Intelligent line intelligent inspection management system and method based on deep learning

Abstract

The invention provides an intelligent line intelligent inspection management system and method based on deep learning. The system comprises a data acquisition device arranged in a to-be-inspected area and used for acquiring monitoring data and environmental data on different power lines; the server is used for receiving the monitoring data and the environment data on different power lines reported by the data acquisition device in real time; according to the monitoring data and the environment data of each power line in the preset time period, performing routing inspection on the power line by adopting a power routing inspection model to obtain a routing inspection result, and pushing the routing inspection result of each power line to related terminal equipment; wherein the inspection result comprises whether a fault occurs or not and a fault type when the fault occurs; the power inspection model is a model which is obtained by performing optimization training on a long short-term memory network by adopting an improved population optimization algorithm and can predict and identify power line faults. The system and the method are used for achieving the effect of reducing the dependence of the power grid safety inspection process on manpower.

Description

Technical Field

[0001] The present application relates to the field of robot intelligent inspection technology, and in particular to a smart line intelligent inspection management system and method based on deep learning. Background Art

[0002] With the acceleration of urbanization and the continuous growth of electricity demand, the scale of urban power grids is expanding and the structure is becoming more complex. Any failure or accident in the power grid will cause serious consequences. Therefore, safety inspection is particularly important in the application of urban power grids.

[0003] Traditional power grid safety inspections are mainly manual, and inspectors need to conduct on-site inspections along power lines, which is not only inefficient but also difficult to fully cover all areas, especially remote areas and areas with complex terrain. In addition, manual inspections also have safety hazards, and inspectors may face risks such as electric shock and falling from heights. Summary of the invention

[0004] The present application provides a deep learning-based smart line intelligent inspection management system and method to reduce the reliance on manual labor in the safety inspection process.

[0005] In a first aspect, the present application provides a smart line intelligent inspection management system based on deep learning, comprising: a data acquisition device arranged in an area to be inspected, the data acquisition device being used to collect monitoring data and environmental data on different power lines;

[0006] A server, the server being connected to the data acquisition device and being used to receive monitoring data and environmental data on different power lines reported in real time by the data acquisition device;

[0007] The server is also used to inspect the power lines using a power inspection model according to the monitoring data and environmental data of each power line within a preset time period to obtain inspection results, wherein the inspection results include whether a fault will occur and the type of fault when a fault occurs; wherein the power inspection model is a model that can predict and identify power line faults obtained by optimizing and training a long short-term memory network using an improved population optimization algorithm;

[0008] The server is also used to push the inspection result of each power line to the relevant terminal equipment.

[0009] In a possible implementation, the data acquisition device includes a plurality of monitoring devices, a temperature sensor, a humidity sensor and an image acquisition device for detecting current, voltage and load information set for each power line;

[0010] Wherein, the plurality of monitoring devices are used to detect the current, voltage and load information of the power line, and the monitoring data of each power line includes the current, voltage and load information;

[0011] The temperature sensor is used to detect the temperature of the environment data where the power line is located, and the humidity sensor is used to detect the humidity of the environment where the power line is located;

[0012] The image acquisition device is used to acquire image data of the power line.

[0013] In a possible implementation manner, the server is specifically used for:

[0014] For each power line, according to the current, voltage, load information, temperature, humidity and image data reported by the data acquisition device, a characteristic data matrix of the power line is obtained, wherein the characteristic data matrix includes the current, voltage, load information, temperature, humidity and image recognition results at multiple time points, and the image recognition results are obtained by performing abnormal recognition on the image data;

[0015] Inputting the characteristic data matrix of the power line into the power inspection model, performing forward feature extraction on the features of each dimension in chronological order through the power inspection model to obtain a forward feature sequence, performing reverse feature extraction processing on the features of each dimension in chronological order to obtain a reverse feature sequence, and fusing the obtained forward feature sequence and reverse feature sequence of the features of each dimension to obtain a fused feature sequence;

[0016] According to the feature weight of each dimension, the fused feature sequences of multiple dimensions are fused again to obtain the time series features;

[0017] According to the time series characteristics, the power grid is analyzed and processed to obtain real-time inspection results of the power grid.

[0018] In a possible implementation manner, the server is further used for:

[0019] Based on the historical data of multiple power lines acquired in advance, a sample set is acquired, wherein the sample set includes multiple samples, each sample includes the current, voltage, load information, temperature, humidity, image recognition result, and fault information of a power line at multiple time points within a set time period, wherein the fault information includes whether a fault occurs within a preset time period after the set time period, and the type of fault when a fault occurs; the length of the set time period is consistent with the length of the preset time period;

[0020] Constructing multiple initial models with exactly the same structure, each initial model includes a bidirectional LSTM network, a self-attention layer, a fully connected layer, and a classification output layer, wherein the bidirectional LSTM network includes a two-layer LSTM network for forward and reverse feature extraction;

[0021] Based on the sample set and the multiple initial models, model training is performed, and during the training process, an improved sand cat population optimization algorithm is used to optimize multiple hyperparameters of the model to obtain the power inspection model.

[0022] In a possible implementation manner, the server is specifically used for:

[0023] Step a, initializing the maximum number of iterations of the improved sand cat swarm optimization algorithm and a plurality of sand cat individuals, the position of each sand cat individual is used to represent a set of parameter configuration schemes, each set of parameter configuration schemes includes a learning rate, the number of hidden layer neurons and a regularization coefficient;

[0024] Step b: for each initial model, a set of parameter configuration schemes is selected to perform parameter configuration to obtain multiple initial inspection models;

[0025] Step c, selecting a plurality of samples from the sample set and inputting them into different initial inspection models for line fault inspection processing, obtaining the inspection results output by each initial inspection model, and determining the fitness value of each group of parameter configuration schemes based on the output results of each initial inspection model and the fault information in the input samples by using the accuracy function of the sand dune cat swarm optimization algorithm;

[0026] Step d: compare the highest fitness value of this round with the highest fitness value of the previous round, and select a set of parameter configuration schemes with the highest fitness value as the target parameter configuration scheme of this round;

[0027] Step e: if the preset iteration termination condition is not met, then according to the target parameter configuration scheme, the R value of each sand dune cat individual in the sand dune cat group optimization algorithm is calculated, and the positions of the multiple sand dune cat individuals are updated using the position update formula to obtain new multiple groups of parameter configuration schemes, and steps b to e are repeated until the iteration termination condition is met, and the target parameter configuration scheme obtained by the last iteration is used as the final parameter configuration scheme;

[0028] Step f: configuring the final parameter configuration scheme in the initial model to obtain the power inspection model.

[0029] In a possible implementation manner, the server is further specifically configured to:

[0030] Based on the position initialization formula, the initial position of each individual sand dune cat is obtained;

[0031] Wherein, the position initialization formula is:

[0032] x ij =lb j +rand(0,1)·(ub j -lb j )

[0033] x ij represents the value of the j-dimensional variable of the i-th sand dune cat individual; lb j represents the upper bound of the j-th dimension variable; ub j represents the lower bound of the j-th dimension variable; j=1 represents the learning rate, j=2 represents the number of neurons in the hidden layer, and j=3 represents the regularization coefficient.

