A Cross-Industry Multi-Source Fault Intelligent Assessment Method and System Based on Large Model

Abstract

This invention discloses a cross-industry multi-source fault intelligent assessment method and system based on a large-scale model, belonging to the field of large-scale model technology. The method includes: retrieving historical fault records of transmission lines; deploying sensors on the transmission lines to obtain target data corresponding to each fault location and physical type; obtaining the associated physical type and association function corresponding to each fault type; extracting all detection points on the target transmission line to obtain the first target value of each detection point; extracting restricted areas on the transmission line to obtain the second target value of each detection point; setting weights to obtain the total target value of each detection point; and obtaining the first and second discrimination criteria to determine whether to issue an early warning for the selected sampling points. This invention analyzes fault records and restricted areas to make the current sampling points more reasonable and representative, intelligently assessing transmission line faults and effectively improving the reliability of transmission line assessment.

Description

Technical Field

[0001] This invention relates to the field of large model technology, specifically to a cross-industry multi-source fault intelligent assessment method and system based on large models. Background Technology

[0002] The safety of power transmission lines is a core prerequisite for ensuring stable, efficient, and accident-free operation of the power transmission process. It directly relates to the reliable transmission of electricity from the generation side to the load side and the power use by people in various industries. In actual transmission line fault detection, because transmission lines are usually long, comprehensive and continuous inspection of the entire line is costly, difficult, and time-consuming. Therefore, sampling points are used for detection. These sampling points are first collected by means of drones, and then handed over to staff for defect identification. This enables timely detection, accurate location, and analysis of faults in the transmission line, and rapid implementation of repair measures to ensure the safe and stable operation of the power grid. However, in the actual sampling process, if sampling points are randomly selected without fully considering the location and degree of defect of each sampling point, the selection of sampling points will not be targeted or representative enough, making it difficult to make timely and effective reliable judgments on the transmission line. Summary of the Invention

[0003] The purpose of this invention is to provide a cross-industry multi-source fault intelligent assessment method and system based on a large model, so as to solve the problems raised in the prior art.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0005] A cross-industry multi-source fault intelligent assessment method based on a large model includes the following steps:

[0006] Retrieve historical fault records of transmission lines and extract the fault time, fault location, and fault type from the fault records;

[0007] Sensors of different physical types are pre-deployed at different locations on the power transmission line to obtain the location of the sensors and historical sensing data, and to obtain the target data corresponding to each physical type for each fault location.

[0008] Obtain the normal value range monitored by the sensor and the defect level of the fault location, and obtain the degree of correlation between each fault type and each physical type based on the target data, and then obtain the associated physical type and correlation function corresponding to each fault type;

[0009] Extract all detection points on the target transmission line, and obtain the target data for each detection point in each physical type based on the location of the sensor and the detection point; obtain the first target value of the detection point based on the target data, the associated physical type, and the association function.

[0010] A three-dimensional model of the transmission line is established, and the controlled areas in the transmission line that restrict personnel access are extracted. Based on the time when personnel enter the controlled areas in historical monitoring videos, the second target value of each detection point is obtained.

[0011] By setting the weights of the first and second target values, the total target value for each detection point is obtained.

[0012] According to the current sampling method, all sampling points are selected from the detection points. Based on the location of each detection point and sampling point, a first discrimination criterion is obtained. Based on the total target value of each detection point and sampling point, a second discrimination criterion is obtained. Based on the first and second discrimination criteria, it is determined whether to issue an early warning to the currently selected sampling point.

[0013] Preferably, the target data corresponding to each fault location in each physical type is obtained, including:

[0014] Extract the fault location P, fault time and fault type of a fault record. Take the range of radius r around the fault location P as the detection range. Obtain all sensors monitoring a certain physical type m within the detection range. Take the sensor data of the sensor closest to the fault location P in the past several days as the target data corresponding to the fault location P in physical type m.

