CNN deep learning-based power transmission channel vegetation intrusion identification method

Abstract

The invention discloses a CNN deep learning-based power transmission channel vegetation intrusion identification method, and relates to the technical field of computer vision and deep learning, and the method comprises the steps: collecting the real-time data of a power transmission channel, carrying out the preprocessing of the real-time data, and integrating the data into a database; a power transmission channel vegetation invasion evaluation model is constructed through the database, and the influence of vegetation invasion on the power transmission channel is evaluated; and comprehensively evaluating the state of the power transmission channel based on a power transmission channel vegetation invasion evaluation result, and analyzing potential problems and maintenance requirements of the power transmission channel. According to the method, the accuracy and reliability of evaluation are improved by constructing a CNN deep learning-based power transmission channel vegetation intrusion evaluation model and combining time sequence analysis and an environmental factor influence function; data features are extracted through the convolutional neural network, an attention mechanism is added, key features of the vegetation invasion area are recognized and enhanced, and the learning ability and recognition accuracy of the model are improved.

Description

Technical Field

[0001] The present invention relates to the field of computer vision and deep learning technology, and specifically to a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning. Background Art

[0002] With the development of modern power systems, the safe management of transmission channels has become increasingly important. Transmission lines usually cross a variety of terrains and environments, such as mountainous areas and forests. Vegetation grows rapidly and variably, which can easily lead to vegetation invading transmission channels. Vegetation intrusion may cause equipment short circuits, fires, and even system failures, seriously affecting the reliability and safety of the power system.

[0003] Traditional vegetation intrusion monitoring methods mainly rely on manual inspections or simple image processing techniques. These methods have disadvantages such as low efficiency, susceptibility to subjective factors, and failure to detect problems in a timely manner. With the rapid development of computer vision and deep learning technology, especially the successful application of convolutional neural networks (CNN) in the field of image recognition, new solutions have been provided to solve the problem of vegetation intrusion identification in power transmission channels.

[0004] The key advantage of deep learning is that it can automatically learn and extract complex features from large-scale data without the need for manual design and feature extraction. CNN, as one of the representative models of deep learning, can effectively capture and learn spatial hierarchical features in images through multi-layer convolution and pooling operations. It is suitable for processing complex remote sensing image data. In the identification of vegetation intrusion in power transmission channels, CNN technology can identify areas covered by vegetation and distinguish which vegetation has invaded the channel area. This method not only improves the accuracy of recognition, but also realizes automated monitoring and timely warning, greatly improving the operation safety and management efficiency of the power system. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is: the existing method for identifying vegetation intrusion in transmission channels has low efficiency, poor accuracy, and poor real-time performance, as well as the problem of how to identify potential problems and maintenance needs of transmission channels, and provide a better maintenance method for transmission channels.

[0007] To solve the above technical problems, the present invention provides the following technical solutions: a method for identifying vegetation intrusion in a transmission channel based on CNN deep learning, comprising collecting real-time data of the transmission channel, preprocessing the real-time data and integrating it into a database; constructing a vegetation intrusion assessment model for the transmission channel through the database to assess the impact of vegetation intrusion on the transmission channel; based on the assessment results of vegetation intrusion in the transmission channel, comprehensively assessing the status of the transmission channel, and analyzing potential problems and maintenance needs of the transmission channel.

[0008] As a preferred solution of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning described in the present invention, the real-time data of the power transmission channel includes environmental impact image data, drone image data, ground sensor image data and fixed camera image data; the environmental impact image data includes environmental image data at different times, environmental image data under different weather conditions and environmental image data in different seasons.