[0034] In a possible implementation manner, the server is further specifically configured to:

[0035] Determine the best position for this iteration according to the target parameter configuration scheme;

[0036] According to the R value of the sand dune cat group optimization algorithm, the behavior of each sand dune cat individual is determined, and the behavior of the sand dune cat individual includes attack behavior and search behavior, wherein the R value is calculated using the following formula:

[0037]

[0038] R=2·r G rand(0,1)-r G

[0039] In the above formula, S M is the sensitivity of the preset sand dune cat individual; T represents the maximum number of iterations; t represents the current number of iterations; r G represents the sensitivity range of individual sand dune cats in the current iteration;

[0040] According to the behavior of each individual sand dune cat, the position is updated using a position update formula corresponding to the behavior to obtain the updated position of each individual sand dune cat.

[0041] In a possible implementation, the system further includes: a plurality of terminal devices and a plurality of inspection robots for managing power lines within different areas;

[0042] The server is also specifically used to generate a regional power inspection report based on the inspection results of each power line, each regional power inspection report includes the inspection results of all power lines within the corresponding area, and the alarm corresponding to the power line that will fail;

[0043] The server is also used to push regional power inspection reports in different areas to corresponding terminal devices for display;

[0044] The server is also used to send a fault warning instruction to the inspection robot within a preset range of the power line where the fault will occur;

[0045] The inspection robot that receives the fault warning instruction will cut off the power supply of the faulty power line according to the fault warning instruction.

[0046] In a possible implementation manner, the server is further used for:

[0047] The power inspection model is optimized based on the inspection result of each power line and the inspection feedback data returned by the terminal equipment corresponding to each power line.

[0048] In a second aspect, the present application provides a smart line intelligent inspection management method based on deep learning, which is applied to a server in a smart line intelligent inspection management system based on deep learning in the first aspect and / or various possible implementations of the first aspect, and the method includes:

[0049] Receive monitoring data and environmental data on different power lines reported in real time by data acquisition devices;

[0050] For each power line, according to the monitoring data and environmental data of the power line within a preset time period, a power inspection model is used to inspect the power line to obtain an inspection result, wherein the inspection result includes whether a fault will occur and the type of fault when a fault occurs; wherein the power inspection model is a model capable of predicting and identifying power line faults obtained by optimizing and training a long short-term memory network using an improved population optimization algorithm;

[0051] Push the inspection results of each circuit line to the relevant terminal equipment.

[0052] The present application provides a smart line intelligent inspection management system and method based on deep learning, the system includes: a data acquisition device set in the area to be inspected, used to collect monitoring data and environmental data on different power lines; a server connected to the data acquisition device, used to receive the monitoring data and environmental data on different power lines reported in real time by the data acquisition device; according to the monitoring data and environmental data of each power line in a preset time period, the power line is inspected using a power inspection model to obtain inspection results, and the inspection results of each power line are pushed to the relevant terminal equipment; wherein the inspection results include whether a fault will occur and the type of fault when a fault occurs; the power inspection model is a model that can predict and identify power line faults obtained by optimizing and training the long short-term memory network using an improved population optimization algorithm. The system is used to reduce the dependence of the safety inspection process on manual labor. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0054] Figure 1 A schematic diagram of the structure of a smart line intelligent inspection management system based on deep learning provided in Example 1 of the present application;

[0055] Figure 2 A flowchart of a smart line intelligent inspection management system and method based on deep learning provided in Example 1 of the present application;

[0056] Figure 3 A flowchart of a smart line intelligent inspection management system and method based on deep learning provided in Example 2 of the present application;

[0057] Figure 4 A structural diagram of a smart line intelligent inspection management system based on deep learning is provided for the third embodiment of the present application;

[0058] Figure 5 A flow chart of a smart line intelligent inspection management system and method based on deep learning is provided for the third embodiment of the present application.

[0059] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0060] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0061] In order to clearly understand the technical solution of the present application, the solution of the prior art is introduced in detail below.

[0062] Grid safety inspection is a key link to ensure the safe and stable operation of the power system. Through grid safety inspection, potential faults and hidden dangers can be discovered and handled in a timely manner, large-scale power outages can be prevented, and the continuity and reliability of power supply can be guaranteed. Usually, inspection personnel use on-site operation, manual transcription, round-trip operation and other forms of daily operation and maintenance of line equipment.

[0063] However, there are many drawbacks to the inspection method that relies on manual labor. Inspectors need to conduct on-site inspections along the power lines, which is not only inefficient, but also difficult to fully cover all areas, especially remote areas and areas with complex terrain. Safety inspections of the power grid are also required at night, which will violate the physiological cycle of inspectors and cause harm to them. In addition, the harsh inspection environment will not only cause physical harm to inspectors (such as coal transportation corridors, water treatment rooms, gas insulated substation (GIS) rooms, etc.), but also expose inspectors to risks such as electric shock and falling from heights (such as harsh terrain such as cliffs and mountains; extreme weather such as rain, snow, and freezing).

[0064] In addition to the above-mentioned manual inspection methods for power grid safety, there are also inspection methods such as manned helicopter inspection, robot inspection, and drone inspection. Among them, manned helicopter inspection has not been widely used due to high investment costs, large manpower, complex management, and complicated technical reserves; while robot inspection and drone inspection, due to the relatively independent data collected, lack of data exchange and sharing mechanism, there is a problem of data islands, so that it is impossible to centrally analyze and manage the data of multiple inspection areas, and it is impossible to coordinate work and make overall decisions for multiple inspection areas.

[0065] Therefore, the prior art lacks a power grid security inspection method that can replace manual inspection and break the "data island" phenomenon.

[0066] Based on the description of the above background technology, the inventors discovered in the process of studying inspection methods that based on deep learning algorithms, they designed a management system for intelligent line inspection that can obtain inspection data in real time and centrally process and analyze the inspection data. This can achieve unattended, controllable and intelligent inspection in environments with complex line environments and harsh operating environments.

[0067] Based on the technical concept of the inventor, the present application provides a smart line intelligent inspection system and method based on deep learning. The system includes a data acquisition device and a server arranged in the area to be inspected, wherein the data acquisition device is used to collect monitoring data and environmental data on different power lines; the server is connected to the data acquisition device and is used to receive the monitoring data and environmental data on different power lines reported in real time by the data acquisition device; according to the monitoring data and environmental data of each power line in a preset time period, the power line is inspected using a power inspection model to obtain inspection results, and the inspection results of each power line are pushed to the relevant terminal equipment; wherein the inspection results include whether a fault will occur and the type of fault when a fault occurs; the power inspection model is a model that can predict and identify power line faults obtained by optimizing and training the long short-term memory network using an improved population optimization algorithm. The inspection method based on this system can not only effectively avoid the disadvantages of traditional manual inspections, but also break the barriers of "data islands" existing in robot inspections and drone inspections.

[0068] The deep learning-based intelligent line intelligent inspection management system provided by this application can perform inspection operations on various power grid environments, such as power grid environments in large urban power grid environments with congested construction, complex line distribution, and wide range of fault impact. This application does not limit the operating environment of the management system.

