Power distribution network operation state monitoring method and system based on visual image

Abstract

The invention relates to a power distribution network operation state monitoring method and system based on a visual image, and belongs to the technical field of visual images, and the method comprises the steps: obtaining the image data of a power distribution network, and obtaining a corresponding feature complexity degree through a feature complexity formula according to the image data; according to the method, the image data of the power distribution network can be quickly responded, the complex feature class and the simple feature class are divided, the proper monitoring model is selected, the image data of the power distribution network can be quickly responded, and the problem that in the prior art, when the power distribution network breaks down, the monitoring model cannot meet the requirement for quick response is effectively solved.

Description

Technical Field

[0001] The invention relates to a method and system for monitoring the operating status of a distribution network based on visual images, and belongs to the technical field of visual images. Background Art

[0002] With the continuous growth of electricity demand and the advancement of intelligent power grid development, the distribution network is a key link in the power system that directly faces users. The accurate monitoring of its operating status is extremely important to ensure the reliability and stability of power supply. The distribution network is large in scale and complex in structure. Traditional monitoring methods are difficult to meet the growing demand for refined management.

[0003] The existing distribution network operation status monitoring system realizes the real-time collection of visual information such as the appearance, operating environment, and line connection status of distribution network equipment by reasonably deploying high-definition image acquisition equipment, such as high-definition cameras and intelligent image sensors, around key distribution network facilities and lines. The image data is deeply analyzed and processed to accurately identify the normal operating status of the equipment, fault characteristics, and abnormal changes in the surrounding environment, so as to comprehensively and intuitively grasp the operating status of the distribution network, timely discover potential safety hazards, and issue early warning signals.

[0004] For example, the invention patent with publication number: CN118367678A discloses a distribution network environmental safety monitoring system and method based on visual images, including: collecting location information corresponding to the distribution network information, formulating sensitive area types based on the distribution network information, formulating monitoring information of the distribution network information based on the monitoring items, and judging the operating status of the distribution network information based on the monitoring information, wherein the operating status includes a qualified state and an abnormal state, and warning the abnormal distribution network information in an abnormal state to the management end and synchronously prompting it in the distribution network model.

[0005] For example, the invention patent with announcement number: CN113241848B discloses a comprehensive monitoring system for a distribution network, including: collecting distribution network power consumption data through a power consumption information collection module, and sending the distribution network power consumption data to an intelligent gateway; a distribution equipment status monitoring module collects distribution equipment operating status data, and sends the distribution equipment operating status data to an intelligent gateway through a communication module; the intelligent gateway sends the distribution network power consumption data and the distribution equipment operating status data to an edge computing module for management; edge computing is performed on the distribution network power consumption data and the distribution equipment operating status data to obtain power consumption data monitoring result data.

[0006] However, in the process of implementing the technical solution of the invention in the embodiments of the present application, the present application found that the above technology has at least the following technical problems:

[0007] In the prior art, the distribution network operation status monitoring system monitors the distribution network by building a monitoring model through a deep learning algorithm. However, in the case of a large amount of complex data, the monitoring model requires a long reasoning time. When a distribution network fails, the monitoring model cannot meet the rapid response requirements. As a result, when the distribution network cannot operate normally, the staff cannot handle it quickly and effectively. Summary of the invention

[0008] According to the problem described in the background, the problem to be solved by the present invention is: to provide a distribution network operation status monitoring method and system based on visual images, which solves the problem in the prior art that the monitoring model cannot meet the demand for rapid response when a distribution network fails, and realizes that a suitable monitoring model can be selected according to the complexity of the features, thereby achieving a rapid response effect.

[0009] To achieve the above object, the present invention provides the following technical solution: a method for monitoring the operating status of a distribution network based on visual images is provided, comprising the following steps:

[0010] (1) photographing the distribution network through a drone, collecting multiple sets of photos of multiple parts of the distribution network, and obtaining multiple sets of image data, wherein the multiple parts of the distribution network include cables, transformers, distribution boxes, and power metering devices;

[0011] (2) transmitting the multiple sets of image data to a ground monitoring center in real time via a wireless network, and then performing denoising and image enhancement preprocessing on the multiple sets of image data to obtain the preprocessed multiple sets of image data;

[0012] (3) extracting features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature types;

[0013] (4) construct monitoring model 1 through deep learning algorithm, and construct monitoring model 2 through support vector machine algorithm;

[0014] (5) The complexity of feature 1, feature 2, ..., feature N is obtained through the feature complexity index formula;

[0015] (6) Classify feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes;

[0016] (7) Inputting the features belonging to the complex feature class into the monitoring model 1, outputting multiple groups of distribution network operation status 1, and inputting the features belonging to the simple feature class into the monitoring model 2, outputting multiple groups of distribution network operation status 2;

[0017] (8) By using the fault state prediction index formula and combining the results of distribution network operation state 1 and distribution network operation state 2, it is determined whether the distribution network operation state is normal.

