Regional tower position abnormity intelligent detection method, device and system based on Beidou differential positioning technology

Abstract

The invention relates to the technical field of intelligent power grids, in particular to an area tower position abnormity intelligent detection method, device and system, and the method is used for carrying out the abnormity detection of the positions of a plurality of towers in a designated area. The method comprises the following steps: acquiring position information of the tower through a positioning terminal arranged at a specified position of the tower, wherein the position information comprises altitude and latitude and longitude coordinates of the positioning terminal; calculating the distance between every two towers according to the position information of the towers; generating a distance matrix based on the distance between every two towers; and calculating the abnormal displacement probability of each tower in the plurality of towers by using a trained artificial intelligence model according to the distance matrix. The method can automatically adapt to different environmental conditions or regional characteristics of the tower, improves the problem of false alarm or missing alarm in a complex geographical environment, effectively eliminates systematic errors of position information, and improves the reliability of anomaly detection.

Description

Technical Field

[0001] The present disclosure relates to the technical field of smart grids, and particularly to an intelligent detection method, device and system for abnormal positions of regional transmission towers. Background Art

[0002] Power line inspection is an important part of power system maintenance work. Traditional power inspection methods mainly rely on manual inspection or drone inspection. However, these methods have problems such as high cost, large workload, and low efficiency. Especially in areas with complex terrain and extensive line distribution, the inspection work is more difficult. To solve these problems, global navigation satellite systems (GNSS), such as the Beidou satellite navigation system (BDS), have gradually been applied to the field of power inspection. GNSS technology can collect the position information of power equipment or transmission towers, and cooperate with corresponding software to analyze the operation status of the line, thereby realizing the automatic inspection and monitoring of the power system. Currently, the existing abnormal detection methods have the following deficiencies:

[0003] Insufficient adaptability: Most of the existing technologies adopt fixed displacement thresholds. When the displacement of the transmission tower measured by the positioning technology exceeds the displacement threshold, it is considered that there is an abnormal displacement. However, the existing technologies cannot automatically adapt to different environmental conditions or regional characteristics, resulting in common false alarms or missed alarms in some complex geographical environments;

[0004] Limited positioning accuracy: Although positioning technologies such as the Beidou system have high-precision positioning capabilities, in the existing solutions, the positioning accuracy decreases in remote areas or environments with strong signal interference, making it difficult to ensure the reliability of abnormal detection. Summary of the Invention

[0005] To solve the problems in the related technologies, embodiments of the present disclosure provide an intelligent detection method, device and system for abnormal positions of regional transmission towers.

[0006] In a first aspect, embodiments of the present disclosure provide an intelligent detection method for abnormal positions of regional transmission towers, which is used to detect abnormal positions of multiple transmission towers in a specified area. The method includes:

[0007] Obtaining the position information of the transmission tower through a positioning terminal set at a specified position of the transmission tower, where the position information includes the altitude and longitude and latitude coordinates of the positioning terminal;

[0008] Calculating the distances between every two of the multiple transmission towers according to the position information of the multiple transmission towers;

[0009] Generating a distance matrix based on the distances between every two of the multiple transmission towers;

[0010] According to the distance matrix, the abnormal displacement probability of each of the multiple towers is calculated using a trained artificial intelligence model, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the multiple towers, and the artificial intelligence model includes an attention layer and a deep neural network. The abnormal displacement probability of each of the multiple towers is calculated using the trained artificial intelligence model, including:

[0011] Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix;

[0012] The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

[0013] According to an embodiment of the present disclosure, the distance matrix is:

[0014]

[0015] Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

[0016] According to an embodiment of the present disclosure, inputting the distance matrix into the attention layer to obtain an attention weighted distance matrix includes:

[0017] Calculate Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer;

[0018] Calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product;

[0019] By combining h with Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

[0020] According to an embodiment of the present disclosure, inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes:

[0021] The attention weighted distance matrix is ​​input into the deep neural network to obtain an N-dimensional vector O=[O 1 ,O 2 ,…O n ,…,O N ], where O 1 ,O 2 ,O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

[0022] According to an embodiment of the present disclosure, the deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

[0023] According to the embodiments of the present disclosure, <O n ≤1, when 0 <O n <0.5, the tower T n There is no abnormal displacement, when 0.5≤O n When ≤1, it is considered that the tower T n There is abnormal displacement.

[0024] According to an embodiment of the present disclosure, the loss function J of the artificial intelligence model is:

[0025] Wherein, N is the number of the plurality of towers,

[0026] The loss function is used to train the artificial intelligence model to optimize the weights of the attention layer and the parameters of the deep neural network.

[0027] According to an embodiment of the present disclosure, it also includes: when the number or installation positions of the plurality of pole towers in the designated area changes, retraining the artificial intelligence model according to new position information of the plurality of pole towers.

[0028] According to an embodiment of the present disclosure, the method further includes:

[0029] When the probability of abnormal displacement of any tower continues to increase, an alarm message is output.

[0030] According to an embodiment of the present disclosure, the method further includes:

[0031] Counting the probability change speed and amplitude of abnormal displacement of the tower;

[0032] A corresponding relationship between the probability change speed and amplitude of abnormal displacement of the pole tower and the characteristics of the pole tower itself and the environmental characteristics is established as reference data for installation and maintenance operations of the pole tower.

