075] After obtaining the time-varying state network, the target detection model can be determined according to the time-varying state network and the candidate states. The target detection model can be determined according to the different model structures of the detection model, or the corresponding candidate states can be determined according to different evaluation requirements. The determination process of this target detection model can be matched with the requirements of the actual application scenario. In one implementation of the embodiments of the present application, a method for determining a target detection model based on the time-varying state network and the candidate states corresponding to the transmission line is also provided. See Figure 2 , and this method may include the following steps:

[0076] S301. Determine the candidate states corresponding to the transmission line according to the operation characteristics of the transmission line and the evaluation requirement information.

[0077] S302. Determine the time-varying state network as the observation sequence.

[0078] S303. Based on the observation sequence, determine the state transition probability matrix.

[0079] S304. Determine the observation probability matrix from the candidate states to the observation sequence.

[0080] S305. Determine the target detection model based on the state transition probability matrix and the observation probability matrix.

[0081] According to the operating characteristics of the transmission line (such as environmental factor characteristics, electrical factor characteristics, etc.) and the evaluation requirement information (such as fault type, health status, etc.), determine the candidate states of the transmission line. The candidate states are the states that cannot be directly observed in the transmission line and can also be called hidden states, such as the normal state, potential fault state, and fault state of the transmission line. For example, the model structure of the target detection model can adopt the hidden Markov state transition model, use the time-varying state network as the observation sequence, and the dimension of the observation sequence matches the number of detection points included in the time-varying state network. The state vector of each detection point is the observation value, so as to construct the observation sequence, which reflects the dynamic change of the detection point state. This observation sequence can be the input of the hidden Markov state transition model for inferring the hidden state (i.e., the candidate state). Analyze the observation sequence, calculate the transition probability between candidate states, and construct the state transition probability matrix to reflect the transition law between candidate states. Then determine the observation probability matrix, which characterizes the relationship between the candidate state and the observation value corresponding to the observation sequence. Calculate the probability distribution from the candidate state to the observation matrix, that is, the occurrence probability of the observation value given the candidate state. The observation probability matrix is another key parameter of the hidden Markov state transition model, which is used to describe the relationship between the candidate state and the observation value. Finally, combine the state transition probability matrix and the observation probability matrix to determine the hidden Markov state transition model. Through model calculation, predict the health status of the transmission line in real time. Combine historical fault data to optimize the model parameters to ensure the accuracy and stability of the model.

[0082] Furthermore, in the embodiments of the present application, in order to ensure the accuracy of the target detection model, the target detection model can also be determined in combination with real-time change data. In one implementation manner, determining the target detection model based on the state transition probability matrix and the observation probability matrix includes: determining the dynamic weight matrix according to the real-time change of the time-varying state network and the fault data within the target time period; generating a preliminary model based on the state transition probability matrix, the observation probability matrix, and the dynamic weight matrix; and optimizing the model parameters of the preliminary model to obtain the target detection model.

[0083] Among them, the dynamic weight matrix is used to adjust the state transition probability matrix and the observation probability matrix. The structure and weight changes of the time-varying state network are monitored in real time. According to the dynamic changes of the detection point state, the state transition and observation probabilities are adjusted. It is also possible to analyze the fault data within the target time period to determine the fault type and fault frequency. For example, the risk probability of each fault type can be calculated using the Poisson distribution function. Based on the real-time changes of the time-varying state network and the fault data, the dynamic weight matrix is determined. The dynamic weight matrix is used to adjust the state transition probability and the observation probability in real time to ensure the balance of state transition. Taking the model structure of the target detection model as the hidden Markov state transition model as an example, the state transition probability matrix and the observation probability matrix are combined to construct a preliminary hidden Markov state transition model. The dynamic weight matrix is introduced into the preliminary model to adjust the state transition probability and the observation probability in real time. A preliminary hidden Markov state transition model is generated, which can predict the health state of the transmission line in real time. Then, the model parameters of the preliminary model can be optimized to obtain the target detection model. For example, the parameters of the state transition probability matrix, the observation probability matrix, and the dynamic weight matrix can be optimized by iterative calculation to ensure the accuracy of the target detection model.