[0009] As a preferred solution of the method for identifying vegetation intrusion in power transmission channels based on CNN deep learning described in the present invention, the preprocessing includes denoising the image data using a Gaussian filter, which is expressed as:

[0010]

[0011] Among them, I filtered (P,Q) is the pixel value with coordinates (P,Q) in the image after Gaussian filtering, I(P+p,Q+q) is the pixel value with coordinates (P+p,Q+q) in the original image, G(p,q) is the weight value with coordinates (p,q) in the Gaussian kernel, and d is the number of Gaussian kernels; the image is normalized and expressed as:

[0012]

[0013] Among them, I norm (P,Q) is the pixel value with coordinates (P,Q) in the standardized image, I(P,Q) is the pixel value with coordinates (P,Q) in the original image, μ is the average value of all pixel values ​​in the original image, and σ is the standard deviation of all pixel values ​​in the original image. The collected real-time data is annotated to mark the vegetation intrusion area and the safe area in the image, distinguish different types of vegetation and the safe distance between different types of vegetation and transmission lines, generate training sets and test sets, and store the preprocessed data in the database.

[0014] As a preferred solution of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning described in the present invention, the construction of a vegetation intrusion assessment model for a power transmission channel includes constructing a vegetation intrusion assessment model for a power transmission channel based on data collected from a database, which is expressed as:

[0015]

[0016] Among them, I VI is the vegetation invasion impact index of the transmission channel, n is the number of image data, L i is the i-th image data, C i is the sensor data of the i-th image, T iis the acquisition time of the i-th image, E i is the soil moisture environmental factor of the i-th image, F i is the temperature environment factor of the i-th image, f VI (L i ,C i ,T i ) is the vegetation intrusion recognition function based on image data and sensor data, g T (T i ) is the time series analysis function, h EF (E i ,F i ) is the environmental factor influence function; calculate the vegetation invasion recognition function f based on image data and sensor data VI (L i ,C i ,T i ), expressed as:

[0017]

[0018] Among them, τ is the activation function, k is the number of convolution kernels, and w j is the weight of the jth convolution kernel, CNN is a convolutional neural network; based on the convolutional neural network, the image data is subjected to convolution layer operations, features are extracted, pooling is performed, features are compressed, fully connected layer processing is performed, features are expanded, and classification is performed through the fully connected layer. Through the attention mechanism, the vegetation invasion area is located; the time series analysis function g is calculated T (T i ), expressed as:

[0019]

[0020] Where T is the length of the time series, a t is the parameter in the time series analysis model; calculate the environmental factor impact function h EF (E i ,F i ), expressed as:

[0021]

[0022] Among them, M is the number of environmental factors, E m is the mth soil moisture environmental factor, F m is the mth temperature environmental factor, b m For E m The weight, c m F m Based on knowledge distillation, the vegetation intrusion assessment model of the power transmission channel is trained in the training set through knowledge distillation, and the loss function is calculated, which is expressed as:

[0023]

[0024] Among them, L KD is the loss function, α is the balance coefficient, L CE is the cross entropy loss function, y is the true label, is the predicted output of the student model, R is the temperature parameter, ρ R is the Softmax function with temperature parameter R, z is the output of the teacher model, and z S is the output of the student model; the vegetation intrusion assessment model of the transmission channel is evaluated in the test set, and the accuracy of the model evaluation index is calculated, which is expressed as:

[0025]

[0026] Among them, Accuracy is the accuracy, TP is the true positive, TN is the true negative, FP is the false positive, and FN is the false negative; the recall rate of the calculation model evaluation index is expressed as:

[0027]

[0028] Among them, Recall is the recall rate; the precision rate of the calculation model evaluation index is expressed as:

[0029]

[0030] Among them, Precision is the accuracy; the calculation model evaluation indicator F1 score is expressed as:

[0031]

[0032] Among them, F1 is the harmonic mean of precision and recall.

[0033] As a preferred solution of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning described in the present invention, wherein: evaluating the impact of vegetation intrusion on the power transmission channel includes evaluating the impact of vegetation intrusion on the power transmission channel based on the output value of the vegetation intrusion evaluation model of the power transmission channel; when I VI When it is equal to 0, it means that there is no vegetation invasion and no impact on the transmission channel; when I VI When it is greater than 0 and less than 1, it means that vegetation has invaded, which has an impact on the transmission channel and needs to be handled by maintenance personnel; when I VI When it is equal to 1, it means that the vegetation invasion is serious, causing a major impact on the transmission channel, and maintenance personnel are required to take emergency measures.