[0018] Preferably, the step (3) is specifically divided into the following steps:

[0019] (3.1) performing image segmentation on the preprocessed multiple sets of image data, extracting each component in the distribution network, and obtaining the segmented area of ​​each component;

[0020] (3.2) Extract the geometric features, color features, texture features and structural features of the segmented area respectively;

[0021] (3.3) Based on the extracted geometric features, color features, texture features and structural features, multiple feature vectors are generated to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers.

[0022] Preferably, the step (4) is specifically divided into the following steps:

[0023] (4.1) obtaining historical information of the distribution network, wherein the historical information of the distribution network includes historical photo information of multiple parts of the distribution network and historical operation status information of the distribution network, and the historical photo information of multiple parts of the distribution network and the historical operation status information of the distribution network are matched one by one;

[0024] (4.2) Preprocessing the historical information of the distribution network;

[0025] (4.3) Based on the preprocessed distribution network historical information, divide the training set and the test set;

[0026] (4.4) Inputting the training set into the deep learning algorithm model and the support vector machine algorithm model for training respectively, and obtaining preliminary monitoring model 1 and preliminary monitoring model 2 respectively;

[0027] (4.5) The test set is input into preliminary monitoring model 1 and preliminary monitoring model 2 for testing respectively to obtain monitoring model 1 and monitoring model 2.

[0028] Preferably, the step (5) is specifically divided into the following steps:

[0029] (5.1) Obtain the image texture complexity, geometric shape complexity, and color change complexity of feature 1, feature 2, ..., feature N respectively, and perform standardization processing;

[0030] (5.2) Obtain the weight factors of the texture complexity, geometric shape complexity, and color change complexity of the standardized image to the feature complexity through the objective weighting method;

[0031] (5.3) The complexity of feature 1, feature 2, ..., feature N is obtained respectively through the feature complexity index formula.

[0032] Preferably, the feature complexity index formula is:

[0033]

[0034] Where FZD is the feature complexity index, WL is the image texture complexity, α1 is the weight factor of WL to FZD, XZ is the geometry complexity, α2 is the weight factor of XZ to FZD, YS is the color change complexity, α3 is the weight factor of YS to FZD, and e is a natural constant.

[0035] Preferably, the step (6) is specifically divided into the following steps:

[0036] (6.1) Compare the complexity values ​​of feature 1, feature 2, ..., feature N with a set threshold 1, and classify the features according to the comparison results;

[0037] (6.2) The features whose complexity value is greater than the set threshold value are classified as complex features;

[0038] (6.3) Features whose feature complexity values ​​are less than the set threshold value are classified as simple features.

[0039] Preferably, the step (8) is specifically divided into the following steps:

[0040] (8.1) respectively obtaining health status assessment indexes, abnormal behavior prediction indexes and reliability indexes of multiple distribution network operation states 1 and multiple distribution network operation states 2, and performing standardization processing;

[0041] (8.2) Obtain the weight factors of the standardized health status assessment index, abnormal behavior prediction index and reliability index for the operating status through the objective weighting method;

[0042] (8.3) Obtaining the fault state prediction indexes of the plurality of distribution network operation states 1 and the plurality of distribution network operation states 2 respectively through the fault state prediction index formula;

[0043] Preferably, the fault state prediction index formula is:

[0044] YXZT=sinh -1 (JK*β1-YC*β2+KX*β3)

[0045] Where YXZT is the fault state prediction index, JK is the health state assessment index, β1 is the weight factor of JK to YXZT, YC is the abnormal behavior prediction index, β2 is the weight factor of YC to YXZT, KX is the color change complexity, and β3 is the weight factor of KX to YXZT.

[0046] Preferably, if there is a group of fault state prediction indexes in the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 that is lower than the set threshold value two, it means that the distribution network cannot operate normally; if the fault state prediction indexes in the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 are not lower than the set threshold value two, it means that the distribution network can operate normally.