[0084] In an implementation manner of the embodiment of the present application, optimizing the model parameters of the preliminary model to obtain the target detection model includes: iteratively optimizing the model parameters of the preliminary model based on the test data corresponding to the transmission line to obtain the target detection model. Among them, the test data includes the characteristic data of the detection points of the transmission line and the state information of the transmission line. The state information of the transmission line corresponding to the test data can be the information of an accurate state. For example, the state information determined through the actual manual maintenance process. The characteristic data in the test data can be input into the preliminarily generated preliminary model, and then the state information obtained based on the preliminary model is compared with the state information in the test data. According to the comparison result, the state transition probability matrix, the observation probability matrix, and the dynamic weight matrix are updated, so as to obtain the target detection model. It should be noted that in the embodiment of the present application, the optimization of the preliminary model adopts an iterative optimization method. By repeatedly applying the corresponding iterative processing method, the difference between the state information estimated by the preliminary model and the state information in the test data is gradually improved, and finally the optimal solution is approximated. In each iteration, the parameters or structure of the solution are adjusted according to the nature of the current solution and the characteristics of the objective function, in order to obtain a better solution in the next iteration.

[0085] The method for generating the detection model according to the embodiment of the present application will be described below in combination with an actual application scenario.

[0086] For example, in a large - scale power transmission network, there is a power transmission line from a mountain power station to an urban substation. The total length of this line is 50 kilometers, passing through complex terrains including mountains, rivers, and farmlands, facing natural environmental factors such as large temperature differences, strong winds, and humidity, and there is a risk of biological corrosion in some areas. Through various sensors installed on the power transmission line, the maintenance records of the line are collected in real - time, including information such as the time, location, maintenance content, and problems found during each maintenance. These data will be transmitted to the data processing center, and then the data is cleaned to remove duplicate and incorrect data, such as incorrect timestamps or obviously unreasonable maintenance records. At the same time, the data is standardized, and the maintenance content in different formats is uniformly converted into a format convenient for analysis. For example, the fault types are classified and coded to prepare for subsequent analysis.

[0087] Then, a time - varying state network of the relationship between the maintenance records of the detection points is established. The attenuation trend is classified into state levels by calculating the attenuation confidence of each detection point (which can also be called the central point), and the dynamic attenuation probability distribution of each detection point is calculated. For example, considering the long length of the power transmission line and the limitations of state measurement, 10 detection points are randomly selected as central points. For each central point, the line integration state index vector is obtained, and these index vectors contain information such as the electrical parameters (such as current, voltage) and environmental parameters (such as temperature, humidity) of the line. Since the environmental factors and the operation conditions of the line are constantly changing, the relationship between the central points changes over time, and the time - varying state network also changes dynamically. The attenuation function of the line integration state index vector is calculated by a computer. For example, it can be known that for a central point located in the mountains, due to the large temperature difference in winter, the attenuation trend of its line integration state index vector is significantly accelerated. By calculating the attenuation confidence, the attenuation trend of this central point is classified into the "severe attenuation" state level, and its dynamic attenuation probability distribution is calculated.

[0088] Taking the target detection model as an example of the hidden Markov state transition model, the hidden states of the hidden Markov state transition model (i.e., the candidate states of the transmission line) can be set, and an improved hidden Markov state transition model can be constructed based on the time-varying state network and the hidden states. For example, according to the constructed time-varying state network, the hidden states of the hidden Markov state transition model are set. Suppose the hidden states include three types: "line normal", "minor fault potential", and "severe fault potential". Taking the time-varying state network as the observation sequence to construct the model, it is determined that the probability distribution between the hidden state and the observation sequence follows a Gaussian distribution. Historical data on the fault frequency caused by environmental and electrical factors of the line hidden state in the neighborhood of the central point are collected. For example, in the strong wind area of the mountainous area, the fault frequency of short circuit caused by the shaking of the line due to strong wind is relatively high. The Poisson distribution function is used to calculate the probability of the occurrence risk of each fault type. For example, the probability of short circuit fault caused by strong wind is 0.3. Then, the joint probability distribution of the fault probability and the dynamic decay probability distribution of the central point is calculated to construct the state transition probability matrix. At the same time, according to the real-time dynamically adjusted time-sequence state network, the relationship between the weight and the line state is updated in real time to obtain the index balance weight, ensuring the balance of state transition between lines.