[0034] As a preferred solution of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning described in the present invention, the comprehensive evaluation of the power transmission channel status includes: based on the evaluation result of vegetation intrusion in the power transmission channel, the output value I of the vegetation intrusion evaluation model of the power transmission channel is converted into VI As an input parameter, the transmission channel status is comprehensively evaluated and the comprehensive score S of the transmission channel status is calculated, which is expressed as:

[0035]

[0036] Where N is the number of transmission channels, ε l is the vegetation invasion weight of the lth transmission channel, I VIl is the vegetation invasion impact index of the lth transmission channel, β l is the maintenance demand factor of the lth transmission channel, D l is the number of failures of the lth transmission channel; when the comprehensive score S of the transmission channel status is equal to 0, it means that the status is the worst and immediate maintenance is required; when the comprehensive score S of the transmission channel status is equal to 1, it means that the status is the best and no maintenance is required.

[0037] As a preferred solution of the method for identifying vegetation intrusion in transmission channels based on CNN deep learning described in the present invention, the analysis of potential problems and maintenance needs of the transmission channel includes conducting data mining and analysis based on the comprehensive scoring results of the transmission channel status, using historical data and real-time data in the database to discover potential problems and risk factors of the transmission channel, predicting the maintenance needs of the transmission channel based on the data mining results, and optimizing the operation and maintenance strategy of the transmission channel according to the analysis results.

[0038] Another object of the present invention is to provide a transmission channel vegetation intrusion identification system based on CNN deep learning, which can construct a transmission channel vegetation intrusion assessment model through a database, evaluate the impact of vegetation intrusion on the transmission channel, and solve the current problem of low efficiency of transmission channel vegetation intrusion identification.

[0039] As a preferred solution of the power transmission channel vegetation intrusion identification system based on CNN deep learning described in the present invention, it includes: a data processing module, a vegetation intrusion assessment module, and a comprehensive assessment module; the data processing module is used to collect real-time data of the power transmission channel, pre-process the real-time data and integrate it into the database; the vegetation intrusion assessment module is used to construct a power transmission channel vegetation intrusion assessment model through the database, and evaluate the impact of vegetation intrusion on the power transmission channel; the comprehensive assessment module is used to comprehensively evaluate the status of the power transmission channel based on the power transmission channel vegetation intrusion assessment result, and analyze the potential problems and maintenance needs of the power transmission channel.

[0040] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning.

[0041] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning.

[0042] The beneficial effects of the present invention are as follows: the power transmission channel vegetation intrusion identification method based on CNN deep learning provided by the present invention improves the accuracy and robustness of model identification by collecting multi-source data, can adapt to different environmental conditions and weather conditions, and realizes all-weather monitoring; through image denoising, normalization and annotation, the data quality, model training effect, model generalization ability and recognition accuracy are improved, and the false alarm rate and missed alarm rate are reduced; by constructing a power transmission channel vegetation intrusion assessment model based on CNN deep learning, and combining time series analysis and environmental factor influence functions, a more comprehensive assessment is achieved, and the accuracy and reliability of the assessment are improved; through convolutional neural network to extract data features and add attention mechanism, the key features of vegetation intrusion areas are identified and strengthened, so that the model pays more attention to key feature important areas during training, improves the learning ability and recognition accuracy of the model, and the present invention achieves better results in terms of accuracy, efficiency and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0044] Figure 1 An overall flow chart of a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning provided in the first embodiment of the present invention;

[0045] Figure 2 A module schematic diagram of a power transmission channel vegetation intrusion identification system based on CNN deep learning provided in the third embodiment of the present invention. DETAILED DESCRIPTION

[0046] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0047] Example 1, reference Figure 1 , is an embodiment of the present invention, and provides a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning, comprising:

[0048] S1: Collect real-time data of the transmission channel, pre-process the real-time data and integrate it into the database.

[0049] Furthermore, the real-time data of the transmission channel includes environmental impact image data, drone image data, ground sensor image data and fixed camera image data; the environmental impact image data includes environmental image data at different times, environmental image data under different weather conditions and environmental image data in different seasons.