[0069] Figure 1 This is a schematic diagram of the structure of the smart line intelligent inspection management system based on deep learning provided in Example 1 of the present application, such as Figure 1 As shown, the deep learning-based smart line intelligent inspection and management system 10 provided in the present application includes a data acquisition device 101 and a server 102, wherein the acquisition device 101 and the server 102 are connected, and the acquisition device 101 is used to collect monitoring data and environmental data on different power lines; the server 102 is used to receive the monitoring data and environmental data on different power lines reported in real time by the data acquisition device 101, and according to the monitoring data and environmental data of each power line within a preset time period, the power line is inspected by using a power inspection model to obtain inspection results, and finally the inspection results of each power line are pushed to the relevant terminal device 103.

[0070] The data acquisition device 101 may be various monitoring devices, sensor devices, image acquisition devices, etc., such as high-definition cameras, infrared sensors, temperature sensors, gas sensors, wind speed sensors, smart surveillance cameras, and IoT smart sensors, etc., which are not limited in this application;

[0071] The data acquisition device 101 and the server 102 use wireless communication to transmit data, such as the fifth generation mobile communication technology (5G), Zigbee technology, wireless network (Wireless Fidelity, Wifi), Bluetooth, etc.

[0072] The terminal device 103 may be a portable mobile communication terminal (such as a smart phone or a tablet computer), a computer device, a cockpit management terminal, a large indicator screen, etc.

[0073] Furthermore, managers can also use applications, applets, etc. on terminal devices to achieve real-time, remote, and intelligent inspection and management of multiple power lines.

[0074] Figure 2 A flowchart of a smart line intelligent inspection management system and method based on deep learning provided in Example 1 of the present application is shown in FIG. Figure 2 As shown, the smart line intelligent inspection management system and method based on deep learning in this embodiment include:

[0075] S201, the data acquisition device collects monitoring data and environmental data on different power lines;

[0076] In this solution, in order to realize intelligent inspection of power lines, the smart line intelligent inspection management system based on deep learning needs to use the data acquisition device in the system to collect relevant data on different power lines in real time, and send the acquired relevant data to the server in the system based on wireless communication technology; the server then analyzes and processes the received relevant data in real time based on the power inspection model to obtain the inspection results of the power lines.

[0077] In this step, in order to avoid manual inspection, multiple data acquisition devices need to be configured in the system to collect monitoring data and environmental data on different power lines in real time. Monitoring data refers to electrical parameter information related to power lines, such as line voltage, current, power, etc., and line status, such as fire, snow cover, disconnection, etc.; environmental data refers to information related to the environment where the power lines are located, such as wind speed, temperature, humidity, pressure, etc.

[0078] In a specific implementation, the data acquisition device may be an electrical parameter monitoring device, an environmental monitoring sensor, an image acquisition device, etc., which are installed at a substation or a transmission line in a power grid. The electrical parameter monitoring device can collect monitoring data of the power line; the environmental monitoring sensor. For example, the electrical parameter monitoring device may be a voltage transformer, a current transformer, a power sensor, a partial discharge sensor, etc.; the environmental monitoring sensor may be a meteorological sensor, a temperature sensor, a humidity sensor, an infrared sensor, a pressure sensor, etc.

[0079] In a specific implementation, the data acquisition device may be an image acquisition device, such as a high-definition camera, an intelligent monitoring camera, etc.; by using the image acquisition device to capture image data of key parts of the power line (such as line channels, tower bases, conductors, insulators, etc.), the line status information can be obtained based on the collected image data, such as line fire, line covered by snow, line disconnection, etc.

[0080] In a specific implementation, the data acquisition device may be an IoT smart sensor. The IoT smart sensor is a new type of sensor that integrates electrical parameter monitoring equipment with environmental monitoring sensors and other acquisition equipment. It can collect monitoring data such as electrical parameters and operating status of power equipment, as well as environmental data.

[0081] In a specific implementation, the data acquisition device includes a plurality of monitoring devices, a temperature sensor, a humidity sensor and an image acquisition device, which are arranged for each power line and are used to detect current, voltage and load information.

[0082] In this implementation, multiple monitoring devices are used to detect the current, voltage and load information of the power line. Accordingly, the monitoring data of each power line collected by the multiple monitoring devices includes current, voltage and load information; the temperature sensor is used to detect the temperature of the environment in which the power line is located, and the humidity sensor is used to detect the humidity of the environment in which the power line is located; the image acquisition device is used to collect image data of the power line.

[0083] Specifically, the multiple monitoring devices may be voltage transformers, current transformers, power sensors, and the like.

[0084] Optionally, one image acquisition device may monitor one power line, or one image acquisition device may monitor multiple power lines at the same time.

[0085] S202: The data collection device sends the collected monitoring data and environmental data on different power lines to the server.

[0086] S202. The server receives monitoring data and environmental data on different power lines reported in real time by a data acquisition device.

[0087] In the above steps, the data acquisition device needs to send the real-time collected monitoring data and environmental data to the server by wireless communication, so that the server can process and analyze the data. Correspondingly, the server needs to receive the monitoring data and environmental data on different power lines reported in real time by the data acquisition device.

[0088] In a specific implementation method, wireless communication methods such as 5G technology, Zigbee technology, Wifi and Bluetooth can be used.

[0089] Optionally, an intelligent gateway can be used to authenticate the data acquisition devices set in the system, and to encrypt the data reported by the data acquisition devices, so as to ensure the stability, reliability and security of the data received by the server. Among them, the intelligent gateway can be an Ankerui series intelligent gateway. Specifically, an authentication mechanism can be used to judge the authorization status of each data acquisition set in the system, so that only authorized data acquisition devices can report data to the server, which not only ensures the reliability of the data and improves the accuracy of the server data analysis and processing, but also effectively avoids illegal access and attacks on the system by external malicious devices, and reduces the security risks of the system being attacked by hackers, virus intrusions, etc.; the data can be encrypted by an encryption algorithm so that the data is transmitted in ciphertext, which effectively prevents the data from being leaked and tampered with, resulting in inaccurate analysis when the subsequent server analyzes the data of each power line in the power grid.

[0090] S203: For each power line, the server inspects the power line using a power inspection model according to monitoring data and environmental data of the power line within a preset time period to obtain an inspection result.

[0091] In this step, after receiving the monitoring data and environmental data of the power line within the preset time period, the server predicts and judges the monitoring data and environmental data of each power line based on the pre-built and trained power inspection model to obtain the inspection results. For a power line, after the monitoring data and environmental data of the power line within the preset time period are input into the power inspection model, it will determine whether the power line is about to have a line fault, and if the power line is about to have a line fault, the fault type of the line will be judged to obtain the inspection results for the supervisor to view.

[0092] Specifically, the inspection results include whether a fault will occur and the type of fault when it occurs; the power inspection model is a model obtained by optimizing and training the long short-term memory network using an improved population optimization algorithm. The model can predict and identify power line faults.

[0093] In a specific implementation, if the data acquisition device includes a plurality of monitoring devices, a temperature sensor, a humidity sensor and an image acquisition device for each power line for detecting current, voltage and load information, correspondingly, the monitoring data and environmental data of each power line include the current, voltage and load information of the power line, the humidity and temperature of the environment in which the power line is located and the image data of the power line; then, the server inspects the acquired data in the following manner, and the method includes:

[0094] S2031. For each power line, obtain a characteristic data matrix of the power line according to the current, voltage, load information, temperature, humidity and image data reported by the data acquisition device.