[0033] According to an embodiment of the present disclosure, the method further includes:

[0034] Constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement;

[0035] The artificial intelligence model is trained based on the sample distance matrix and the corresponding sample label vector.

[0036] According to an embodiment of the present disclosure, the normal position information includes: position information manually set according to the reference position information of the tower and the preset normal displacement condition, and / or position information acquired by the corresponding positioning terminal when the tower is actually found to be in the normal position;

[0037] The abnormal position information includes: position information manually set according to the reference position information of the tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the abnormal displacement of the tower is actually found.

[0038] According to an embodiment of the present disclosure, constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes:

[0039] For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower;

[0040] For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower;

[0041] Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UPns one;

[0042] For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector corresponding to the sample matrix OS=[OS 1 ,OS 2 ,…OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

[0043] In a second aspect, an embodiment of the present disclosure provides a regional tower position abnormality intelligent detection device, which is used to detect abnormalities in the positions of multiple towers in a specified area, and the device includes:

[0044] An acquisition module is configured to acquire the location information of the pole tower through a positioning terminal arranged at a designated position of the pole tower, wherein the location information includes the altitude and longitude and latitude coordinates of the positioning terminal;

[0045] A first calculation module is configured to calculate the distance between two of the plurality of towers according to the position information of the plurality of towers;

[0046] A generating module, configured to generate a distance matrix based on the distances between the plurality of towers;

[0047] The second calculation module is configured to calculate the abnormal displacement probability of each of the plurality of towers using a trained artificial intelligence model according to the distance matrix, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the plurality of towers, the artificial intelligence model comprising an attention layer and a deep neural network, and the use of the trained artificial intelligence model to calculate the abnormal displacement probability of each of the plurality of towers comprises:

[0048] Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix;

[0049] The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

[0050] According to an embodiment of the present disclosure, the distance matrix is as follows:

[0051]

[0052] where h is the distance matrix, N is the number of the plurality of poles and towers, h 11 is zero, h N1 is the distance between the positioning terminal of the Nth pole and tower and the positioning terminal of the 1st pole and tower, h 1N is the distance between the positioning terminal of the 1st pole and tower and the positioning terminal of the Nth pole and tower, h NN is zero.

[0053] According to an embodiment of the present disclosure, the step of inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix includes:

[0054] Calculating Q = hW Q , K = hW K , W Q and W K are the weights of the attention layer;

[0055] Calculating the normalized weight Softmax(QK T ), where QK T is the dot product of Q and K T ;

[0056] Calculating the attention-weighted distance matrix by multiplying h with Softmax(QK T ).

[0057] According to an embodiment of the present disclosure, the step of inputting the attention-weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each pole and tower includes:

[0058] Inputting the attention-weighted distance matrix into the deep neural network to obtain an N-dimensional vector O = [O 1 , O 2 , … O n , …, O N , where O 1 , O 2 , O n , O N respectively represent the probabilities of abnormal displacement of the 1st, 2nd, nth, and Nth poles and towers.

[0059] According to an embodiment of the present disclosure, the deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

[0060] According to an embodiment of the present disclosure, 0 < O n≤1 when 0 < O n <0.5, it is considered that the pole tower T n has no abnormal displacement. When 0.5 ≤ O n ≤1, it is considered that the pole tower T n has abnormal displacement.

[0061] According to an embodiment of the present disclosure, the loss function J of the artificial intelligence model is:

[0062] where N is the number of the multiple pole towers, and the loss function is used to train the artificial intelligence model to optimize the weights of the attention layer and the parameters of the deep neural network.

[0063] According to an embodiment of the present disclosure, it further includes:

[0064] A retraining module configured to retrain the artificial intelligence model according to the new position information of the multiple pole towers when the number or installation position of the multiple pole towers in the specified area changes;

[0065] An alarm module configured to output an alarm message when the abnormal displacement probability of any pole tower continuously increases;

[0066] A statistics module configured to statistically calculate the change speed and amplitude of the abnormal displacement probability of the pole tower, and establish a correspondence between the change speed and amplitude of the abnormal displacement probability of the pole tower and the self-characteristics and environmental characteristics of the pole tower as reference data for the installation and maintenance operations of the pole tower.

[0067] According to an embodiment of the present disclosure, it further includes:

[0068] A construction module configured to construct a sample distance matrix based on the normal position information of each pole tower in the multiple pole towers when in a normal position and the abnormal position information when having abnormal displacement, and construct a sample label vector corresponding to the sample distance matrix based on whether each pole tower in the multiple pole towers has abnormal displacement;

[0069] A training module configured to train the artificial intelligence model based on the sample distance matrix and the corresponding sample label vector.

[0070] According to an embodiment of the present disclosure, the normal position information includes: position information manually set according to the reference position information of the pole tower and a preset normal displacement condition, and / or position information obtained by the corresponding positioning terminal when it is actually found that the pole tower is in a normal position;

[0071] The abnormal position information includes: position information manually set according to the reference position information of the tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the abnormal displacement of the tower is actually found.

[0072] According to an embodiment of the present disclosure, constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes:

[0073] For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower;

[0074] For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower;

[0075] Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UP ns one;

[0076] For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector corresponding to the sample matrix OS=[OS 1 ,OS 2 ,…OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

[0077] In a third aspect, a monitoring device is provided in an embodiment of the present disclosure, comprising a memory and a processor; wherein the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement any method described in the first aspect.