[0015] If there is no sensor monitoring a certain physical type n within the detection range, acquire all sensors monitoring physical type n in the transmission line, take the reciprocal of the distance between the location of each sensor and the fault location P as the characteristic distance corresponding to each sensor, add all characteristic distances to obtain the characteristic value, and divide each characteristic distance by the characteristic value to obtain the distance weight corresponding to each sensor; acquire the sensing value of each sensor at each moment over several historical days, and obtain the target data corresponding to physical type n at fault location P based on the distance weight.

[0016] Since the closer the distance, the closer the fault location P is to the sensor value, the reciprocal of the distance is used as the feature distance to calculate the distance weight for each sensor. Then, based on the sensor values ​​and distance weights of several sensors at each moment, the sensor value of the fault location at each moment is obtained, and thus the target data is obtained.

[0017] Preferably, the associated physical type and associated function corresponding to each fault type are obtained, including:

[0018] Obtain the normal value range monitored by the sensor for a certain physical type q, extract the target data Dte corresponding to the fault location P in physical type q, take the fault type corresponding to the fault location P as h, and take the ratio of the number of times the sensor value in the target data Dte exceeds the normal value range to the total number of times in the target data Dte as the target ratio of the fault location P in fault type h and physical type q.

[0019] Establish a neural network model, obtain several fault location datasets belonging to fault type h but with different defect levels, and train the neural network model; substitute the fault location P into the trained neural network model to obtain the target defect level of fault location P in fault type h.

[0020] Based on historical fault records, target ratios and target defect levels corresponding to several fault types h and physical types q are obtained. A linear function is fitted to show the change of target defect level with target ratio. The goodness of fit during the fitting process is taken as the degree of correlation between fault type h and physical type q. Then, the degree of correlation between fault type h and each physical type is obtained. The physical type with the highest degree of correlation is taken as the associated physical type of fault type h. The corresponding linear function is taken as the correlation function of fault type h.

[0021] Goodness of fit R 2 As a core evaluation metric for function fitting, the goodness-of-fit R-squared is used to quantify the regression model's ability to interpret observed data, i.e., the degree of agreement between the model's predicted values ​​and the actual observed values. 2 The larger the value, the more closely the corresponding fault type matches the physical type in a causal relationship.

[0022] Preferably, obtaining the first target value of the detection point includes: extracting all detection points on the target transmission line, obtaining the corresponding detection range based on the position of a certain detection point DP, and obtaining the target data of the detection point DP for each physical type based on the detection range and the position of each sensor.

[0023] Obtain the target ratio in the target data corresponding to a certain physical type w, and then, based on the associated physical type and association function corresponding to each fault type, obtain the fault level of the detection point DP in each fault type, and thus obtain the first target value of the detection point DP: Where D is the number of fault types, e is the natural constant, and R is the number of fault types. d Let d be the fault level of the d-th fault type; then, the first target value of each detection point can be obtained.

[0024] Preferably, obtaining the total target value of each detection point includes: establishing a three-dimensional model of the transmission line, obtaining the control area restricting personnel entry corresponding to each detection point according to the location of each detection point; acquiring the monitoring videos within a number of historical days, obtaining the duration of personnel entering the control area corresponding to each detection point, and normalizing the duration to obtain the second target value of each detection point; setting the weights of the first target value and the second target value, and obtaining the total target value of each detection point according to the first target value and the second target value of each detection point.

[0025] Preferably, determining whether to give an early warning prompt for the currently selected sampling point includes:

[0026] Obtaining the positions of all current sampling points, extracting the sampling point b closest to a certain sampling point a among them, taking the distance between the sampling point a and the sampling point b as the target distance of the sampling point a, and marking the sampling point a; obtaining the unmarked sampling point c closest to the sampling point b, taking the distance between the sampling point b and the sampling point c as the target distance of the sampling point b, and marking the sampling point b as well, and so on, obtaining the target distance of each sampling point, and calculating the average value L1; then, according to the positions of each detection point, obtaining the target distance of each detection point, and calculating the average value L2;

[0027] Calculating the average value P1 according to the total target value of each sampling point, calculating the average value P2 according to the total target value of each detection point, if there is L1 < k1 * L2 or P1 < k2 * P2, where k1 and k2 are the first ratio and the second ratio respectively, then give an early warning prompt for the currently selected sampling point.