[0050] It should be noted that the preprocessing includes denoising the image data using a Gaussian filter, which can be expressed as:

[0051]

[0052] Among them, I filtered (P,Q) is the pixel value with coordinates (P,Q) in the image after Gaussian filtering, I(P+p,Q+q) is the pixel value with coordinates (P+p,Q+q) in the original image, G(p,q) is the weight value with coordinates (p,q) in the Gaussian kernel, and d is the number of Gaussian kernels; the image is normalized and expressed as:

[0053]

[0054] Among them, I norm (P,Q) is the pixel value with coordinates (P,Q) in the standardized image, I(P,Q) is the pixel value with coordinates (P,Q) in the original image, μ is the average value of all pixel values ​​in the original image, and σ is the standard deviation of all pixel values ​​in the original image. The collected real-time data is annotated to mark the vegetation intrusion area and the safe area in the image, distinguish different types of vegetation and the safe distance between different types of vegetation and transmission lines, generate training sets and test sets, and store the preprocessed data in the database.

[0055] It should also be noted that the real-time data of the transmission channel is obtained using drones, ground sensors and fixed cameras., collect image and video data of the transmission channel from different perspectives. The drone can flexibly fly above and around the transmission lines to obtain data from high-altitude perspectives and capture hard-to-reach areas. Ground sensors and fixed cameras are installed on transmission towers or other key locations, and can continuously capture data from fixed angles, which is particularly suitable for long-term monitoring and timely detection of abnormal situations. The collected data covers different time, weather and lighting conditions, forming a diverse data set, providing rich samples for subsequent model evaluation and analysis. By using Gaussian filters to denoise the image data, the clarity and usability of the image are improved. By normalizing the images, images from different devices and different perspectives can be uniformly processed and analyzed by the system, avoiding errors caused by inconsistent data formats. By annotating the collected images and video data in detail, key information about vegetation intrusion areas and safe areas is clearly identified. This step provides important reference data for subsequent machine learning model training and risk assessment.

[0056] S2: Through the database, a vegetation intrusion assessment model for the transmission channel is constructed to evaluate the impact of vegetation intrusion on the transmission channel.

[0057] Furthermore, constructing a transmission channel vegetation intrusion assessment model includes constructing a transmission channel vegetation intrusion assessment model based on data collected from the database through a database, which is expressed as:

[0058]

[0059] Among them, I VI is the vegetation invasion impact index of the transmission channel, n is the number of image data, L i is the i-th image data, C i is the sensor data of the i-th image, T i is the acquisition time of the i-th image, E i is the soil moisture environmental factor of the i-th image, F i is the temperature environment factor of the i-th image, f VI (L i ,C i ,T i ) is the vegetation intrusion recognition function based on image data and sensor data, g T (T i ) is the time series analysis function, h EF (E i ,F i ) is the environmental factor influence function; calculate the vegetation invasion recognition function f based on image data and sensor data VI (L i ,C i ,T i ), expressed as:

[0060]

[0061] Among them, τ is the activation function, k is the number of convolution kernels, and w j is the weight of the jth convolution kernel, CNN is a convolutional neural network; based on the convolutional neural network, the image data is subjected to convolution layer operations, features are extracted, pooling is performed, features are compressed, fully connected layer processing is performed, features are expanded, and classification is performed through the fully connected layer. Through the attention mechanism, the vegetation invasion area is located; the time series analysis function g is calculated T (T i ), expressed as:

[0062]

[0063] Where T is the length of the time series, a t is the parameter in the time series analysis model; calculate the environmental factor impact function h EF (E i ,F i ), expressed as:

[0064]

[0065] Among them, M is the number of environmental factors, E m is the mth soil moisture environmental factor, F m is the mth temperature environmental factor, b m For E m The weight, c m F m Based on knowledge distillation, the vegetation intrusion assessment model of the power transmission channel is trained in the training set through knowledge distillation, and the loss function is calculated, which is expressed as:

[0066]