[0089] Then, according to the improved Hidden Markov state transition model (i.e., the target detection model), the state of the transmission line is predicted in real time and dynamically. Among them, the state of the transmission line can include the risk probability of fault type transition. Further, the state evaluation data of the transmission line collected in real time and the response changes of the transmission line can be pooled and stored, and then the big data platform is used for distributed storage and management. The cross-region data interaction technology is adopted to exchange the operation status and fault detection results of transmission lines in different regions. Through detailed data quantification and iteration, the fault types are classified, the state of the transmission line is carefully detected and fault analyzed, and a sound multi-state evaluation system is constructed to predict the faults of the transmission line in real time. Through the Internet of Things, cross-region information transmission is carried out, so as to improve the training accuracy of the transmission line state evaluation system. For example, using the constructed improved Hidden Markov state transition model, the real-time data of the time-varying state network is input. If the model predicts that a certain section of the line near the river has a relatively high risk probability of transferring from "minor fault hidden danger" to "serious fault hidden danger" due to the long-term influence of the humid environment, reaching 0.6. This prediction result is timely fed back to the operation and maintenance personnel to remind them to focus on the line in this area. The evaluation data collected in real time (such as the real-time state indicators of each center point, the predicted fault risk probability, etc.) and the response changes of the distribution line (such as the actual response of the line to the fault risk) are pooled and stored through the big data platform for distributed storage and management. Using the cross-region data interaction technology, the operation status and fault detection results of transmission lines in other regions are exchanged. For example, by comparing with another transmission line passing through a similar humid environment, it is found that the other party has taken measures to install moisture-proof equipment to effectively reduce the fault risk. The operation and maintenance personnel of this line refer to this experience and plan to take similar measures for the line in the area affected by humidity, so as to continuously optimize the line state evaluation and operation and maintenance strategies.

[0090] Correspondingly, in the process of determining the target detection model (e.g., the improved hidden Markov state transition model), the time-varying state network is used as the observation sequence to construct the hidden Markov state transition model of the transmission line. During the construction of the hidden Markov state transition model, the probability distribution from the hidden state to the observation sequence follows a Gaussian distribution, and the dimension of the observation sequence is the number of central points of the static graph of the time-varying state network at the τ time state. Then, calculate the probability distribution from the observation sequence to the hidden state, that is, divide the state level number n of the line fusion state index vector corresponding to the hidden state through the dynamic decay probability distribution, and calculate the probability distribution that the observation sequence corresponding to the time-varying state network at the τ time state is transferred to the hidden state. Further, collect the historical data of the frequencies of faults caused by environmental factors and electrical factors in the hidden state on the lines in the neighborhood of the detection point. The fault types caused by environmental factors include temperature difference, humidity, wind force, environmental electron aggregation, and biological corrosion. The fault types caused by electrical factors include the fault frequency caused by current load imbalance, the fault frequency caused by voltage load imbalance, the fault frequency caused by short circuit, and the fault frequency caused by harmonic and noise interference. Use the Poisson distribution function to calculate the probabilities of the risks of occurrence of each fault type respectively, and calculate the joint probability distribution of the probabilities of the risks of occurrence of each fault type and the dynamic decay probability distribution of the corresponding detection point. If the joint probability distribution satisfies the first condition and the second condition, the first condition is not less than 0 and lower than the preset probability threshold, and the second condition is that the sum of all joint probabilities is equal to 1. Then construct the state transition probability matrix. For example, the mathematical expression of the state transition probability matrix is:

[0091]

[0092] Among them, P(τ) is the state transition probability matrix at the τ time state, and Pnm is the probability that the nth observation sequence is transferred to the mth hidden state.

[0093] The traditional hidden Markov state transition model usually assumes that the relationship between the observed value and the hidden state is fixed. However, in actual situations, the change of the observed value will affect the transition probability of the hidden state. To avoid some states being too concentrated or overly declining, it is necessary to ensure that the transitions between all states are balanced. In the embodiments of the present application, through 19) the weight matrix adjusts the relationship between the weight and the line state in real time according to the real-time dynamic time-sequence state network, and obtains the index balance weight updated in real time to ensure that the state transitions between lines are balanced. Form the risk probability matrix of the fault type. The corresponding mathematical expression of the risk probability matrix can be:

[0094] P(t)=[P1(t), P2(t),…, P j (t)]

[0095] Wherein, t represents a time parameter, and j represents the number of fault types. The characteristic matrix and corresponding eigenvalues of the risk probability matrix are solved by using the Laplace transform. The eigenvalues are the model parameters of the improved hidden Markov state transition model, and a line health state set is obtained. The line health state set includes a fault state and the risk probability corresponding to the fault state.