[0028] When the sampling points are too dense, it means that the sampling points are not set reasonably. Therefore, in this solution, L1 < k1 * L2 is used to compare the sampling points with the detection points as the first discrimination criterion; when the total target values of each sampling point are too small, it means that the sampling points are not set reasonably. Therefore, in this solution, P1 < k2 * P2 is used to compare the sampling points with the detection points as the second discrimination criterion, and combining the first discrimination criterion and the second discrimination criterion, determine whether to give an early warning prompt for the currently selected sampling point.

[0029] The cross-industry multi-source fault intelligent judgment system based on the large model includes a target data acquisition module, a first target value calculation module, a total target value calculation module, and an early warning prompt module;

[0030] The target data acquisition module: is used to retrieve the historical fault records of the transmission line, extract the fault time, fault location, and fault type of the fault records; deploy sensors for monitoring different physical types at different positions on the transmission line in advance, obtain the positions of the sensors and the historical sensing data, and obtain the target data corresponding to each physical type at each fault location;

[0031] The first target value calculation module is used to obtain the normal value range monitored by the sensor and the defect level of the fault location, and to obtain the degree of correlation between each fault type and each physical type based on the target data, thereby obtaining the associated physical type and correlation function corresponding to each fault type.

[0032] Extract all detection points on the target transmission line, and obtain the target data for each detection point in each physical type based on the location of the sensor and the detection point; obtain the first target value of the detection point based on the target data, the associated physical type, and the association function.

[0033] Total target value calculation module: used to build a three-dimensional model of the transmission line, extract the control area in the transmission line where personnel are restricted from entering, obtain the second target value for each detection point based on the time when personnel enter the control area in historical monitoring videos; set the weights of the first target value and the second target value to obtain the total target value for each detection point;

[0034] Early warning module: It is used to select all sampling points from the detection points according to the current sampling method, obtain the first discrimination criterion based on the position of each detection point and sampling point, obtain the second discrimination criterion based on the total target value of each detection point and sampling point, and determine whether to issue an early warning to the currently selected sampling point according to the first discrimination criterion and the second discrimination criterion.

[0035] Preferably, the target data acquisition module includes a target data acquisition unit;

[0036] The target data acquisition unit is used to extract the fault location, fault time, and fault type of a fault record, obtain the detection range, determine whether there is a sensor monitoring a certain physical type within the detection range, and obtain the target data corresponding to each fault location for each physical type based on the sensor's sensing data over several historical days and the sensor's location.

[0037] Preferably, the total target value calculation module includes a second target value calculation unit and a total target value calculation unit;

[0038] The second target value calculation unit is used to establish a three-dimensional model of the transmission line, obtain the control area for restricted personnel access corresponding to each detection point based on the location of each detection point, acquire monitoring videos from several historical days, obtain the duration of personnel entering the control area corresponding to each detection point, normalize the duration, and obtain the second target value for each detection point.

[0039] Total target value calculation unit: used to set the weights of the first target value and the second target value, and to obtain the total target value of each detection point based on the first target value and the second target value of each detection point.

[0040] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention provides a cross-industry multi-source fault intelligent judgment method and system based on a large model, including: retrieving historical fault records of transmission lines; deploying sensors on the transmission lines to obtain target data corresponding to each fault location in each physical type; obtaining the associated physical type and association function corresponding to each fault type; extracting all detection points on the target transmission line to obtain the first target value of the detection points; extracting the control areas in the transmission line where personnel access is restricted to obtain the second target value of each detection point; setting weights to obtain the total target value of each detection point; obtaining the first and second discrimination criteria to determine whether to issue an early warning to the selected sampling points. This invention, by analyzing fault records and control areas, makes the current sampling points more reasonable and representative, intelligently judges transmission line faults, and effectively improves the reliability of transmission line judgment. Attached Figure Description

[0041] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 This is a flowchart illustrating the intelligent assessment method for cross-industry multi-source faults based on a large model, as described in this invention. Detailed Implementation

[0043] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0044] Example: Figure 1 As shown, this invention provides a technical solution for intelligent cross-industry multi-source fault assessment based on a large model, including the following steps:

[0045] 1. Retrieve historical fault records of transmission lines and extract the fault time, fault location, and fault type from the fault records; pre-deploy sensors at different locations on the transmission lines to monitor different physical types, obtain the sensor locations and historical sensing data, and obtain the target data corresponding to each fault location for each physical type.