[0067] Among them, L KD is the loss function, α is the balance coefficient, L CE is the cross entropy loss function, y is the true label, is the predicted output of the student model, R is the temperature parameter, ρ R is the Softmax function with temperature parameter R, z is the output of the teacher model, and z S is the output of the student model; the vegetation intrusion assessment model of the transmission channel is evaluated in the test set, and the accuracy of the model evaluation index is calculated, which is expressed as:

[0068]

[0069] Among them, Accuracy is the accuracy, TP is the true positive, TN is the true negative, FP is the false positive, and FN is the false negative; the recall rate of the calculation model evaluation index is expressed as:

[0070]

[0071] Among them, Recall is the recall rate; the precision rate of the calculation model evaluation index is expressed as:

[0072]

[0073] Among them, Precision is the accuracy; the calculation model evaluation indicator F1 score is expressed as:

[0074]

[0075] Among them, F1 is the harmonic mean of precision and recall.

[0076] It should be noted that the assessment of the impact of vegetation intrusion on the transmission channel includes assessing the impact of vegetation intrusion on the transmission channel based on the output value of the vegetation intrusion assessment model of the transmission channel; when I VI When it is equal to 0, it means that there is no vegetation invasion and no impact on the transmission channel; when I VI When it is greater than 0 and less than 1, it means that vegetation has invaded, which has an impact on the transmission channel and needs to be handled by maintenance personnel; when I VI When it is equal to 1, it means that the vegetation invasion is serious, causing a major impact on the transmission channel, and maintenance personnel are required to take emergency measures.

[0077] It should also be noted that the transmission channel vegetation intrusion assessment model integrates the transmission channel vegetation intrusion data and environmental influencing factors to provide a comprehensive environmental impact assessment. This comprehensive assessment is very important for transmission channel maintenance because it can better understand the actual impact of vegetation intrusion on the transmission channel and formulate corresponding policies and measures accordingly. By considering the time series analysis function and analyzing the vegetation growth trend, it helps to identify long-term trends and potential environmental risks, so as to take preventive measures. By considering the environmental factor impact function and analyzing the impact of environmental factors on vegetation intrusion, the prediction ability of the model is also effectively improved. The model is based on a convolutional neural network, and an attention mechanism module is introduced in the convolutional layer to make the model pay more attention to potential vegetation intrusion areas and enhance recognition capabilities. The knowledge distillation technology is used to transfer complex teacher model knowledge to a lightweight student model, reducing the complexity and volume of the model while maintaining high performance. This method improves the operating efficiency and response speed of the model in a resource-constrained environment, and effectively solves the problem of model overfitting. The transmission channel vegetation intrusion assessment model determines the degree of impact of vegetation intrusion on the transmission channel by calculating the output value.

[0078] S3: Based on the results of the vegetation intrusion assessment on the transmission corridor, comprehensively assess the status of the transmission corridor and analyze the potential problems and maintenance needs of the transmission corridor.

[0079] Furthermore, the comprehensive evaluation of the transmission channel status includes converting the output value I of the transmission channel vegetation intrusion evaluation model into the output value I of the transmission channel vegetation intrusion evaluation model based on the transmission channel vegetation intrusion evaluation result. VI As an input parameter, the transmission channel status is comprehensively evaluated and the comprehensive score S of the transmission channel status is calculated, which is expressed as:

[0080]

[0081] Where N is the number of transmission channels, ε l is the vegetation invasion weight of the lth transmission channel, I VIl is the vegetation invasion impact index of the lth transmission channel, β l is the maintenance demand factor of the lth transmission channel, D l is the number of failures of the lth transmission channel; when the comprehensive score S of the transmission channel status is equal to 0, it means that the status is the worst and immediate maintenance is required; when the comprehensive score S of the transmission channel status is equal to 1, it means that the status is the best and no maintenance is required.

[0082] It should be noted that the analysis of potential problems and maintenance needs of transmission channels includes conducting data mining and analysis based on the comprehensive scoring results of the transmission channel status, using historical data and real-time data in the database to discover potential problems and risk factors of the transmission channel, predicting the maintenance needs of the transmission channel based on the data mining results, and optimizing the operation and maintenance strategy of the transmission channel according to the analysis results.