[0095] In this step, the characteristic data matrix includes current, voltage, load information, temperature, humidity and image recognition results at multiple time points, wherein the image recognition results are obtained by performing abnormality recognition on the image data.

[0096] The number of time points in each feature matrix is ​​determined by the preset data collection frequency and the duration of the preset time period. For example, the model is preset to process a multidimensional time series within 10 minutes at a time, and the sample collection frequency is 1 time / second, then the number of time points in each time feature matrix is ​​600.

[0097] In one possible implementation, in order to reduce the amount of data processed by the server and improve the server data processing speed, the server may no longer receive data reported by a data collection device whose collected data value remains constant as the power line runs.

[0098] In one possible implementation, the image recognition result includes the state type of the power line, such as a normal state, and abnormal states such as a fire state and a line disconnection state. The present application does not limit the feasibility of the abnormal state type that can be included in the image recognition result. The image recognition result can be obtained based on a pre-trained image recognition model.

[0099] Specifically, based on a pre-trained image recognition model, the image data at the current moment reported by the data acquisition device is recognized and processed to obtain the image features at the current moment extracted from the image data; the extracted image features at the current moment are compared with the pre-extracted image features of the power line when normal to obtain the difference features of the above two image features; based on the difference features of the above two images, the image data at the current moment is classified to obtain the image recognition results of the power line; wherein the image recognition results include the status information of the power line.

[0100] Furthermore, the image recognition model may be a convolutional neural network (CNN) model. The following steps may be used to train the CNN model:

[0101] Step 1: Collect image data of normal power lines and image data of various abnormal power lines, and label the state type of each image data, for example, the normal state is labeled as 0; the fire state is labeled as 1; the disconnected state is labeled as 2, etc., to obtain a training sample set for the image recognition model.

[0102] Step 2: Based on the constructed initial convolutional neural network model, during the model training process, the training sample set of the image recognition model is used to train the model to obtain the image recognition model. The initial convolutional neural network model includes a convolution layer, a pooling layer, and a fully connected layer.

[0103] In this implementation, the CNN model is used to perform image recognition processing on the real-time image data of each power line taken by multiple cameras to obtain image recognition results that can indicate the status of each power line, thereby ensuring the accuracy of image data recognition. Furthermore, the image recognition result is used as a dimensional feature input in the subsequent power inspection model, which can help the power inspection model to more quickly judge whether the power line is about to (or has) failed. At the same time, based on the power line recognition result, the accuracy of the fault type judgment can be further improved.

[0104] Optionally, the image recognition result may include the location information of at least one power line covered in each image data. Correspondingly, the inspection result may include the location information of the power line that is about to fail (or has failed), which is used to indicate the location of the power line to be repaired so that the corresponding power line can be quickly processed.

[0105] S2032. Input the characteristic data matrix of the power line into the power inspection model, perform forward feature extraction on the features of each dimension in chronological order through the power inspection model to obtain a forward feature sequence, perform reverse feature extraction on the features of each dimension in chronological order to obtain a reverse feature sequence, and fuse the forward feature sequence and reverse feature sequence of the features of each dimension to obtain a fused feature sequence.

[0106] In this step, the core of the electric power inspection model used is a bidirectional long short memory (Short Term Memory, LSTM) network. For the time series of the feature of one dimension within a time step, after inputting it into the electric power inspection model, firstly, according to the forward time order, the forward feature extraction of the time series is performed in the forward layer, and the forward feature vector of the dimension within the time step is obtained, and it is saved in the hidden layer of the forward layer. At the same time, according to the reverse time order, the reverse feature extraction of the time series is performed in the backward layer, and the reverse feature vector of the dimension within the given time step is obtained, and it is saved in the hidden layer of the backward layer. Based on this, the forward feature vector and the reverse feature vector of the time series of the feature of one dimension within a time step are obtained. Further, the forward feature vector and the reverse feature vector of the feature of each dimension within multiple time steps are obtained in this way to obtain the forward feature sequence and the reverse feature sequence of the feature of each dimension. Finally, the forward feature sequence and the reverse feature sequence of the feature of each dimension saved in the hidden layer are integrated to obtain the fusion feature sequence within multiple time steps of each dimension, and as the output of the bidirectional LSTM network. Among them, the fused feature sequence output by the bidirectional LSTM network contains fused feature vectors corresponding to the features of each dimension in multiple time steps.

[0107] It should be understood that compared with the LSTM network, the bidirectional LSTM network extracts features from both directions of the time series at the same time based on the LSTM network, and can extract the features of the time series more comprehensively, so as to accurately predict whether the power line will fail within a preset time in the future.

[0108] S2033. According to the feature weight of each dimension, the fused feature sequences of multiple dimensions are fused again to obtain time series features.

[0109] In this step, in order to obtain more accurate and reasonable prediction results in the power inspection model, in the output layer of the bidirectional LSTM network, according to the fused feature sequence output by the bidirectional LSTM network and the feature weight of each dimension, the fused feature sequence of each dimension is fused again to obtain the final time series features.

[0110] In a specific implementation, the output layer of the bidirectional LSTM network is a self-attention layer. For the fused feature sequence composed of the fused feature vectors corresponding to the features of each dimension within a time step, the self-attention layer compares each vector in each fused feature sequence input with the weight matrix W of the linear transformation learned by the layer in the model training phase. Q , W K , W VMultiply to obtain the query vector Q, matching vector K and target vector V of each vector in the sequence of fused features;

[0111] Using formula Get the self-attention score of each fused feature vector in the fused feature sequence. T Represents a 1×d vector consisting of all vectors in the fused feature sequence k dimensional matrix; d k Indicates the number of vectors in the fused feature sequence;

[0112] The Softmax function is used to normalize the self-attention score of each fused feature to obtain the self-attention weight of each fused feature vector in the fused feature sequence; based on the self-attention weight of each fused feature vector and the target vector, the final time series features are obtained.

[0113] In this implementation, since the contribution of features of different dimensions to the final result is often different, the self-attention layer dynamically assigns different weights to the features of each dimension based on the internal structure of the data of each dimension in the input data learned during the training process and the correlation between data of different dimensions, so as to distinguish the importance of the features of each dimension. At the same time, through the self-attention layer, more valuable information can be better screened out from the data, which can improve the efficiency of the power inspection model in analyzing and processing data.

[0114] S2034. Analyze and process the electric power grid according to the time series characteristics to obtain real-time inspection results of the electric power grid.

[0115] In this step, in the power inspection model, the time series features acquired in real time are input into the subsequent layer for feature classification, and the time series features of multiple dimensions of each power line in the power grid are classified and processed, so that the inspection results corresponding to the time series features of multiple dimensions of each power line can be obtained in real time.

[0116] In a possible implementation, in the power inspection model, the output layer of the attention layer is a fully connected layer, and the output layer of the fully connected layer is a classification output layer. For each power line, the fully connected layer and the classification output layer analyze and process the time series features output from the attention layer, and the classification result output by the classification output layer is the inspection result of the power line.