[0046] Extract the fault location P, fault time and fault type of a fault record. Take the range of radius r around the fault location P as the detection range. Obtain all sensors monitoring a certain physical type m within the detection range. Take the sensor data of the sensor closest to the fault location P in the past several days as the target data corresponding to the fault location P in physical type m.

[0047] If there is no sensor monitoring a certain physical type n within the detection range, acquire all sensors monitoring physical type n in the transmission line, take the reciprocal of the distance between the location of each sensor and the fault location P as the characteristic distance corresponding to each sensor, add all characteristic distances to obtain the characteristic value, and divide each characteristic distance by the characteristic value to obtain the distance weight corresponding to each sensor; acquire the sensing value of each sensor at each moment over several historical days, and obtain the target data corresponding to physical type n at fault location P based on the distance weight.

[0048] In this embodiment, the fault types include wire strand breakage, corrosion, and overheating, while the physical types include wind, humidity, and temperature. Each fault type has a corresponding physical type. For example, when wind is frequently strong, wires are prone to strand breakage but not corrosion. Therefore, the purpose of obtaining the target data in this solution is to find the associated physical type for each fault type and to obtain the correlation function between the two. The specific steps are as follows:

[0049] 2. Obtain the normal value range monitored by the sensor and the defect level of the fault location, and obtain the correlation between each fault type and each physical type based on the target data, and then obtain the associated physical type and correlation function corresponding to each fault type.

[0050] Obtain the normal value range monitored by the sensor for a certain physical type q, extract the target data Dte corresponding to the fault location P in physical type q, take the fault type corresponding to the fault location P as h, and take the ratio of the number of times the sensor value in the target data Dte exceeds the normal value range to the total number of times in the target data Dte as the target ratio of the fault location P in fault type h and physical type q.

[0051] Establish a neural network model, obtain several fault location datasets belonging to fault type h but with different defect levels, and train the neural network model; substitute the fault location P into the trained neural network model to obtain the target defect level of fault location P in fault type h.

[0052] Here, fault locations with different defect levels are defined as follows: the higher the defect level, the more severe the defect. A neural network model is established. In this embodiment, the training dataset and the validation dataset are divided into a 7:3 ratio to train the neural network model. The specific establishment and validation process will not be elaborated here. Substituting the fault location P into the finally trained neural network model, the target defect level of the fault location P in fault type h can be obtained.

[0053] Based on historical fault records, target ratios and target defect levels corresponding to several fault types h and physical types q are obtained. A linear function is fitted to show the change of target defect level with target ratio. The goodness of fit during the fitting process is taken as the degree of correlation between fault type h and physical type q. Then, the degree of correlation between fault type h and each physical type is obtained. The physical type with the highest degree of correlation is taken as the associated physical type of fault type h. The corresponding linear function is taken as the correlation function of fault type h.

[0054] 3. Extract all detection points on the target transmission line, and obtain the target data for each detection point in each physical type based on the location of the sensor and the detection point; obtain the first target value of the detection point based on the target data, the associated physical type and the association function.

[0055] Extract all detection points on the target transmission line, obtain the corresponding detection range based on the location of a certain detection point DP, and obtain the target data corresponding to the detection point DP for each physical type based on the detection range and the location of each sensor.

[0056] Obtain the target ratio in the target data corresponding to a certain physical type w, and then, based on the associated physical type and association function corresponding to each fault type, obtain the fault level of the detection point DP in each fault type, and thus obtain the first target value of the detection point DP: Where D is the number of fault types, e is the natural constant, and R is the number of fault types. d Let d be the fault level of the d-th fault type; then, the first target value for each detection point can be obtained.