[0047] The present invention also provides a distribution network operation status monitoring system based on visual images, characterized in that it includes a data acquisition module, a data transmission and preprocessing module, a feature extraction module, a feature classification module, and an operation status monitoring module;

[0048] The data acquisition module collects multiple sets of image data of the power distribution network through the drone;

[0049] The data transmission and preprocessing module transmits the collected image data to the ground monitoring center in real time through the wireless network and performs preprocessing;

[0050] The feature extraction module extracts features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers;

[0051] The feature classification module constructs monitoring model 1 through a deep learning algorithm, constructs monitoring model 2 through a support vector machine algorithm, obtains the complexity of feature 1, feature 2, ..., feature N through a feature complexity index formula, and classifies feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes;

[0052] The operation status monitoring module inputs the features belonging to the complex feature class into the monitoring model 1, outputs multiple groups of distribution network operation status 1, inputs the features belonging to the simple feature class into the monitoring model 2, outputs multiple groups of distribution network operation status 2, and uses the fault state prediction index formula to combine the results of the distribution network operation status 1 and the distribution network operation status 2 to determine whether the distribution network operation status is normal.

[0053] The beneficial effects of the present invention are:

[0054] 1. By acquiring distribution network image data and using the feature complexity formula according to the image data to derive the corresponding feature complexity, the complex feature class and the simple feature class are divided and the appropriate monitoring model is selected, thereby achieving rapid response to the distribution network image data, effectively solving the problem in the prior art that the monitoring model cannot meet the demand for rapid response when a distribution network failure occurs.

[0055] 2. Through feature extraction and classification methods, combined with the standardization and weighted processing of image texture complexity, geometric shape complexity and color change complexity, accurate complexity assessment of each component of the distribution network is achieved, thereby improving the accuracy and real-time performance of monitoring, and effectively avoiding misjudgment caused by overly simple or overly complex features in traditional image processing methods.

[0056] 3. By combining the fault status prediction index formula with the health status assessment index, abnormal behavior prediction index and reliability index for comprehensive evaluation, the accurate operation status judgment of the distribution network is achieved, which effectively solves the problem that the operation status judgment of the distribution network in the existing technology is not accurate enough and potential abnormalities cannot be discovered in real time. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 is a flow chart of the method of the present invention;

[0058] Figure 2 It is a schematic diagram of the system structure of the present invention; DETAILED DESCRIPTION

[0059] The embodiments of the present invention are further described below in conjunction with the accompanying drawings:

[0060] Example 1

[0061] like Figure 1 As shown, the present invention provides a method for monitoring the operation status of a distribution network based on visual images, comprising the following steps:

[0062] (1) photographing the distribution network through a drone, collecting multiple sets of photos of multiple parts of the distribution network, and obtaining multiple sets of image data, wherein the multiple parts of the distribution network include cables, transformers, distribution boxes, and power metering devices;

[0063] Step (1) By using drones to photograph the distribution network, a large distribution network area can be quickly covered, avoiding the time and manpower required for manual inspections. At the same time, drones can perform photography tasks at high altitudes, avoiding manual entry into dangerous areas (such as high-voltage towers, substations, etc.), reducing safety risks for workers;

[0064] The distribution network consists of multiple parts (cables, transformers, distribution boxes, power metering devices), each of which may have different types of faults and problems. By taking special photos of different parts, each part can be checked in detail to avoid missing possible fault points.

[0065] (2) transmitting the multiple sets of image data to a ground monitoring center in real time via a wireless network, and then performing denoising and image enhancement preprocessing on the multiple sets of image data to obtain the preprocessed multiple sets of image data;

[0066] Step (2) smoothing the multiple sets of image data by using a Gaussian function to reduce random noise in the multiple sets of image data, the steps are as follows: calculating a weighted average of each pixel in the multiple sets of image data and surrounding pixels, using a Gaussian function as a weight distribution, and smoothing the noisy pixels in the multiple sets of image data, which helps to remove stray noise in the multiple sets of image data, enhance the visibility of device details, and improve the accuracy of subsequent algorithms (such as feature extraction and state assessment);

[0067] The distribution of grayscale values ​​of multiple groups of image data is adjusted through a histogram equalization algorithm to enhance the contrast of multiple groups of image data. The steps are as follows: first, the grayscale histograms in the multiple groups of image data are calculated, and then the brightness of the images in the multiple groups of image data is adjusted according to the distribution of grayscale values, so that the grayscale range of the image is more uniform; it helps to enhance the details of the image and make important components in the distribution network (such as cables, transformers, distribution boxes, etc.) clearer, which is convenient for subsequent feature extraction and status evaluation, thereby improving the monitoring system's ability to identify and diagnose equipment.