[0117] S204: The server pushes the inspection result of each circuit line to the relevant terminal device.

[0118] In this step, in order to facilitate supervisors to respond to inspection results in a timely manner and perform timely maintenance and repair of faulty lines, the inspection results of each power line need to be pushed to the terminal device corresponding to each power line for supervisors to view.

[0119] Optionally, only the relevant information of the faulty power line may be pushed to the terminal device corresponding to the faulty line to reduce the transmission of unnecessary data and alleviate the processing burden of the terminal device.

[0120] In a specific implementation, the terminal device 20 may be a portable mobile communication terminal (such as a smart phone or a tablet computer), a computer device, a cockpit management terminal, a large indicator screen, etc.

[0121] Optionally, managers can also use applications, applets, etc. on terminal devices to achieve real-time, remote, and intelligent inspection and management of multiple power lines.

[0122] The smart line intelligent inspection management system and method based on deep learning provided in the embodiment of the present application collects and reports the monitoring data and environmental data on different power lines in real time through the server receiving data acquisition equipment, and inspects the power lines using the power inspection model for the monitoring data and environmental data of each power line within a preset time period, and obtains inspection results including whether each power line will fail and the type of fault when a failure occurs. Among them, the power inspection model is a model that can predict and identify power line faults obtained by optimizing and training the long short-term memory network using an improved population optimization algorithm; finally, the inspection results of each circuit line are pushed to the relevant terminal equipment to achieve the technical effect of intelligent safety inspection of the power grid and reduce the dependence on manual inspection, effectively avoiding many disadvantages brought by traditional manual inspection. At the same time, the method provided in this embodiment, through the setting of a smart line intelligent inspection management system based on deep learning, simultaneously obtains relevant data of power lines in different areas, and uniformly processes the data based on the power inspection model in the server, breaking the "data island" barrier that exists in robot inspection and drone inspection, and realizing centralized management of relevant data of power lines in multiple regions, which is helpful for the overall planning of the power grid.

[0123] Figure 3 A flowchart of a smart line intelligent inspection management system and method based on deep learning provided in Example 2 of the present application is shown in FIG. Figure 3 As shown, based on the above embodiments, the deep learning-based intelligent line intelligent inspection management system and method provided by the present application also includes a method for obtaining a power inspection model, which is applied to a server of the system and includes:

[0124] S301. Acquire a sample set based on pre-acquired historical data of multiple power lines.

[0125] In this step, the historical data of multiple power lines include monitoring data and environmental data on different power lines historically collected by the data collection device. The sample set obtained from the historical data includes multiple samples, each of which includes the current, voltage, load information, temperature, humidity, image recognition results, and fault information of a power line at multiple time points within a set time period, wherein the fault information includes whether a fault occurs within a preset time period after the set time period, and the type of fault in the event of a fault; wherein the length of the set time period is consistent with the length of the preset time period.

[0126] Specifically, in the sample set, the current, voltage, load information, temperature, humidity, and image recognition results of a power line of each sample at multiple time points within a set time period are used as the input of the model, and the fault information corresponding to the current, voltage, load information, temperature, humidity, and image recognition results at the aforementioned multiple time points is used as the output of model training.

[0127] The fault type in the fault information is represented in the form of a label code. For example, the label code of an overload fault is 1, the label code of a short circuit fault is 2, the label code of a ground fault is 3, the label code of a disconnection fault is 4, the label code of a high resistance fault is 5, and the label code of a lightning fault is 6.

[0128] In a specific implementation, the sample set is divided into a training set, a test set and a validation set for model training. Exemplarily, 80% of the sample set is used as a training set, 20% as a test set, and 20% of the training set as a validation set.

[0129] In a possible implementation, in order to ensure the reliability of the data in the sample set, the data in the sample set needs to be preprocessed. The preprocessing process includes: removing data with obvious errors; removing the mean and normalizing the variance of the data of each dimension of each power line. The specific formula is as follows:

[0130]

[0131] In the above formula, X represents the original sample data of a certain dimensional feature corresponding to a certain power line; μ represents the mean of all sample data of a certain dimensional feature corresponding to a certain power line; σ represents the standard deviation of all sample data of a certain dimensional feature corresponding to a certain power line; X * Represents the normalized sample data of a certain dimensional feature corresponding to a certain power line.

[0132] S302. Construct multiple initial models with exactly the same structure, each of which includes a bidirectional LSTM network, a self-attention layer, a fully connected layer, and a classification output layer. The bidirectional LSTM network includes a two-layer LSTM network for forward and reverse feature extraction.

[0133] In this step, multiple identical initial models need to be built at the same time, so that the optimal parameter configuration scheme can be obtained from multiple parameter configuration schemes in the subsequent hyperparameter optimization process. The constructed initial model includes a bidirectional LSTM network, a self-attention layer, a fully connected layer, and a classification output layer.

[0134] Optionally, the classification output layer is a Softmax classifier, which can perform classification processing tasks with multiple labels and multiple classifications. Exemplarily, there can be multiple fault types for a power line.

[0135] Specifically, the main structure of each LSTM network includes a gate control unit, which includes a forget gate, an input gate, and an output gate.

[0136] S303, based on the sample set and multiple initial models, model training is performed, and during the training process, multiple hyperparameters of the model are optimized using an improved sand cat population optimization algorithm to obtain a power inspection model.

[0137] In this step, for an initial model, a hyperparameter combination needs to be allocated for a period of time to obtain the initial inspection model, and during its training process, training is performed according to pre-set training parameters, wherein the hyperparameter combination is a combination of learning rate, number of hidden layer neurons, and regularization coefficient, and the pre-set training parameters include batch size and training iteration rounds. This application does not limit the setting of batch size and training iteration rounds. Exemplarily, the pre-set training parameters can be: batch size is set to 64; training iteration rounds are set to 100 rounds.

[0138] Specifically, based on the improved sand cat population optimization algorithm, the best hyperparameter combination is obtained during the model training process, and the training model corresponding to the hyperparameter combination is used as the final power inspection model.

[0139] In a specific implementation, the following method is used to optimize multiple hyperparameters of the model during model training to obtain a reasonable parameter configuration scheme for the learning rate, the number of hidden layer nodes, and the regularization coefficient. The method includes:

[0140] Step a, initializing the maximum number of iterations of the improved sand cat swarm optimization algorithm and a plurality of sand cat individuals, the position of each sand cat individual is used to represent a set of parameter configuration schemes, each set of parameter configuration schemes includes a learning rate, the number of hidden layer neurons and a regularization coefficient;

[0141] Step b: for each initial model, a set of parameter configuration schemes is selected to perform parameter configuration to obtain multiple initial inspection models;

[0142] In this step, during the first iteration, the initial position of each sand dune cat individual can be obtained based on the position initialization formula, and multiple groups of parameter configuration schemes corresponding to multiple initial positions can be configured in multiple initial models to obtain multiple initial inspection models; in each subsequent iteration, based on the position update method in step e, the new position of each sand dune cat individual is obtained, and multiple groups of parameter configuration schemes corresponding to the new positions of multiple sand dune cat individuals are configured in multiple initial inspection models to obtain multiple initial inspection models.