[0057] Since not all sections of a transmission line are meaningful for testing, the testing points on the transmission line are the points that have a significant impact on the transmission of the line. Sampling points extracted from the testing points are more reliable. Therefore, this scheme needs to analyze the testing points to obtain the first target value and the second target value, and then obtain the total target value of the testing points. Based on the total target value of the testing points and the total target value of the selected sampling points, it is then determined whether the selected sampling points are reasonable.

[0058] 4. Establish a three-dimensional model of the transmission line, extract the control area in the transmission line where personnel are restricted from entering, and obtain the second target value for each detection point based on the time when personnel enter the control area in historical monitoring videos; set the weights of the first and second target values ​​to obtain the total target value for each detection point.

[0059] Establish a three-dimensional model of the transmission line, and obtain the control area corresponding to each detection point that restricts personnel access based on the location of each detection point; acquire monitoring videos from several historical days to obtain the duration of personnel entering the control area corresponding to each detection point, and normalize the duration to obtain the second target value for each detection point.

[0060] Set the weights of the first target value and the second target value, and obtain the total target value of each detection point based on the first target value and the second target value of each detection point.

[0061] In this embodiment, the weights of the first target value and the second target value are set to W1 and W2, respectively, with W1 + W2 = 1. Based on the first target value X1 and the second target value X2 of the detection point, the total target value X of the detection point is obtained as X = W1 × X1 + W2 × X2. The first target value represents the situation where a defect accident occurs at the detection point, and the second target value represents the situation where a dangerous accident occurs in the controlled area of ​​the detection point. Furthermore, the larger the first target value and the second target value, the greater the regulatory effort should be placed on the detection point. Therefore, the larger the total target value, the greater the regulatory effort should be placed on the detection point.

[0062] 5. Based on the current sampling method, select all sampling points from the detection points. Based on the location of each detection point and sampling point, obtain the first discrimination criterion. Based on the total target value of each detection point and sampling point, obtain the second discrimination criterion. Based on the first and second discrimination criteria, determine whether to issue an early warning to the currently selected sampling point.

[0063] Obtain the positions of all current sampling points, extract the sampling point b that is closest to a given sampling point a, and use the distance between sampling point a and sampling point b as the target distance of sampling point a, and mark sampling point a; obtain the unmarked sampling point c that is closest to sampling point b, use the distance between sampling point b and sampling point c as the target distance of sampling point b, and mark sampling point b as well, and so on, to obtain the target distance of each sampling point, and calculate the average value L1; then, based on the position of each detection point, obtain the target distance of each detection point, and calculate the average value L2;

[0064] Calculate the average value P1 according to the total target value of each sampling point, and calculate the average value P2 according to the total target value of each detection point. If there is L1 < k1 * L2 or P1 < k2 * P2, where k1 and k2 are the first ratio and the second ratio respectively, then give a warning prompt for the currently selected sampling point.

[0065] This embodiment also provides a cross-industry multi-source fault intelligent judgment system based on a large model, including a target data obtaining module, a first target value calculation module, a total target value calculation module, and a warning prompt module. The target data obtaining module includes a target data obtaining unit, and the total target value calculation module includes a second target value calculation unit and a total target value calculation unit; when the system executes a computer program, it implements the above-mentioned cross-industry multi-source fault intelligent judgment method based on a large model. Since this cross-industry multi-source fault intelligent judgment method based on a large model has been introduced in detail above, it will not be elaborated here.