[0078] In a fourth aspect, an embodiment of the present disclosure provides a regional tower position abnormality intelligent detection system for detecting abnormalities in the positions of multiple towers in a specified area, wherein the system includes multiple positioning terminals and the monitoring device according to the third aspect, wherein:

[0079] The multiple positioning terminals are respectively arranged at designated positions of the multiple towers;

[0080] The positioning terminal obtains the position information of the corresponding tower and sends the position information to the monitoring device, wherein the position information includes the altitude and longitude and latitude coordinates of the positioning terminal.

[0081] In a fifth aspect, an embodiment of the present disclosure provides a computer-readable storage medium on which computer instructions are stored. When the computer instructions are executed by a processor, the method described in any one of the first aspects is implemented.

[0082] According to the technical solution provided by the embodiment of the present disclosure, a method, device and system for intelligent detection of regional tower position anomalies are provided. The method is used to detect anomalies in the positions of multiple towers in a specified area. The method includes: obtaining the position information of the tower through a positioning terminal set at a specified position of the tower, and the position information includes the altitude and longitude and latitude coordinates of the positioning terminal; calculating the distance between the multiple towers according to the position information of the multiple towers; generating a distance matrix based on the distance between the multiple towers; and calculating the abnormal displacement probability of each tower in the multiple towers using a trained artificial intelligence model according to the distance matrix. The present disclosure is based on using the distance data between the multiple towers to calculate the abnormal displacement probability with a trained deep neural network model based on the attention mechanism, which does not rely on a fixed displacement threshold. By using the sample data generated according to the position information of the multiple towers for model training, it can automatically adapt to the different environmental conditions or regional characteristics of the towers, and can especially improve the problem of false alarms or missed alarms in complex geographical environments. In addition, by using the distances between multiple towers to generate a distance matrix for detection, the systematic errors of location information can be effectively eliminated and the reliability of anomaly detection can be improved in remote areas or environments with large signal interference.

[0083] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0084] Other features, objectives and advantages of the present disclosure will become more apparent through the following detailed description of non-limiting embodiments in conjunction with the accompanying drawings. In the accompanying drawings:

[0085] Figure 1A schematic diagram showing the structure of a regional tower position abnormality intelligent detection system according to an embodiment of the present disclosure is shown;

[0086] Figure 2 A schematic diagram showing the use of a trained artificial intelligence model to calculate the probability of abnormal displacement of each of the plurality of towers according to an embodiment of the present disclosure;

[0087] Figure 3 A flow chart of a method for intelligently detecting abnormality of regional tower positions according to an embodiment of the present disclosure is shown;

[0088] Figure 4 A structural block diagram of an intelligent detection device for abnormal regional tower position according to an embodiment of the present disclosure is shown;

[0089] Figure 5 A structural block diagram of an electronic device according to an embodiment of the present disclosure is shown. DETAILED DESCRIPTION

[0090] Hereinafter, exemplary embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement them. In addition, for the sake of clarity, parts not related to the description of the exemplary embodiments are omitted in the accompanying drawings.

[0091] In the present disclosure, it should be understood that terms such as "include" or "have" are intended to indicate the presence of features, numbers, steps, behaviors, components, parts, or a combination thereof disclosed in the present specification, and are not intended to exclude the possibility that one or more other features, numbers, steps, behaviors, components, parts, or a combination thereof exist or are added.

[0092] It should also be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0093] As mentioned above, most existing technologies use a fixed displacement threshold. When the tower displacement measured by positioning technology exceeds the displacement threshold, it is considered that abnormal displacement occurs. However, existing technologies cannot automatically adapt to different environmental conditions or regional characteristics, resulting in common false alarms or missed alarms in certain complex geographical environments. In addition, although positioning technologies such as the Beidou system have high-precision positioning capabilities, in existing solutions, the positioning accuracy is reduced in remote areas or environments with large signal interference, making it difficult to ensure the reliability of abnormal detection.

[0094] In order to solve the above problems in the prior art, the disclosed embodiment provides an efficient, intelligent and adaptive regional pole tower position abnormality intelligent detection system, which adopts an artificial intelligence algorithm, based on the distance data between the multiple pole towers, and uses a trained deep neural network model based on the attention mechanism to calculate the probability of abnormal displacement, which does not rely on a fixed displacement threshold. By using the sample distance matrix and sample label vector of the multiple pole towers for model training, it can automatically adapt to the different environmental conditions or regional characteristics of the towers, and can especially improve the problem of false alarms or missed alarms in complex geographical environments. In addition, by using the distance between the multiple pole towers to generate a distance matrix for detection, in remote areas or environments with large signal interference, the systematic error of the location information can be effectively eliminated, and the reliability of abnormality detection can be improved.

[0095] Figure 1 A schematic structural diagram of a regional tower position abnormality intelligent detection system according to an embodiment of the present disclosure is shown.

[0096] Figure 2 A schematic diagram showing the use of a trained artificial intelligence model to calculate the probability of abnormal displacement of each of the plurality of towers according to an embodiment of the present disclosure.

[0097] like Figure 1 and Figure 2 As shown, the system includes multiple positioning terminals and monitoring equipment. The multiple positioning terminals are respectively arranged at designated positions of the multiple pole towers, and are used to obtain the position information of the corresponding pole towers and send the position information to the monitoring equipment. The position information includes the altitude and longitude and latitude coordinates of the positioning terminal.