[0143] In a specific implementation, the initial position of each individual sand dune cat can be obtained in the following manner:

[0144] Based on the position initialization formula, the initial position of each individual sand dune cat is obtained;

[0145] Wherein, the position initialization formula is:

[0146] x ij =lb j +rand(0,1)·(ub j -lb j )

[0147] x ij represents the value of the j-dimensional variable of the i-th sand dune cat individual; lb j represents the upper bound of the j-th dimension variable; ub j represents the lower bound of the j-th dimension variable; j=1 represents the learning rate, j=2 represents the number of neurons in the hidden layer, and j=3 represents the regularization coefficient.

[0148] Specifically, the upper and lower bounds of the learning rate, the number of neurons in the hidden layer, and the regularization coefficient are pre-set, and the values ​​are not limited in this application. For example, the range of the learning rate is pre-set to [0.001, 0.99], the range of the number of neurons in the hidden layer is [3, 5], and the range of the regularization coefficient is [-1, 1].

[0149] In a possible implementation, in order to make the initial position distribution of multiple sand dune cat individuals more uniform and expand the coverage of the search space, the following method can be used to obtain multiple three-dimensional vectors to replace rand(0,1) in the position initialization formula. The method includes:

[0150] Based on the random sequence generation formula, and a randomly obtained three-dimensional vector Get N three-dimensional vectors; N is the number of pre-set sand dune cat individuals, M 11 , M 12 , M 13 are all random numbers between (0,1); the random sequence generation formula is:

[0151] M (i+1)j =aM ij (1-M ij )

[0152] In the above formula, M ij represents the random term of the j-dimensional variable of the i-th sand dune cat individual, 1≤i≤N, 1≤j≤3; a is the control parameter; M T represents the value of the Tth iteration;

[0153] Correspondingly, the position initialization formula corresponds to: x ij =lb j +M ij ·(ub j -lb j ).

[0154] In this implementation, the random sequence generation formula is used to obtain random items that are more evenly distributed than those of the ordinary rand(0,1) algorithm, so that the initial positions of the sand cat individuals are more evenly distributed, thereby expanding the coverage of the search space and helping to accelerate the efficiency of the hyperparameter optimization process.

[0155] Step c, selecting multiple samples from the sample set and inputting them into different initial inspection models for line fault inspection processing, obtaining the inspection results output by each initial inspection model, and determining the fitness value of each group of parameter configuration schemes based on the output results of each initial inspection model and the fault information in the input samples by using the accuracy function of the sand dune cat swarm optimization algorithm;

[0156] In this step, it is necessary to select some sample data from the sample set to train and verify different initial inspection models, and determine the precision of each initial inspection model according to the number of true positive examples and false positive examples in the verification results, and then determine the fitness value of the parameter configuration scheme corresponding to each initial inspection model, where the precision function is: Prec = TP / (TP+FP), TP represents the number of predicted true positive examples; FP represents the number of predicted false positive examples.

[0157] Step d: compare the highest fitness value of this round with the highest fitness value of the previous round, and select a set of parameter configuration schemes with the highest fitness value as the target parameter configuration scheme of this round;

[0158] Step e, if the preset iteration termination condition is not met, then according to the target parameter configuration scheme, the R value of each sand dune cat individual in the sand dune cat group optimization algorithm, the position of the multiple sand dune cat individuals is updated using the position update formula to obtain new multiple groups of parameter configuration schemes, and steps b to e are repeated until the iteration termination condition is met, and the target parameter configuration scheme obtained by the last iteration is used as the final parameter configuration scheme;

[0159] In this step, in order to more quickly determine the best hyperparameter combination, it is necessary to continuously adjust the positions of the sand dune cat individuals so that multiple sand dune cat individuals continuously search for the optimal solution (the behavior corresponding to the sand dune cat individual is the search behavior), or move toward the optimal solution (the behavior corresponding to the sand dune cat individual is the attack behavior). If the current number of iterations is less than the maximum number of iterations, it is determined that the preset iteration condition is not met.

[0160] Optionally, the position of each individual sand dune cat can be updated as follows:

[0161] Step e.1, determine the best position for this iteration according to the target parameter configuration scheme;

[0162] Step e.2, calculate the R value of each individual sand dune cat according to the sand dune cat group optimization algorithm, determine the behavior of each individual sand dune cat, wherein the behavior of the individual sand dune cat includes attack behavior and search behavior, and the R value is calculated using the following formula:

[0163]

[0164] R=2·r G rand(0,1)-r G

[0165] In the above formula, S M is the sensitivity of the preset sand dune cat individual; T represents the maximum number of iterations; t represents the current number of iterations; r G represents the sensitivity range of individual sand dune cats in the current iteration;

[0166] In this step, the random number rand(0,1) taken by each sand dune cat individual when calculating the R value may be different. When |R|≤1, the behavior of the sand dune cat individual is attacking behavior; when |R|>1, the behavior of the sand dune cat individual is searching behavior.

[0167] Step e.3: According to the behavior of each individual sand dune cat, the position is updated using the position update formula corresponding to the behavior to obtain the updated position of each individual sand dune cat.

[0168] In this step, for each individual sand cat, if the behavior of the individual sand cat is an attack, the position of the individual sand cat in the next iteration is calculated using the following formula:

[0169]

[0170] in, Indicates the position of the individual sand dune cat in the next iteration; Indicates the best position of this iteration; rand(0,1) indicates any random number between (0,1); Pos rnd represents the random position of individual sand dune cats; represents the current position of the individual sand dune cat; θ represents the predation angle of the randomly obtained individual sand dune cat, and θ∈(0°~360°);

[0171] If the behavior of the sand dune cat is a search behavior, the following formula is used to calculate the new position of the sand dune cat individual:

[0172]

[0173] In the above formula, Q is a random number that follows a normal distribution; L is a randomly generated 1×3 dimensional matrix, in which all values ​​are between (0,1).

[0174] Specifically, based on the above position update formula, the position of each individual sand dune cat is dynamically adjusted (that is, multiple parameter configuration schemes are continuously updated), so that multiple sand dune cat individuals continuously search for the optimal solution, or move towards the optimal solution, and quickly determine the best hyperparameter combination to achieve the optimization of the hyperparameters of the power inspection model.

[0175] In one possible implementation, in order to improve the iteration speed of the sand cat group optimization algorithm, before the position of the above-mentioned sand cat individual is updated, the sand cat individual farthest from the best position of the iteration can be determined based on the distance between each sand cat individual and the best position of the iteration, and the position of the sand cat individual can be directly updated to the best position.

[0176] Step f: configure the final parameter configuration scheme in the initial model to obtain a power inspection model.

[0177] In this step, the hyperparameter combination in the final parameter configuration scheme is the best hyperparameter combination. Therefore, it is necessary to configure the best hyperparameter combination in the initial model and use the complete sample set to carry out sample training on the model to obtain the final power inspection model.