[0096] Correspondingly, in an embodiment of the present application, a method for determining the state of a transmission line is further provided. Refer to Figure 3 , and the method may include the following steps:

[0097] S401. Obtain the detection data of the detection points corresponding to the transmission line.

[0098] S402. Process the monitoring data based on the target detection model to obtain the target state of the transmission line.

[0099] Wherein, the generation process of the target processing model includes: obtaining the maintenance record data of the transmission line in the target time period, where the target time period represents the time period before the current moment; determining a time-varying state network based on the characteristic data of the detection points in the maintenance record data, where the time-varying state network includes the relationships between the detection points in the transmission line and the situation of the state of the transmission line changing over time; determining the target detection model based on the time-varying state network and the candidate states corresponding to the transmission line. Correspondingly, the generation process of the target processing model can refer to the foregoing embodiments and will not be elaborated here.

[0100] In this embodiment, the monitoring data of the detection points corresponding to the obtained transmission line may be the data collected in real time by the Internet of Things, or the data obtained after preprocessing the real-time collected data. The target state of the transmission line may include representing the health state of the transmission line, or may also include the risk probability of fault type transition. In the embodiment of the present application, while predicting the transmission line in real time, the target detection model can be continuously improved, and the integrity of the line multi-state evaluation system can be updated, thereby improving the accuracy of determining the state of the transmission line.

[0101] In another embodiment of the present application, a device for generating a detection model is further provided. Refer to Figure 4 , including:

[0102] A first acquisition unit 501, configured to obtain the maintenance record data of the transmission line in the target time period, where the target time period represents the time period before the current moment;

[0103] A first determination unit 502, configured to determine a time-varying state network based on the characteristic data of the detection points in the maintenance record data, where the time-varying state network includes the relationships between the detection points in the transmission line and the situation of the state of the transmission line changing over time;

[0104] A second determination unit 503, configured to determine a target detection model based on the time-varying state network and candidate states corresponding to the transmission line, where the target detection model is used to detect the state of the transmission line.

[0105] Optionally, the first determination unit includes:

[0106] A first generation subunit, configured to generate a target vector for each detection point based on the feature data of the detection points in the maintenance record data, where the target vector characterizes the operating state of the transmission line at the location where the detection point is located;

[0107] A first determination subunit, configured to determine a state attenuation trend parameter of the detection point based on the target vector of the detection point, where the state attenuation trend parameter characterizes the change trend of the state corresponding to the detection point over time;

[0108] A second determination subunit, configured to determine a relationship model between detection points according to the relationship between each detection point, where the relationship model characterizes the mutual influence between the states of the detection points and the information on the change of the mutual influence over time;

[0109] A third determination subunit, configured to determine a time-varying state network based on the state attenuation trend parameter of the detection point and the relationship model between the detection points.

[0110] Optionally, the third determination subunit is configured to:

[0111] Determine a dynamic attenuation probability distribution of the detection point based on the state attenuation trend parameter of the detection point;

[0112] Determine a time-varying state network based on the dynamic attenuation probability distribution of the detection point and the relationship model between the detection points; where nodes of the time-varying state network represent detection points, edges between nodes of the time-varying state network represent state relationships between corresponding detection points of the relationship model; and weights of edges of the time-varying state network represent state transition probabilities.

[0113] Optionally, the apparatus further includes:

[0114] A network update unit, configured to update the time-varying state network when the maintenance record data is updated.