[0098] According to an embodiment of the present disclosure, the positioning terminal uses Beidou differential positioning technology to obtain the location information of the corresponding tower. Beidou differential positioning technology is a high-precision positioning technology based on the Beidou satellite navigation system. It uses a base station receiver and a mobile receiver with a known position to simultaneously receive satellite signals, and eliminates errors by comparing the differences between the two. Specifically, the position of the base station receiver is known and can be corrected by actual measurement, while the position of the mobile receiver needs to be measured, and its error can be eliminated by comparing it with the base station receiver. Beidou differential positioning technology has high precision and can achieve centimeter-level positioning accuracy. It is suitable for occasions that require higher positioning accuracy, such as navigation and surveying and mapping with precise positioning. At the same time, Beidou differential positioning technology also has high stability. Based on the differential positioning method, it can stably provide more accurate and continuous location information, avoiding errors caused by uncontrollable factors such as signal interference and weather changes.

[0099] According to an embodiment of the present disclosure, the monitoring device is configured to: calculate the distance between each of the multiple towers based on the position information of the multiple towers; generate a distance matrix based on the distance between each of the multiple towers; and calculate the abnormal displacement probability of each of the multiple towers using a trained artificial intelligence model based on the distance matrix, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample data generated according to the position information of the multiple towers, and the artificial intelligence model includes an attention layer and a deep neural network.

[0100] According to an embodiment of the present disclosure, the use of the trained artificial intelligence model to calculate the abnormal displacement probability of each of the multiple towers includes: inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix with normalized weights, and the normalized weights are obtained based on the distance matrix; inputting the attention-weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower.

[0101] Specifically, according to an embodiment of the present disclosure, the distance matrix is:

[0102]

[0103] Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

[0104] According to an embodiment of the present disclosure, the inputting the distance matrix into the attention layer to obtain the attention weighted distance matrix includes: calculating Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer; calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product; by adding h to Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

[0105] According to an embodiment of the present disclosure, the inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes: inputting the attention weighted distance matrix into the deep neural network to obtain an N-dimensional vector O=[O 1 ,O 2 ,…O n ,…,O N ], where O 1 ,O 2 ,O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

[0106] Each tower in the region has different environmental conditions (such as ground height, foundation stability, tower inclination, distance from other towers, density of nearby towers, vegetation type near towers, temperature, humidity, wind speed near towers, etc.) and positioning terminal signal quality, which have different impacts on the accuracy of the prediction results of the artificial intelligence model. These influencing factors are difficult to analyze qualitatively or quantitatively through mathematical modeling due to their large number and complex relationships. By introducing the attention weighting matrix into the artificial intelligence model and weighting the distance between towers, possible influencing factors can be automatically converted into weights of the artificial intelligence model through training, thereby obtaining more accurate prediction results.

[0107] According to an embodiment of the present disclosure, the deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

[0108] According to the embodiments of the present disclosure, <O n ≤1, when 0 <O n <0.5, it is considered that there is no abnormal displacement of the tower. n When ≤1, it is considered that the tower has abnormal displacement.

[0109] According to an embodiment of the present disclosure, the monitoring device includes an edge device arranged in the designated area. When the monitoring device finds a pole tower with abnormal displacement, the monitoring device reports the relevant information of the pole tower to a monitoring center server, and the monitoring center server is connected to monitoring devices in multiple designated areas; or, the monitoring device includes a monitoring center server, and the monitoring center server is connected to positioning terminals in multiple designated areas.

[0110] According to the embodiments of the present disclosure, by implementing the monitoring device as an edge device in a specified area, distributed detection of abnormal displacement of the tower can be achieved, reducing the load of the monitoring center server, and having higher detection efficiency and real-time performance.

[0111] According to an embodiment of the present disclosure, the monitoring device is also configured to: when the number or installation position of the multiple towers in the designated area changes, retrain the artificial intelligence model according to the new position information of the multiple towers.

[0112] The existing inspection process of power transmission line towers mainly uses GPS positioning as the main means, combined with drones, manual inspections and other means, and the main detection data is the settlement, displacement and other data of the towers themselves, ignoring the linkage changes between towers. These existing detection data are easy to deviate from the actual data in remote areas or environments with large signal interference, resulting in a decrease in positioning accuracy, making it difficult to ensure the reliability of abnormal detection, and unable to deeply analyze the trend of tower position changes or potential hidden dangers. The tower position abnormality detection system disclosed in the present invention calculates the distance between the multiple towers based on the position information of the multiple towers, and generates a distance matrix based on the distance between the multiple towers, which significantly improves the accuracy of power tower position monitoring and can detect small displacement changes. At the same time, the present invention introduces an artificial intelligence model that can automatically adapt to the different environmental conditions or regional characteristics of the towers, improving the reliability of abnormal detection and the efficiency of inspection.

[0113] According to an embodiment of the present disclosure, the positioning terminal can be arranged at the top of the tower. When the tower is displaced, the displacement at the top is usually greater than that at other positions of the tower body. Therefore, arranging the positioning terminal at the top of the tower helps to improve the sensitivity and accuracy of detection.

[0114] According to an embodiment of the present disclosure, the monitoring device is further configured to output an alarm message when the abnormal displacement probability of any tower indicates that the tower has abnormal displacement, or when the abnormal displacement probability of the tower continues to increase or abnormal fluctuation occurs.