[0083] It should also be noted that by comprehensively considering multiple aspects of vegetation intrusion and transmission channel operation, an accurate assessment of the status of the transmission channel can be provided, which is crucial for power system managers because it can help identify which transmission channels need priority attention and maintenance. Through a comprehensive assessment of the status of the transmission channel, managers can allocate resources more effectively and give priority to those transmission channels with severe vegetation intrusion or the most unstable operation. This not only improves the efficiency of maintenance work, but also helps to improve the overall performance of the transmission channel. By continuously monitoring and evaluating the vegetation intrusion of the transmission channel, it will help the power system develop in a more reliable direction and ensure that safe operation is fully considered in the management of the transmission channel. When the status score of the transmission channel exceeds the equipment status limit, it means that emergency intervention is required. Such a mechanism ensures timely maintenance of potential high-risk transmission channels, thereby reducing the negative impact that may be caused by transmission channel failures. Through historical data and real-time data in the database, data mining and analysis are carried out to discover potential problems and risk factors of the transmission channel, which can help power system managers better understand the patterns and trends of transmission channel failures and maintenance needs, thereby formulating more effective long-term maintenance strategies.

[0084] Example 2 is an embodiment of the present invention, which provides a method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculation and simulation experiments.

[0085] First, real-time data of the transmission channel was collected, including selecting 10 transmission lines in a certain area of ​​my country, using drones, ground sensors, and fixed camera equipment to collect environmental impact image data at different times (morning, noon, and evening), different weather (sunny, cloudy, and rainy), and different seasons (spring, summer, autumn, and winter). A total of 10,000 image data were collected. Next, preprocessing was performed, including denoising the collected image data, filtering the image with a Gaussian filter, and then normalizing it. The vegetation intrusion area and safety area in the image were marked. A total of 5,000 data were marked, including 2,000 vegetation intrusion areas and 3,000 safety areas. Next, based on the preprocessed data, a transmission channel vegetation intrusion assessment model was constructed, and the transmission channel vegetation intrusion impact index I was output. VIThe model adopts CNN deep learning algorithm, extracts features through structures such as convolution layer, pooling layer, and fully connected layer, and uses attention mechanism to locate vegetation intrusion areas. The knowledge distillation method is used to train the model, calculate the loss function, evaluate the model performance on the test set, and calculate the accuracy, recall rate, precision and F1 score. Next, the transmission channel status is comprehensively evaluated based on the vegetation intrusion impact index of the transmission channel, and the comprehensive score S of the transmission channel status is output. Finally, based on the comprehensive score result of the transmission channel status, the historical data and real-time data in the database are used to conduct data mining and analysis to discover potential problems and risk factors of the transmission channel. Based on the data mining results, the maintenance needs of the transmission channel are predicted. According to the analysis results, the operation and maintenance strategy of the transmission channel is optimized.

[0086] Refer to Table 1 to record and analyze the experimental data.

[0087] Table 1 Experimental data record table

[0088]

[0089]

[0090] Through the comparison of 10 transmission lines, it can be seen that the model accuracy rates are all above 0.95, and the highest can reach 0.99, indicating that the method of the present invention has high accuracy in identifying vegetation intrusion. In terms of recall rate, precision rate and F1 score, the method of the present invention also shows good performance, all of which are above 0.95, which shows that the method of the present invention has high recall rate and precision rate when identifying vegetation intrusion. The vegetation intrusion impact index I VI The distribution ranges from 0.2 to 0.5, which indicates that the vegetation invasion degree of different transmission lines is different. VI The higher the value, the greater the impact of vegetation invasion on the transmission channel. The comprehensive score S of the transmission channel status is distributed between 0.6 and 0.8, reflecting the overall status of the transmission channel. The lower the S value, the worse the status of the transmission channel is, requiring more maintenance and attention. By comparing I VI value and S value, we can find that I VI Transmission lines with higher values ​​tend to have lower S values, which indicates that vegetation intrusion does have a negative impact on the state of the transmission channel. The method of the present invention identifies vegetation intrusion and calculates I VI , which helps to discover potential risks in advance and take measures to improve the status of the transmission channel. Compared with the existing technology, the method of the present invention has significant advantages in identifying vegetation intrusion compared with the traditional manual inspection method. Manual inspection is easily affected by subjective factors and has low efficiency. The method of the present invention is based on CNN deep learning technology, which can realize fast and accurate identification of vegetation intrusion and improve the operation and maintenance efficiency of the transmission channel. Therefore, our invention is creative.