[0115] Optionally, the second determination unit includes:

[0116] A fourth determination subunit, configured to determine candidate states corresponding to the transmission line according to the operating characteristics of the transmission line and evaluation requirement information;

[0117] A fifth determination subunit, configured to determine the time-varying state network as an observation sequence, where the dimension of the observation sequence matches the number of detection points included in the time-varying state network;

[0118] A sixth determination subunit, configured to determine a state transition probability matrix based on the observation sequence, where the state transition probability matrix represents the transition probability from the current candidate state to the next candidate state;

[0119] A seventh determination subunit, configured to determine an observation probability matrix from the candidate state to the observation sequence, where the observation probability matrix represents the relationship between the candidate state and the observed value corresponding to the observation sequence;

[0120] An eighth determination subunit, configured to determine a target detection model based on the state transition probability matrix and the observation probability matrix.

[0121] Optionally, the eighth determination subunit is configured to:

[0122] Determine a dynamic weight matrix according to the real-time change of the time-varying state network and the fault data within the target time period, where the dynamic weight matrix is used to adjust the state transition probability matrix and the observation probability matrix;

[0123] Generate a preliminary model based on the state transition probability matrix, the observation probability matrix, and the dynamic weight matrix;

[0124] Optimize the model parameters of the preliminary model to obtain a target detection model.

[0125] Optionally, the optimizing the model parameters of the preliminary model to obtain a target detection model includes:

[0126] Iteratively optimize the model parameters of the preliminary model based on the test data corresponding to the transmission line to obtain a target detection model, where the test data includes the feature data of the detection points of the transmission line and the state information of the transmission line.

[0127] It should be noted that the specific implementation of each unit and subunit in this embodiment can refer to the corresponding content in the foregoing, which will not be elaborated here.

[0128] Correspondingly, in an embodiment of the present application, there is also provided a device for determining the state of a transmission line. Refer to Figure 5 , and the device includes:

[0129] A second acquisition unit 601, configured to acquire monitoring data of detection points corresponding to a transmission line;

[0130] A processing unit 602, configured to process the monitoring data based on a target detection model to obtain the target state of the transmission line;

[0131] The generation process of the target processing model includes: obtaining the maintenance record data of the transmission line during the target time period, where the target time period represents the time period before the current moment; determining a time-varying state network based on the feature data of the detection points in the maintenance record data, where the time-varying state network includes the relationships between the detection points in the transmission line and the change of the state of the transmission line over time; and determining a target detection model based on the time-varying state network and the candidate states corresponding to the transmission line.

[0132] It should be noted that the specific implementation of each unit and sub-unit in this embodiment can refer to the corresponding content in the previous text, and will not be elaborated here.

[0133] In another embodiment of the present application, a readable storage medium is further provided, on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned detection model generation method and transmission line state determination method are implemented.

[0134] In another embodiment of the present application, an electronic device is further provided. The electronic device may include:

[0135] A memory for storing an application program and the data generated by the running of the application program;

[0136] A processor for executing the application program to implement the detection model generation method as described above.

[0137] It should be noted that the specific implementation of the processor in this embodiment can refer to the corresponding content in the previous text, and will not be elaborated here.

[0138] In this specification, the various embodiments are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The same or similar parts among the various embodiments can be referred to each other. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple. For the relevant parts, reference can be made to the description of the method part.

[0139] Those skilled in the art can further realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed in this article can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the components and steps of the examples have been generally described according to their functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0140] The steps of the methods or algorithms described in connection with the embodiments disclosed herein may be implemented directly in hardware, in a software module executed by a processor, or in a combination thereof. The software module may be disposed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0141] The foregoing description of the disclosed embodiments enables those skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for generating a detection model, comprising: Obtaining maintenance record data of the power transmission line within a target time period, wherein the target time period represents a time period before a current moment; Determine a time-varying state network based on characteristic data of the detection points in the maintenance record data, wherein the time-varying state network includes the relationship between the detection points in the transmission line and the change of the state of the transmission line over time; Based on the candidate states corresponding to the time-varying state network and the power transmission line, a target detection model is determined, and the target detection model is used to detect the state of the power transmission line.

2. The method according to claim 1, wherein determining the time-varying state network based on the characteristic data of the detection points in the maintenance record data comprises: Based on the characteristic data of the detection point in the maintenance record data, a target vector for each detection point is generated, wherein the target vector represents the operating state of the power transmission line at the location of the detection point; Based on the target vector of the detection point, determining a state decay trend parameter of the detection point, wherein the state decay trend parameter represents a change trend of the state corresponding to the detection point over time; Determine a relationship model between the detection points according to the relationship between the detection points, wherein the relationship model represents the mutual influence between the detection point states and the information of the mutual influence changing over time; A time-varying state network is determined based on the state decay trend parameters of the detection points and the relationship model between the detection points.