[0115] The abnormal displacement detection of towers in the prior art can usually only give an alarm when abnormal displacement occurs, while the abnormal displacement detection system of the disclosed embodiment can detect hidden problems such as long-term slight tilt or gradual settlement of towers in advance according to the change in the probability of abnormal displacement of towers, thereby preventing problems before they occur, which is of great positive significance for effectively preventing power failures caused by tower collapse in severe weather such as rain, snow and frost.

[0116] According to an embodiment of the present disclosure, the monitoring device is also configured to train the artificial intelligence model, including: constructing a sample distance matrix based on normal position information of each of the multiple towers when it is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement; training the artificial intelligence model based on the sample distance matrix and the corresponding sample label vector.

[0117] According to an embodiment of the present disclosure, the normal position information includes: position information manually set according to the reference position information of the pole tower and the preset normal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to be in the normal position; the abnormal position information includes: position information manually set according to the reference position information of the pole tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to have abnormal displacement.

[0118] According to an embodiment of the present disclosure, constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes:

[0119] For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower;

[0120] For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower;

[0121] Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UP ns one;

[0122] For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector corresponding to the sample matrix OS=[OS 1 ,OS 2 ,…OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

[0123] In a specific example, the reference position information of the pole tower may be the position information obtained by the corresponding positioning terminal when the pole tower is installed. The position information may be considered as the position information of the pole tower in a normal position state, and may be used as a reference for subsequent judgment of whether the pole tower has abnormal displacement. Normal displacement conditions may be preset based on experience, for example, when the position information of the pole tower changes by a certain amount relative to the reference position information, it may be considered as normal displacement. Abnormal displacement conditions may also be preset based on experience, for example, when the position information of the pole tower changes by a certain amount relative to the reference position information, it may be considered as abnormal displacement.

[0124] In order to train the artificial intelligence model, a large amount of sample data is required. However, it is obviously unrealistic to artificially displace each tower in the area in order to obtain sample data. Therefore, some position information can be manually set according to the benchmark position information of the tower and the preset normal displacement conditions. The position information contains normal position information and abnormal position information, and the corresponding sample distance matrix and sample label vector are constructed based on the position information to train the artificial intelligence model.

[0125] Furthermore, in order to make the prediction results of the artificial intelligence model more in line with reality, the actual position information of the tower obtained daily and the labels of whether the tower has abnormal displacement can also be used as samples for model training to enhance the model's adaptability to the environment.

[0126] According to an embodiment of the present disclosure, the loss function J used for artificial intelligence model training is:

[0127]

[0128] Where N is the number of towers, OS n For the tower T n The sample label of O n For the tower T n prediction results.

[0129] Figure 3 A flow chart of a method for intelligently detecting abnormality in regional tower locations according to an embodiment of the present disclosure is shown.

[0130] like Figure 3 As shown, in an embodiment of the present disclosure, a method for intelligently detecting abnormality of a regional pole tower position is provided, which is used to detect abnormality of the positions of multiple pole towers in a specified area. The method for intelligently detecting abnormality of a regional pole tower position includes the following steps S310 to S340:

[0131] In step S310, the location information of the pole tower is obtained by a positioning terminal arranged at a designated position of the pole tower, and the location information includes the altitude and longitude and latitude coordinates of the positioning terminal.

[0132] In step S320, the distances between any two of the plurality of towers are calculated according to the position information of the plurality of towers.

[0133] In step S330, a distance matrix is ​​generated based on the distances between each of the plurality of towers.

[0134] In step S340, according to the distance matrix, the abnormal displacement probability of each of the multiple towers is calculated using the trained artificial intelligence model, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the multiple towers, and the artificial intelligence model includes an attention layer and a deep neural network. The use of the trained artificial intelligence model to calculate the abnormal displacement probability of each of the multiple towers includes:

[0135] Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix;

[0136] The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

[0137] According to an embodiment of the present disclosure, the distance matrix is:

[0138]

[0139] Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

[0140] According to an embodiment of the present disclosure, inputting the distance matrix into the attention layer to obtain an attention weighted distance matrix includes:

[0141] Calculate Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer;

[0142] Calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product;

[0143] By combining h with Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

[0144] According to an embodiment of the present disclosure, inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes:

[0145] The attention weighted distance matrix is ​​input into the deep neural network to obtain an N-dimensional vector O=[O 1 ,O 2 ,…O n ,…,O N ], where O 1 ,O 2 ,O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

[0146] According to an embodiment of the present disclosure, the deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

[0147] According to the embodiments of the present disclosure, <O n ≤1, when 0 <O n <0.5, it is considered that there is no abnormal displacement of the tower. n When ≤1, it is considered that the tower has abnormal displacement.

[0148] According to an embodiment of the present disclosure, the method further includes: when the probability of abnormal displacement of any tower continues to increase, outputting alarm information.

[0149] According to an embodiment of the present disclosure, the method further includes: counting the speed and amplitude of the probability change of abnormal displacement of the pole tower; establishing a corresponding relationship between the speed and amplitude of the probability change of abnormal displacement of the pole tower and the inherent characteristics and environmental characteristics of the pole tower as reference data for the installation and maintenance operations of the pole tower.