[0178] In this implementation, based on the powerful global search capability of the improved Dune Cat Swarm Optimization Algorithm, a comprehensive and rapid exploration is conducted in the search space of the hyperparameters to achieve the optimization of the hyperparameters of the power inspection model. This method effectively avoids falling into the situation of local optimal solutions, ensuring that the hyperparameter combination adopted by the power inspection model is the best combination. Furthermore, a reasonable learning rate parameter can reduce the problem of slow convergence or failure to converge during the training process; reasonable hidden layer node parameters can not only enhance the analysis ability of the model, but also prevent the model from overfitting or underfitting; a reasonable regularization coefficient helps to improve the generalization ability of the model. Based on this, when the power inspection model using the best hyperparameter combination is used in the subsequent data analysis and processing, more accurate predictions and judgments can be made on the data in different power lines, improving the accuracy and rapidity of the power inspection model in analyzing and processing data.

[0179] In a specific implementation, in the actual application of the power model, the power inspection model can be optimized based on the inspection results of each power line and the inspection feedback data returned by the terminal equipment corresponding to each power line.

[0180] In this step, in order to further optimize and adjust the power inspection model to improve its prediction accuracy, the terminal device corresponding to each power line can use the accuracy, precision, recall and F value four indicators to evaluate the acquired power inspection model, and return the inspection feedback data to the server according to the evaluation results; the server optimizes the power inspection model based on the inspection feedback data returned by the terminal device. Specifically, the following formula can be used to calculate the evaluation index:

[0181] Acc=(TP+TN) / (TP+TN+FP+FN)

[0182] Prec=TP / (TP+FP)

[0183] Rec=TP / (TP+FN)

[0184] F value = 2 × (Prec × Rec) / (Prec + Rec)

[0185] In the above formula, Acc represents accuracy; Prec represents precision; Rec represents recall; TP represents the number of predicted true positive examples; FP represents the number of predicted false positive examples; TN represents the number of predicted true negative examples; and FN represents the number of predicted false negative examples.

[0186] Specifically, the hyperparameters of the model may be optimized based on the above-mentioned performance evaluation data fed back during the actual application of the power inspection model.

[0187] In this implementation, the power inspection model is optimized and adjusted based on the inspection results of the model in actual applications and the inspection feedback data corresponding to the inspection results, so as to continuously improve the prediction accuracy and stability of the model and reduce the error rate of the inspection results.

[0188] The smart line inspection management system and method based on deep learning provided in the embodiment of the present application, on the basis of the above embodiment, also provides a method for constructing and training a power inspection model. In the process of model training, based on the global search strategy and algorithm advantages of the improved population optimization algorithm, the best hyperparameter combination is obtained, and then the power inspection model with the best performance is obtained. Based on the power inspection model, the data reported by the data acquisition device is analyzed and processed, and the inspection results of the power grid are obtained in real time. Based on the performance advantages of the power inspection model itself, the accuracy and stability of data analysis can be ensured.

[0189] Based on the above embodiments, Figure 4 A structural diagram of a smart line intelligent inspection management system based on deep learning is provided for the third embodiment of the present application, such as Figure 4 As shown, the deep learning-based smart line intelligent inspection and management system 10 also includes: multiple terminal devices 103 and multiple inspection robots 104 for managing power lines in different areas.

[0190] Among them, the terminal device 103 can be a portable mobile communication terminal (such as a smart phone or a tablet computer), a computer device, a cockpit management terminal, and a large indicator screen; the inspection robot 104 can include an indoor inspection robot and an outdoor inspection robot.

[0191] Figure 5 A flowchart of a smart line intelligent inspection management system and method based on deep learning is provided for the third embodiment of the present application, such as Figure 5 As shown, based on the above embodiments, the smart line intelligent inspection management system based on deep learning provided by the present application can also take the following measures to warn and process abnormal inspection results, and the method includes:

[0192] S401. The server generates a regional power inspection report based on the inspection results of each power line. Each regional power inspection report includes the inspection results of all power lines within the corresponding area and the alarm corresponding to the power line that may fail.

[0193] In this step, for all power lines within a region, based on the power inspection model, the fault information of each power line can be obtained. Furthermore, if there is a line that is about to fail in the power line, the server will also add the alarm information corresponding to the line that is about to fail to the inspection report to instruct the supervisor to perform corresponding maintenance.

[0194] S402. The server pushes regional power inspection reports of different areas to corresponding terminal devices for display.

[0195] S403: The terminal device displays a regional power inspection report within the corresponding area.

[0196] In this solution, supervisors in the corresponding area can promptly perform maintenance on the lines based on the alarm information in the power inspection report displayed on the terminal device, thereby achieving supervision of all power lines in their area.

[0197] Alternatively, the inspection robot can quickly respond to the fault information in the inspection results, as follows:

[0198] S501. The server sends a fault warning instruction to the inspection robot within a preset range of the power line where a fault may occur.

[0199] In this step, each power line is equipped with a patrol robot within a preset range. When the server finds that there is a power line that is about to fail based on the power patrol results, it sends a fault warning instruction to the patrol robot within the preset range where the circuit that is about to fail is located.

[0200] The preset range for each power line can be determined according to the actual configuration of the robot and the size of the inspection area, and this application does not limit this. For example, when the inspection area is small, or the inspection area is large but there are many inspection robots configured in it, the preset range for the power line within the inspection area can be small, for example, 100 meters.

[0201] S502: The inspection robot that receives the fault warning instruction cuts off the power line where the fault will occur according to the fault warning instruction.

[0202] In this scheme, if the inspection robot receives a fault warning instruction, it will use the power line identification information in the instruction information to cut off the power line that is about to fail, effectively preventing the adverse effects of the line failure and achieving rapid supervision of the power inspection model in its area.

[0203] In a specific implementation method, after receiving a fault warning instruction, the inspection robot can send a tripping instruction to the intelligent circuit breaker installed on the power line that is about to fail as indicated by the instruction information through wireless communication; the intelligent circuit breaker controls the relay protection device or circuit breaker and other devices of the power line based on the tripping instruction to trigger the tripping operation.

[0204] The deep learning-based smart line intelligent inspection management system and method provided in the embodiment of the present application pushes the alarm information in the power inspection results to the terminal equipment corresponding to the line that is about to fail, so that the supervisor can repair the line in time based on the alarm information; or sends it to the inspection robot corresponding to the faulty line, and the inspection robot cuts off the power to the line that is about to fail in time based on the alarm information, thereby realizing rapid early warning and maintenance of the line that is about to become abnormal. This solution not only reduces the dependence of power inspection on manual labor, but also can complete the inspection task more quickly and accurately compared to manual inspection, and can predict and warn whether the line is about to fail before the line fails, thereby reducing the maintenance cost of the line failure and reducing the unnecessary losses caused by the line failure to a certain extent.

[0205] Those skilled in the art will readily appreciate other embodiments of the present application after considering the specification and practicing the invention disclosed herein. The present application is intended to cover any modification, use or adaptation of the present application, which follows the general principles of the present application and includes common knowledge or customary techniques in the art that are not disclosed in the present application. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present application are indicated by the following claims.

[0206] It should be understood that the present application is not limited to the precise structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present application is limited only by the appended claims.