3. The method according to claim 2, wherein determining the time-varying state network based on the state decay trend parameter of the detection point and the relationship model between the detection points comprises: Determining a dynamic attenuation probability distribution of the detection point based on a state attenuation trend parameter of the detection point; Based on the dynamic attenuation probability distribution of the detection points and the relationship model between the detection points, a time-varying state network is determined; wherein the nodes of the time-varying state network represent the detection points, and the edges between the nodes of the time-varying state network represent the state relationship between the detection points corresponding to the relationship model; the weights of the edges of the time-varying state network represent the state transition probability.

4. The method according to claim 1, further comprising: When the maintenance record data is updated, the time-varying state network is updated.

5. The method according to claim 1, wherein determining a target detection model based on the candidate states corresponding to the time-varying state network and the transmission line comprises: Determine the candidate states corresponding to the transmission line according to the operation characteristics of the transmission line and the evaluation demand information; Determining the time-varying state network as an observation sequence, the dimension of the observation sequence matches the number of detection points included in the time-varying state network; Based on the observation sequence, determining a state transition probability matrix, wherein the state transition probability matrix represents a transition probability from a current candidate state to a next candidate state; Determine an observation probability matrix from the candidate state to the observation sequence, wherein the observation probability matrix represents a relationship between the candidate state and the observation values ​​corresponding to the observation sequence; Based on the state transition probability matrix and the observation probability matrix, a target detection model is determined.

6. The method according to claim 5, wherein determining the target detection model based on the state transition probability matrix and the observation probability matrix comprises: Determine a dynamic weight matrix according to the real-time changes of the time-varying state network and the fault data in the target time period, wherein the dynamic weight matrix is ​​used to adjust the state transition probability matrix and the observation probability matrix; Generate a preliminary model based on the state transition probability matrix, the observation probability matrix and the dynamic weight matrix; The model parameters of the preliminary model are optimized to obtain a target detection model.

7. According to the method of claim 6, the step of optimizing the model parameters of the preliminary model to obtain the target detection model comprises: The model parameters of the preliminary model are iteratively optimized based on the test data corresponding to the transmission line to obtain the target detection model, wherein the test data includes the feature data of the detection points of the transmission line and the status information of the transmission line.

8. A method for determining a state of a power transmission line, comprising: Obtain monitoring data of corresponding detection points of transmission lines; Processing the monitoring data based on a target detection model to obtain a target state of the transmission line; The generation process of the target processing model includes: obtaining maintenance record data of the transmission line within a target time period, wherein the target time period represents the time period before the current moment; determining a time-varying state network based on the characteristic data of the detection points in the maintenance record data, wherein the time-varying state network includes the relationship between the various detection points in the transmission line and the state of the transmission line changing over time; and determining a target detection model based on the candidate states corresponding to the time-varying state network and the transmission line.

9. A device for generating a detection model, comprising: A first acquisition unit is used to obtain maintenance record data of the power transmission line within a target time period, where the target time period represents a time period before a current moment; A first determining unit is used to determine a time-varying state network based on characteristic data of the detection points in the maintenance record data, wherein the time-varying state network includes the relationship between the detection points in the transmission line and the change of the state of the transmission line over time; The second determination unit is used to determine a target detection model based on the candidate states corresponding to the time-varying state network and the transmission line, and the target detection model is used to detect the state of the transmission line.

10. A device for determining a state of a power transmission line, comprising: A second acquisition unit is used to obtain monitoring data of a corresponding detection point of the transmission line; A processing unit, configured to process the monitoring data based on a target detection model to obtain a target state of the transmission line; The generation process of the target processing model includes: obtaining maintenance record data of the transmission line within a target time period, wherein the target time period represents the time period before the current moment; determining a time-varying state network based on the characteristic data of the detection points in the maintenance record data, wherein the time-varying state network includes the relationship between the various detection points in the transmission line and the state of the transmission line changing over time; and determining a target detection model based on the candidate states corresponding to the time-varying state network and the transmission line.

Images