[0150] The abnormal displacement detection of towers in the prior art can usually only give an alarm when abnormal displacement occurs, while the abnormal displacement detection system of the disclosed embodiment can detect hidden problems such as long-term slight tilt or gradual settlement of towers in advance according to the change in the probability of abnormal displacement of towers, thereby preventing problems before they occur, which is of great positive significance for effectively preventing power failures caused by tower collapse in severe weather such as rain, snow and frost.

[0151] According to an embodiment of the present disclosure, the method also includes: constructing a sample distance matrix based on normal position information of each of the multiple towers when it is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement; training the artificial intelligence model based on the sample distance matrix and the corresponding sample label vector.

[0152] According to an embodiment of the present disclosure, the normal position information includes: position information manually set according to the reference position information of the pole tower and the preset normal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to be in the normal position; the abnormal position information includes: position information manually set according to the reference position information of the pole tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to have abnormal displacement.

[0153] According to an embodiment of the present disclosure, constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes:

[0154] For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information;

[0155] For each of the plurality of towers, a plurality of abnormal position information UP when the tower has abnormal displacement is obtained. ns , 1≤s≤M2, M2 is the number of abnormal location information;

[0156] Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UP ns one;

[0157] For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector corresponding to the sample matrix OS=[OS 1 ,OS 2 ,…OS n ,…,OS N], where when the position information of the nth tower is normal, OS n = 0, when the position information of the nth tower is abnormal, OS n =1.

[0158] According to an embodiment of the present disclosure, the loss function J used for artificial intelligence model training is:

[0159]

[0160] Where N is the number of towers, OS n For the tower T n The sample label of O n For the tower T n During the training process, the weights of the attention layer and the parameters of the deep neural network are optimized according to the loss function J.

[0161] According to an embodiment of the present disclosure, the method further includes: when the number or installation positions of the plurality of pole towers in the designated area changes, retraining the artificial intelligence model according to new position information of the plurality of pole towers.

[0162] This disclosure achieves adaptive detection capability of abnormal tower position by introducing distance matrix, combining deep learning algorithm and attention mechanism. Different from the traditional abnormality detection method based on fixed threshold, the deep learning algorithm combined with attention mechanism can adaptively adjust the weight according to the influence of the distance between different towers on the detection results, so as to automatically adapt to the environmental characteristics of different regions. Compared with the traditional detection method, it greatly reduces the false alarm and missed alarm. This adaptive capability can provide more accurate detection results in the complex environment where there are frequent occurrences in my country.

[0163] Figure 4 The structural block diagram of the regional tower position abnormality intelligent detection device according to the embodiment of the present disclosure is shown. The device can be implemented as part or all of the electronic device through software, hardware or a combination of both.

[0164] like Figure 4 As shown, in an embodiment of the present disclosure, a regional tower position abnormality intelligent detection device is provided, which is used to detect abnormalities in the positions of multiple towers in a specified area, and the device includes:

[0165] An acquisition module is configured to acquire the location information of the pole tower through a positioning terminal arranged at a designated position of the pole tower, wherein the location information includes the altitude and longitude and latitude coordinates of the positioning terminal;

[0166] A first calculation module is configured to calculate the distance between two of the plurality of towers according to the position information of the plurality of towers;

[0167] A generating module, configured to generate a distance matrix based on the distances between the plurality of towers;

[0168] The second calculation module is configured to calculate the abnormal displacement probability of each of the plurality of towers using a trained artificial intelligence model according to the distance matrix, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the plurality of towers, the artificial intelligence model comprising an attention layer and a deep neural network, and the use of the trained artificial intelligence model to calculate the abnormal displacement probability of each of the plurality of towers comprises:

[0169] Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix;

[0170] The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

[0171] According to an embodiment of the present disclosure, the distance matrix is:

[0172]

[0173] Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

[0174] According to an embodiment of the present disclosure, inputting the distance matrix into the attention layer to obtain an attention weighted distance matrix includes:

[0175] Calculate Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer;

[0176] Calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product;

[0177] By combining h with Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

[0178] According to an embodiment of the present disclosure, inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes:

[0179] The attention weighted distance matrix is ​​input into the deep neural network to obtain an N-dimensional vector O=[O 1 ,O 2 ,…O n ,…,O N ], where O 1 ,O 2 ,O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

[0180] According to an embodiment of the present disclosure, the deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

[0181] According to the embodiments of the present disclosure, <O n ≤1, when 0 <O n <0.5, the tower T n There is no abnormal displacement, when 0.5≤O n When ≤1, it is considered that the tower T n There is abnormal displacement.

[0182] According to an embodiment of the present disclosure, the loss function J of the artificial intelligence model is: The loss function is used to train the artificial intelligence model to optimize the weights of the attention layer and the parameters of the deep neural network.

[0183] According to an embodiment of the present disclosure, the device further includes:

[0184] A retraining module, configured to retrain the artificial intelligence model according to the location information of the plurality of towers when the number or installation locations of the plurality of towers in the designated area changes;

[0185] An alarm module is configured to output an alarm message when the probability of abnormal displacement of any tower continues to increase;

[0186] The statistical module is configured to count the probability change speed and amplitude of abnormal displacement of the pole tower, establish the corresponding relationship between the probability change speed and amplitude of abnormal displacement of the pole tower and the own characteristics and environmental characteristics of the pole tower, and use it as reference data for installation and maintenance operations of the pole tower.

[0187] According to an embodiment of the present disclosure, the device further includes:

[0188] A construction module is configured to construct a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and to construct a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement;

[0189] The training module is configured to train the artificial intelligence model based on the sample distance matrix and the corresponding sample label vector.