[0066] It should be noted that: the above sequence of the embodiments of the present invention is only for description and does not represent the advantages or disadvantages of the embodiments. And the above description of specific embodiments of this specification is made. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in a different order than in the embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0067] Each embodiment in this specification is described in a progressive manner. The same or similar parts between each embodiment can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0068] The above are only the preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A cross-industry multi-source fault intelligent assessment method based on a large model, characterized in that, Includes the following steps: Retrieve historical fault records of transmission lines and extract the fault time, fault location, and fault type from the fault records; Sensors of different physical types are pre-deployed at different locations on the power transmission line to obtain the location of the sensors and historical sensing data, and to obtain the target data corresponding to each physical type for each fault location. Obtain the normal value range monitored by the sensor and the defect level of the fault location, and obtain the degree of correlation between each fault type and each physical type based on the target data, and then obtain the associated physical type and correlation function corresponding to each fault type; Extract all detection points on the target transmission line, and obtain the target data for each detection point in each physical type based on the location of the sensor and the detection point; obtain the first target value of the detection point based on the target data, the associated physical type, and the association function. A three-dimensional model of the transmission line is established, and the controlled areas in the transmission line that restrict personnel access are extracted. Based on the time when personnel enter the controlled areas in historical monitoring videos, the second target value of each detection point is obtained. By setting the weights of the first and second target values, the total target value for each detection point is obtained. According to the current sampling method, all sampling points are selected from the detection points. Based on the location of each detection point and sampling point, a first discrimination criterion is obtained. Based on the total target value of each detection point and sampling point, a second discrimination criterion is obtained. Based on the first and second discrimination criteria, it is determined whether to issue an early warning to the currently selected sampling point.

2. The intelligent cross-industry multi-source fault assessment method based on a large model according to claim 1, characterized in that, Obtain the target data for each fault location in each physical type, including: Extract the fault location P, fault time and fault type of a fault record. Take the range of radius r around the fault location P as the detection range. Obtain all sensors monitoring a certain physical type m within the detection range. Take the sensor data of the sensor closest to the fault location P in the past several days as the target data corresponding to the fault location P in physical type m. If no sensor monitoring a certain physical type n is found within the detection range, acquire all sensors monitoring physical type n in the transmission line, take the reciprocal of the distance between the location of each sensor and the fault location P as the characteristic distance corresponding to each sensor, add all characteristic distances to obtain the characteristic value, and divide each characteristic distance by the characteristic value to obtain the distance weight corresponding to each sensor; acquire the sensing value of each sensor at each moment over several historical days, and obtain the target data corresponding to physical type n at fault location P based on the distance weight.

3. The intelligent cross-industry multi-source fault assessment method based on a large model according to claim 2, characterized in that, Obtain the associated physical type and associated function for each fault type, including: Obtain the normal value range monitored by the sensor for a certain physical type q, extract the target data Dte corresponding to the fault location P in physical type q, take the fault type corresponding to the fault location P as h, and take the ratio of the number of times the sensor value in the target data Dte exceeds the normal value range to the total number of times in the target data Dte as the target ratio of the fault location P in fault type h and physical type q. Build a neural network model, obtain a number of fault location datasets belonging to fault type h but with different defect levels, and train the neural network model; substitute the fault location P into the trained neural network model to obtain the target defect level of the fault location P in fault type h. According to the historical fault records, obtain the target ratios and target defect levels corresponding to a number of fault types h and physical types q, and perform a linear function fitting of the change of the target defect level with the target ratio. Take the goodness of fit corresponding to the fitting process as the correlation degree between fault type h and physical type q; furthermore, obtain the correlation degrees between fault type h and each physical type, and take the physical type corresponding to the largest correlation degree as the associated physical type of the fault type h, and take the corresponding linear function of the fitting as the associated function of the fault type h.

4. The intelligent cross-industry multi-source fault assessment method based on a large model according to claim 3, characterized in that, Obtain the first target value of the detection point, including: extract all detection points on the target transmission line, obtain the corresponding detection range according to the position of a certain detection point DP, and obtain the target data corresponding to each physical type of the detection point DP according to the detection range and the position of each sensor. The target ratio in the target data corresponding to a certain physical type w is obtained. Then, based on the associated physical type and association function corresponding to each fault type, the fault level of the detection point DP in each fault type is obtained, and thus the first target value of the detection point DP is obtained. Where D is the number of fault types, e is the natural constant, and R is the number of fault types. d Let d be the fault level of the d-th fault type; then, the first target value of each detection point can be obtained.