[0091] Example 3, reference Figure 2 , is an embodiment of the present invention, and provides a power transmission channel vegetation intrusion identification system based on CNN deep learning, including a data processing module, a vegetation intrusion assessment module, and a comprehensive assessment module.

[0092] The data processing module is used to collect real-time data of the transmission channel, pre-process the real-time data and integrate it into the database; the vegetation intrusion assessment module is used to construct a vegetation intrusion assessment model for the transmission channel through the database, and evaluate the impact of vegetation intrusion on the transmission channel; the comprehensive assessment module is used to comprehensively evaluate the status of the transmission channel based on the vegetation intrusion assessment results of the transmission channel, and analyze the potential problems and maintenance needs of the transmission channel.

[0093] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0094] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0095] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering, or processing in another suitable manner as necessary, and then stored in a computer memory.

[0096] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc. It should be noted that the above embodiments are only used to illustrate the technical solution of the present invention and are not limited. Although the present invention is described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

[0097] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying vegetation intrusion in power transmission channels based on CNN deep learning, characterized in that: include: Collect real-time data of power transmission channels, pre-process the real-time data and integrate it into the database; Through the database, a vegetation intrusion assessment model for power transmission channels is constructed to assess the impact of vegetation intrusion on power transmission channels; Based on the results of the vegetation intrusion assessment on the transmission channel, the status of the transmission channel is comprehensively assessed, and the potential problems and maintenance needs of the transmission channel are analyzed.

2. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 1, characterized in that: The real-time data of the power transmission channel includes environmental impact image data, drone image data, ground sensor image data and fixed camera image data; The environmental impact image data includes environmental image data at different times, environmental image data at different weather conditions, and environmental image data at different seasons.

3. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 2, characterized in that: The preprocessing includes using a Gaussian filter to perform denoising on the image data, which is expressed as: Among them, I filtered (P,Q) is the pixel value with coordinates (P,Q) in the image after Gaussian filtering, I(P+p,Q+q) is the pixel value with coordinates (P+p,Q+q) in the original image, G(p,q) is the weight value with coordinates (p,q) in the Gaussian kernel, and d is the number of Gaussian kernels; The image is normalized and expressed as: Among them, I norm (P,Q) is the pixel value with coordinates (P,Q) in the standardized image, I(P,Q) is the pixel value with coordinates (P,Q) in the original image, μ is the average value of all pixel values ​​in the original image, and σ is the standard deviation of all pixel values ​​in the original image; The collected real-time data is annotated to mark the vegetation intrusion areas and safe areas in the image, distinguish different types of vegetation and the safe distance between different types of vegetation and transmission lines, generate training sets and test sets, and store the preprocessed data in the database.

4. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 3, characterized in that: The construction of the transmission channel vegetation intrusion assessment model includes constructing the transmission channel vegetation intrusion assessment model through a database based on data collected from the database, which is expressed as: Among them, I VI is the vegetation invasion impact index of the transmission channel, n is the number of image data, L i is the i-th image data, C i is the sensor data of the i-th image, T i is the acquisition time of the i-th image, E i is the soil moisture environmental factor of the i-th image, F i is the temperature environment factor of the i-th image, f VI (L i ,C i ,T i ) is the vegetation intrusion recognition function based on image data and sensor data, g T (T i ) is the time series analysis function, H EF (E i ,F i ) is the environmental factor influence function; Calculate the vegetation intrusion identification function f based on image data and sensor data VI (L i ,C i ,T i ), expressed as: Among them, τ is the activation function, k is the number of convolution kernels, and w j is the weight of the jth convolution kernel, CNN is a convolutional neural network; Based on the convolutional neural network, the image data is subjected to convolution layer operations, features are extracted, pooling processing is performed, features are compressed, fully connected layer processing is performed, features are expanded, and classification is performed through the fully connected layer. Through the attention mechanism, the vegetation intrusion area is located; Calculate the time series analysis function g T (T i ), expressed as: Among them, T is the length of the time series, a t Parameters in time series analysis models; Calculate the environmental factor impact function h EF (E i ,F i ), expressed as: Among them, M is the number of environmental factors, E m is the mth soil moisture environmental factor, F m is the mth temperature environmental factor, b m For E m The weight of m F m The weight of Based on knowledge distillation, the vegetation intrusion assessment model of the power transmission channel is trained in the training set through knowledge distillation, and the loss function is calculated, which is expressed as: Among them, L KD is the loss function, α is the balance coefficient, L CE is the cross entropy loss function, y is the true label, is the predicted output of the student model, R is the temperature parameter, ρ R is the Softmax function with temperature parameter R, z is the output of the teacher model, and z S Output for the student model; The transmission channel vegetation intrusion assessment model is evaluated in the test set, and the accuracy of the model evaluation index is calculated, which is expressed as: Among them, Accuracy is the accuracy, TP is the true positive, TN is the true negative, FP is the false positive, and FN is the false negative; Calculate the recall rate of the model evaluation indicator, expressed as: Among them, Recall is the recall rate; The accuracy of the calculation model evaluation index is expressed as: Among them, Precision is the accuracy; Calculate the model evaluation indicator F1 score, expressed as: Among them, F1 is the harmonic mean of precision and recall.

5. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 4, characterized in that: The assessing the impact of vegetation intrusion on the power transmission channel includes assessing the impact of vegetation intrusion on the power transmission channel based on the output value of the power transmission channel vegetation intrusion assessment model; When I VI When it is equal to 0, it means that there is no vegetation invasion and no impact on the transmission channel; When I VI When it is greater than 0 and less than 1, it indicates that vegetation has invaded, which has an impact on the power transmission channel and needs to be handled by maintenance personnel; When I VI When it is equal to 1, it means that the vegetation invasion is serious, causing a major impact on the transmission channel, and maintenance personnel are required to take emergency measures.

6. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 5, characterized in that: The comprehensive evaluation of the transmission channel status includes: based on the transmission channel vegetation intrusion evaluation result, the transmission channel vegetation intrusion evaluation model output value I VI As an input parameter, the transmission channel status is comprehensively evaluated and the comprehensive score S of the transmission channel status is calculated, which is expressed as: Where N is the number of transmission channels, ε l is the vegetation invasion weight of the lth transmission channel, I VIl is the vegetation invasion impact index of the lth transmission channel, β l is the maintenance demand factor of the lth transmission channel, D l is the number of failures in the lth transmission channel; When the comprehensive score S of the transmission channel status is equal to 0, it means the status is the worst and immediate maintenance is required; When the comprehensive score S of the transmission channel status is equal to 1, it means that the status is optimal and no maintenance is required.

7. The method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in claim 6, characterized in that: The analysis of potential problems and maintenance needs of the transmission channel includes performing data mining and analysis based on the comprehensive scoring results of the transmission channel status, using historical data and real-time data in the database to discover potential problems and risk factors of the transmission channel, predicting the maintenance needs of the transmission channel based on the data mining results, and optimizing the operation and maintenance strategy of the transmission channel according to the analysis results.

8. A system using the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as claimed in any one of claims 1 to 7, characterized in that: It includes data processing module, vegetation invasion assessment module and comprehensive assessment module; The data processing module is used to collect real-time data of the power transmission channel, pre-process the real-time data and integrate it into the database; The vegetation intrusion assessment module is used to construct a transmission channel vegetation intrusion assessment model through a database to assess the impact of vegetation intrusion on the transmission channel; The comprehensive assessment module is used to comprehensively assess the status of the power transmission channel based on the assessment results of vegetation intrusion into the power transmission channel, and analyze potential problems and maintenance needs of the power transmission channel.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for identifying vegetation intrusion in a power transmission channel based on CNN deep learning as described in any one of claims 1 to 7 are implemented.

Images