[0190] According to an embodiment of the present disclosure, the normal position information includes: position information manually set according to the reference position information of the pole tower and the preset normal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to be in the normal position; the abnormal position information includes: position information manually set according to the reference position information of the pole tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the pole tower is actually found to have abnormal displacement.

[0191] According to an embodiment of the present disclosure, constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes:

[0192] For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower;

[0193] For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower;

[0194] Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NPnj and UP ns one;

[0195] For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector corresponding to the sample matrix OS=[OS 1 ,OS 2 ,…OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

[0196] Figure 5 FIG. 2 shows a structural block diagram of a monitoring device according to an embodiment of the present disclosure. Figure 5 As shown, the monitoring device includes a memory and a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions are executed by the processor to implement the regional tower position abnormality intelligent detection method as described in any one of the above method embodiments.

[0197] The present disclosure also provides a computer-readable storage medium, which may be a computer-readable storage medium included in the electronic device or computer system in the above embodiment; or a computer-readable storage medium that exists independently and is not assembled into a device. The computer-readable storage medium stores one or more programs, and the programs are used by one or more processors to execute the regional tower position abnormality intelligent detection method described in the present disclosure.

[0198] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the present disclosure is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the inventive concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other.

Claims

1. A method for intelligently detecting abnormality of regional tower positions, characterized in that: For detecting abnormality at positions of a plurality of towers in a specified area, the method comprises: Acquiring location information of the pole tower by means of a positioning terminal arranged at a designated position of the pole tower, wherein the location information includes the altitude and longitude and latitude coordinates of the positioning terminal; Calculate the distance between each of the plurality of towers according to the position information of the plurality of towers; Generate a distance matrix based on the distances between the plurality of towers; According to the distance matrix, the abnormal displacement probability of each of the multiple towers is calculated using a trained artificial intelligence model, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the multiple towers, and the artificial intelligence model includes an attention layer and a deep neural network. The abnormal displacement probability of each of the multiple towers is calculated using the trained artificial intelligence model, including: Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix; The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

2. The method according to claim 1, characterized in that The distance matrix is: Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

3. The method according to claim 1, characterized in that The step of inputting the distance matrix into the attention layer to obtain an attention weighted distance matrix comprises: Calculate Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer; Calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product; By combining h with Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

4. The method according to claim 1, characterized in that: The step of inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes: The attention weighted distance matrix is ​​input into the deep neural network to obtain an N-dimensional vector O=[O1, O2, ... n ,…,O N ], where O1, O2, O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

5. The method according to claim 4, characterized in that The deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

6. The method according to claim 4, characterized in that 0 <O n ≤1, when 0 <O n <0.5, the tower T n There is no abnormal displacement, when 0.5≤O n When ≤1, it is considered that the tower T n There is abnormal displacement.

7. The method according to claim 6, characterized in that The loss function J of the artificial intelligence model is: Wherein, N is the number of the plurality of towers, The loss function is used to train the artificial intelligence model to optimize the weights of the attention layer and the parameters of the deep neural network.

8. The method according to claim 1, characterized in that Also includes: When the number or installation positions of the plurality of poles in the designated area changes, the artificial intelligence model is retrained according to new position information of the plurality of poles.

9. The method according to claim 1, characterized in that: The method further comprises: When the probability of abnormal displacement of any tower continues to increase, an alarm message is output.

10. The method according to claim 1, characterized in that The method further comprises: Counting the probability change speed and amplitude of abnormal displacement of the tower; A corresponding relationship between the probability change speed and amplitude of abnormal displacement of the pole tower and the characteristics of the pole tower itself and the environmental characteristics is established as reference data for installation and maintenance operations of the pole tower.

11. The method according to claim 1, characterized in that: The method further comprises: Constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement; The artificial intelligence model is trained based on the sample distance matrix and the corresponding sample label vector.

12. The method according to claim 11, characterized in that: The normal position information includes: position information manually set according to the reference position information of the tower and the preset normal displacement condition, and / or position information obtained by the corresponding positioning terminal when the tower is actually found to be in the normal position; The abnormal position information includes: position information manually set according to the reference position information of the tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the abnormal displacement of the tower is actually found.

13. The method according to claim 11, characterized in that The step of constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes: For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower; For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower; Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UP ns one; For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector OS=[OS1, OS2, …OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

14. An intelligent detection device for abnormal regional tower position, characterized in that: Used to detect abnormality at the positions of multiple towers in a specified area, the device comprises: An acquisition module is configured to acquire the location information of the pole tower through a positioning terminal arranged at a designated position of the pole tower, wherein the location information includes the altitude and longitude and latitude coordinates of the positioning terminal; A first calculation module is configured to calculate the distance between two of the plurality of towers according to the position information of the plurality of towers; A generating module, configured to generate a distance matrix based on the distances between the plurality of towers; The second calculation module is configured to calculate the abnormal displacement probability of each of the plurality of towers using a trained artificial intelligence model according to the distance matrix, wherein the artificial intelligence model is obtained by training a deep neural network model based on an attention mechanism using sample distance matrices and sample label vectors of the plurality of towers, the artificial intelligence model comprising an attention layer and a deep neural network, and the use of the trained artificial intelligence model to calculate the abnormal displacement probability of each of the plurality of towers comprises: Inputting the distance matrix into the attention layer to obtain an attention-weighted distance matrix, wherein the attention-weighted distance matrix is ​​obtained by multiplying the distance matrix by normalized weights, wherein the normalized weights are obtained based on the distance matrix; The attention weighted distance matrix is ​​input into the deep neural network to obtain the abnormal displacement probability of each tower.