5. The intelligent cross-industry multi-source fault assessment method based on a large model according to claim 1, characterized in that, Obtain the total target value of each detection point, including: build a three-dimensional model of the transmission line, obtain the control area restricting personnel entry corresponding to each detection point according to the position where each detection point is located; obtain the monitoring videos in a number of historical days, obtain the duration of personnel entering the control area corresponding to each detection point, and normalize the duration to obtain the second target value of each detection point; set the weights of the first target value and the second target value, and obtain the total target value of each detection point according to the first target value and the second target value of each detection point.

6. The intelligent cross-industry multi-source fault assessment method based on a large model according to claim 1, characterized in that, Judge whether to give a warning prompt for the currently selected sampling point, including: Obtain the positions of all current sampling points, extract the sampling point b closest to a certain sampling point a, take the distance between the sampling point a and the sampling point b as the target distance of the sampling point a, and mark the sampling point a; obtain the unmarked sampling point c closest to the sampling point b, take the distance between the sampling point b and the sampling point c as the target distance of the sampling point b, and mark the sampling point b as well, and so on, to obtain the target distance of each sampling point and calculate the average value L1; then obtain the target distance of each detection point according to the position of each detection point and calculate the average value L2. Calculate the average value P1 according to the total target value of each sampling point, calculate the average value P2 according to the total target value of each detection point. If there is L1 < k1 * L2 or P1 < k2 * P2, where k1 and k2 are the first ratio and the second ratio respectively, then give a warning prompt for the currently selected sampling point.

7. A cross-industry multi-source fault intelligent assessment system, used to execute the cross-industry multi-source fault intelligent assessment method based on a large model as described in any one of claims 1-6, characterized in that, The system includes a target data obtaining module, a first target value calculation module, a total target value calculation module, and a warning prompt module. The target data obtaining module: is used to retrieve the historical fault records of the transmission line and extract the fault time, fault location, and fault type of the fault records. Sensors of different physical types are pre-deployed at different locations on the power transmission line to obtain the location of the sensors and historical sensing data, and to obtain the target data corresponding to each physical type for each fault location. The first target value calculation module is used to obtain the normal value range monitored by the sensor and the defect level of the fault location, and to obtain the degree of correlation between each fault type and each physical type based on the target data, thereby obtaining the associated physical type and correlation function corresponding to each fault type. Extract all detection points on the target transmission line, and obtain the target data for each detection point in each physical type based on the location of the sensor and the detection point. Based on the target data, associated physical type, and association function, the first target value of the detection point is obtained; Overall target value calculation module: used to build a three-dimensional model of the transmission line, extract the control area in the transmission line where personnel are restricted from entering, and obtain the second target value for each detection point based on the time when personnel enter the control area in historical monitoring videos; By setting the weights of the first and second target values, the total target value for each detection point is obtained. Early warning module: It is used to select all sampling points from the detection points according to the current sampling method, obtain the first discrimination criterion based on the position of each detection point and sampling point, obtain the second discrimination criterion based on the total target value of each detection point and sampling point, and determine whether to issue an early warning to the currently selected sampling point according to the first discrimination criterion and the second discrimination criterion.

8. The cross-industry multi-source fault intelligent analysis system according to claim 7, characterized in that, The target data acquisition module includes a target data acquisition unit; The target data acquisition unit is used to extract the fault location, fault time, and fault type of a fault record, obtain the detection range, determine whether there is a sensor monitoring a certain physical type within the detection range, and obtain the target data corresponding to each fault location for each physical type based on the sensor's sensing data over several historical days and the sensor's location.

9. The cross-industry multi-source fault intelligent analysis system according to claim 7, characterized in that, The total target value calculation module includes a second target value calculation unit and a total target value calculation unit; The second target value calculation unit is used to establish a three-dimensional model of the transmission line, obtain the control area for restricted personnel access corresponding to each detection point based on the location of each detection point, acquire monitoring videos from several historical days, obtain the duration of personnel entering the control area corresponding to each detection point, normalize the duration, and obtain the second target value for each detection point. Total target value calculation unit: used to set the weights of the first target value and the second target value, and to obtain the total target value of each detection point based on the first target value and the second target value of each detection point.

Images