Claims

1. A smart line intelligent inspection management system based on deep learning, characterized in that: include: A data acquisition device is arranged in the area to be inspected, and is used to collect monitoring data and environmental data on different power lines; A server, the server being connected to the data acquisition device and being used to receive monitoring data and environmental data on different power lines reported in real time by the data acquisition device; The server is also used to inspect the power lines using a power inspection model according to the monitoring data and environmental data of each power line within a preset time period to obtain inspection results, wherein the inspection results include whether a fault will occur and the type of fault when a fault occurs; wherein the power inspection model is a model that can predict and identify power line faults obtained by optimizing and training a long short-term memory network using an improved population optimization algorithm; The server is also used to push the inspection result of each power line to the relevant terminal equipment.

2. The system according to claim 1, characterized in that The data acquisition device includes a plurality of monitoring devices, a temperature sensor, a humidity sensor and an image acquisition device for detecting current, voltage and load information set for each power line; Wherein, the plurality of monitoring devices are used to detect the current, voltage and load information of the power line, and the monitoring data of each power line includes the current, voltage and load information; The temperature sensor is used to detect the temperature of the environment data where the power line is located, and the humidity sensor is used to detect the humidity of the environment where the power line is located; The image acquisition device is used to acquire image data of the power line.

3. The system according to claim 2, characterized in that The server is specifically used for: For each power line, according to the current, voltage, load information, temperature, humidity and image data reported by the data acquisition device, a characteristic data matrix of the power line is obtained, wherein the characteristic data matrix includes the current, voltage, load information, temperature, humidity and image recognition results at multiple time points, and the image recognition results are obtained by performing abnormal recognition on the image data; Inputting the characteristic data matrix of the power line into the power inspection model, performing forward feature extraction on the features of each dimension in chronological order through the power inspection model to obtain a forward feature sequence, performing reverse feature extraction processing on the features of each dimension in chronological order to obtain a reverse feature sequence, and fusing the obtained forward feature sequence and reverse feature sequence of the features of each dimension to obtain a fused feature sequence; According to the feature weight of each dimension, the fused feature sequences of multiple dimensions are fused again to obtain the time series features; According to the time series characteristics, the power grid is analyzed and processed to obtain real-time inspection results of the power grid.

4. The system according to any one of claims 1 to 3, characterized in that: The server is also used to: Based on the historical data of multiple power lines acquired in advance, a sample set is acquired, wherein the sample set includes multiple samples, each sample includes the current, voltage, load information, temperature, humidity, image recognition result, and fault information of a power line at multiple time points within a set time period, wherein the fault information includes whether a fault occurs within a preset time period after the set time period, and the type of fault when a fault occurs; the length of the set time period is consistent with the length of the preset time period; Constructing multiple initial models with exactly the same structure, each initial model includes a bidirectional LSTM network, a self-attention layer, a fully connected layer, and a classification output layer, wherein the bidirectional LSTM network includes a two-layer LSTM network for forward and reverse feature extraction; Based on the sample set and the multiple initial models, model training is performed, and during the training process, an improved sand cat population optimization algorithm is used to optimize multiple hyperparameters of the model to obtain the power inspection model.

5. The system according to claim 4, characterized in that The server is specifically used for: Step a, initializing the maximum number of iterations of the improved sand cat swarm optimization algorithm and a plurality of sand cat individuals, the position of each sand cat individual is used to represent a set of parameter configuration schemes, each set of parameter configuration schemes includes a learning rate, the number of hidden layer neurons and a regularization coefficient; Step b: for each initial model, a set of parameter configuration schemes is selected to perform parameter configuration to obtain multiple initial inspection models; Step c, selecting a plurality of samples from the sample set and inputting them into different initial inspection models for line fault inspection processing, obtaining the inspection results output by each initial inspection model, and determining the fitness value of each group of parameter configuration schemes based on the output results of each initial inspection model and the fault information in the input samples by using the accuracy function of the sand dune cat swarm optimization algorithm; Step d: compare the highest fitness value of this round with the highest fitness value of the previous round, and select a set of parameter configuration schemes with the highest fitness value as the target parameter configuration scheme of this round; Step e: if the preset iteration termination condition is not met, then according to the target parameter configuration scheme, the R value of each sand dune cat individual in the sand dune cat group optimization algorithm, the position of the multiple sand dune cat individuals is updated using the position update formula to obtain new multiple groups of parameter configuration schemes, and steps b to e are repeated until the iteration termination condition is met, and the target parameter configuration scheme obtained by the last iteration is used as the final parameter configuration scheme; Step f: configuring the final parameter configuration scheme in the initial model to obtain the power inspection model.

6. The system according to claim 5, characterized in that The server is also specifically used for: Based on the position initialization formula, the initial position of each individual sand dune cat is obtained; Wherein, the position initialization formula is: x ij =lb j +rand(0,1)·(ub j -lb j ) x ij represents the value of the j-dimensional variable of the i-th sand dune cat individual; lb j represents the upper bound of the j-th dimension variable; ub j represents the lower bound of the j-th dimension variable; j=1 represents the learning rate, j=2 represents the number of neurons in the hidden layer, and j=3 represents the regularization coefficient.

7. The system according to claim 5, characterized in that The server is also specifically used for: Determine the best position for this iteration according to the target parameter configuration scheme; The R value of each individual sand dune cat is calculated according to the sand dune cat group optimization algorithm, and the behavior of each individual sand dune cat is determined. The behavior of the individual sand dune cat includes attack behavior and search behavior, wherein the R value is calculated using the following formula: R=2·r G ·rand(0,1)-r G In the above formula, S M is the sensitivity of the preset sand dune cat individual; T represents the maximum number of iterations; t represents the current number of iterations; r G represents the sensitivity range of individual sand dune cats in the current iteration; According to the behavior of each individual sand dune cat, the position is updated using a position update formula corresponding to the behavior to obtain the updated position of each individual sand dune cat.

8. The system according to any one of claims 1 to 3, characterized in that: The system further includes: a plurality of terminal devices and a plurality of inspection robots for managing power lines within different areas; The server is also specifically used to generate a regional power inspection report based on the inspection results of each power line, each regional power inspection report includes the inspection results of all power lines within the corresponding area, and the alarm corresponding to the power line that will fail; The server is also used to push regional power inspection reports in different areas to corresponding terminal devices for display; The server is also used to send a fault warning instruction to the inspection robot within a preset range of the power line where the fault will occur; The inspection robot that receives the fault warning instruction will cut off the power supply of the faulty power line according to the fault warning instruction.

9. The system according to any one of claims 1 to 3, characterized in that: The server is also used to: The power inspection model is optimized based on the inspection result of each power line and the inspection feedback data returned by the terminal equipment corresponding to each power line.

10. A smart line intelligent inspection management method based on deep learning, characterized in that: The method applied to the server in the smart line intelligent inspection management system based on deep learning according to any one of claims 1 to 9 comprises: Receive monitoring data and environmental data on different power lines reported in real time by data acquisition devices; For each power line, according to the monitoring data and environmental data of the power line within a preset time period, a power inspection model is used to inspect the power line to obtain an inspection result, wherein the inspection result includes whether a fault will occur and the type of fault when a fault occurs; wherein the power inspection model is a model capable of predicting and identifying power line faults obtained by optimizing and training a long short-term memory network using an improved population optimization algorithm; Push the inspection results of each circuit line to the relevant terminal equipment.

Images