15. The device according to claim 14, characterized in that The distance matrix is: Wherein, h is the distance matrix, N is the number of the plurality of towers, and h 11 is zero, h N1 h is the distance between the positioning terminal of the Nth tower and the positioning terminal of the first tower, 1N h is the distance between the positioning terminal of the first tower and the positioning terminal of the Nth tower, NN is zero.

16. The device according to claim 14, characterized in that The step of inputting the distance matrix into the attention layer to obtain an attention weighted distance matrix comprises: Calculate Q = hW Q , K = hW K , W Q and W K is the weight of the attention layer; Calculate the normalized weight Softmax (QK T ), where QK T To combine Q and K T Do the dot product; By combining h with Softmax(QK T ) performs dot multiplication to calculate the attention weighted distance matrix.

17. The device according to claim 14, characterized in that The step of inputting the attention weighted distance matrix into the deep neural network to obtain the abnormal displacement probability of each tower includes: The attention weighted distance matrix is ​​input into the deep neural network to obtain an N-dimensional vector O=[O1, O2, ... n ,…,O N ], where O1, O2, O n ,O N They represent the probability of abnormal displacement of the 1st, 2nd, nth, and Nth towers respectively.

18. The device according to claim 17, characterized in that The deep neural network includes a three-layer convolutional neural network, the first layer includes 80 neurons, the second layer includes 120 neurons, and the third layer includes 100 neurons.

19. The device according to claim 17, characterized in that 0 <O n ≤1, when 0 <O n <0.5, the tower T n There is no abnormal displacement, when 0.5≤O n When ≤1, it is considered that the tower T n There is abnormal displacement.

20. The device according to claim 19, characterized in that: The loss function J of the artificial intelligence model is: Wherein, N is the number of the multiple towers, and the loss function is used to train the artificial intelligence model to optimize the weight of the attention layer and the parameters of the deep neural network.

21. The device according to claim 14, characterized in that Also includes: A retraining module, configured to retrain the artificial intelligence model according to new position information of the plurality of towers when the number or installation positions of the plurality of towers in the designated area changes; An alarm module is configured to output an alarm message when the probability of abnormal displacement of any tower continues to increase; The statistical module is configured to count the probability change speed and amplitude of abnormal displacement of the pole tower, establish the corresponding relationship between the probability change speed and amplitude of abnormal displacement of the pole tower and the own characteristics and environmental characteristics of the pole tower, and use it as reference data for installation and maintenance operations of the pole tower.

22. The device according to claim 14, characterized in that Also includes: A construction module is configured to construct a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and to construct a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement; The training module is configured to train the artificial intelligence model based on the sample distance matrix and the corresponding sample label vector.

23. The device according to claim 22, characterized in that: The normal position information includes: position information manually set according to the reference position information of the tower and the preset normal displacement condition, and / or position information obtained by the corresponding positioning terminal when the tower is actually found to be in the normal position; The abnormal position information includes: position information manually set according to the reference position information of the tower and the preset abnormal displacement condition, and / or position information acquired by the corresponding positioning terminal when the abnormal displacement of the tower is actually found.

24. The device according to claim 22, characterized in that The step of constructing a sample distance matrix based on normal position information when each of the multiple towers is in a normal position and abnormal position information when there is abnormal displacement, and constructing a sample label vector corresponding to the sample distance matrix based on whether each of the multiple towers has abnormal displacement, includes: For each of the plurality of towers T n , obtain multiple normal position information NP when the tower is in multiple normal positions nj , 1≤j≤M1, M1 is the number of normal position information of the tower; For each of the plurality of towers T n , obtain multiple abnormal position information UP when the tower has abnormal displacement ns , 1≤s≤M2, M2 is the number of abnormal position information of the tower; Based on the plurality of normal position information and the plurality of abnormal position information of the plurality of towers, a plurality of different sample states of the plurality of towers are constructed. In each sample state, each tower T of the plurality of towers n The location information is NP nj and UP ns one; For each sample state, according to each tower T in the sample state n The position information of the sample state is determined by determining the sample distance matrix corresponding to the sample state and the sample label vector OS=[OS1, OS2, …OS n ,…,OS N ], where when the tower T n When the location information is normal, OS n =0, when the tower T n If the location information is abnormal, OS n =1.

25. A monitoring device, characterized in that: The method comprises a memory and a processor; wherein the memory is used to store one or more computer instructions, and the one or more computer instructions are executed by the processor to implement the method according to any one of claims 1 to 13.

26. An intelligent detection system for regional tower position anomalies, characterized in that: For detecting abnormality of the positions of a plurality of pole towers in a specified area, the system comprises a plurality of positioning terminals and a monitoring device according to claim 25, wherein: The multiple positioning terminals are respectively arranged at designated positions of the multiple towers; The positioning terminal obtains the position information of the corresponding tower and sends the position information to the monitoring device, wherein the position information includes the altitude and longitude and latitude coordinates of the positioning terminal.

27. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the method according to any one of claims 1 to 13 is implemented.

Images