[0068] (3) extracting features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature types;

[0069] Step (3) is specifically divided into the following steps:

[0070] (3.1) performing image segmentation on the preprocessed multiple sets of image data, extracting each component in the distribution network, and obtaining the segmented area of ​​each component;

[0071] (3.2) Extract the geometric features, color features, texture features and structural features of the segmented area respectively;

[0072] (3.3) Based on the extracted geometric features, color features, texture features and structural features, multiple feature vectors are generated to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers.

[0073] Step (3) uses a convolutional neural network image segmentation method to segment the distribution network components (cables, transformers, distribution boxes, power metering devices, etc.) of each image in the multiple sets of image data, including the following steps: using an edge detection algorithm to extract edge information of each image in the multiple sets of image data; then, converting the edge information into a binary image through thresholding processing to obtain a clear object boundary; finally, using image contour analysis and region growing and other techniques to further extract the segmented area, respectively identify the components such as cables, transformers, distribution boxes, power metering devices, etc., and complete the object segmentation, thereby improving the recognition accuracy;

[0074] Extracting the geometric features of the segmented area includes the following steps: extracting the geometric features of the object using the shape descriptor, the geometric features including roundness, convexity and aspect ratio, which are helpful to distinguish objects of different shapes and can clearly identify the components of the distribution network with specific geometric shapes, such as the rectangular shape of the transformer and the linear shape of the cable;

[0075] Extracting the color features of the segmented region includes the following steps: converting each image in the multiple sets of image data into a different color space (such as RGB, HSV or Lab), calculating the color mean, standard deviation, color difference and other statistical features of each color space, and describing the color distribution in the region by using color moments, which helps to distinguish objects with different appearance colors, such as special color markings of distribution boxes, or color differences of transformers and power metering devices;

[0076] Extracting texture features of the segmented region includes the following steps: by calculating the gray level co-occurrence matrix, describing the spatial relationship between the gray levels in each image in multiple sets of image data, extracting texture features such as contrast, homogeneity, and entropy, using local binary patterns (LBP) to extract local texture features of each image, using wavelet transform to analyze texture information of different frequencies, extracting texture features of different scales, and being able to effectively identify objects with complex surface structures, such as the metal surface texture of transformers, the insulation material of cables, etc.;

[0077] Extracting the structural features of the segmented area includes the following steps: using an edge detection algorithm to obtain the edge information of the object, further processing the structure of the object through morphological operations, extracting shape structural features, and then extracting the key structural features of each image in multiple groups of image data through a corner detection method. The key structural features include edge density, corner distribution, etc., which help to identify the key structural parts of the object, such as the interface location of the power metering device, the connection point of the transformer, etc.

[0078] (4) construct monitoring model 1 through deep learning algorithm, and construct monitoring model 2 through support vector machine algorithm;

[0079] Step (4) is specifically divided into the following steps:

[0080] (4.1) obtaining historical information of the distribution network, wherein the historical information of the distribution network includes historical photo information of multiple parts of the distribution network and historical operation status information of the distribution network, and the historical photo information of multiple parts of the distribution network and the historical operation status information of the distribution network are matched one by one;

[0081] (4.2) Preprocessing the historical information of the distribution network;

[0082] (4.3) Based on the preprocessed distribution network historical information, divide the training set and the test set;

[0083] (4.4) Inputting the training set into the deep learning algorithm model and the support vector machine algorithm model for training respectively, and obtaining preliminary monitoring model 1 and preliminary monitoring model 2 respectively;

[0084] (4.5) Input the test set into preliminary monitoring model 1 and preliminary monitoring model 2 for testing, and obtain monitoring model 1 and monitoring model 2.

[0085] The deep learning algorithm of step (4) can handle a large number of complex nonlinear relationships and high-dimensional data, and is suitable for feature extraction and pattern recognition of image data. It can automatically extract hierarchical features from distribution network image data without the need for manual feature extraction, and has higher expressiveness on large-scale data sets, but requires a long training time and a large amount of data. The support vector machine is a classic supervised learning algorithm that is good at classifying small samples and high-dimensional data and is suitable for classification tasks of known categories. The support vector machine is quickly trained on a smaller data set and obtains better performance, especially when there is not enough labeled data. The complex features are monitored through monitoring model 1, and the simple features are monitored through monitoring model 2, and are reasonably allocated to effectively improve the response speed of the monitoring model, so that when a failure occurs in the distribution network, an alarm can be issued in a timely manner.

[0086] (5) The complexity of feature 1, feature 2, ..., feature N is obtained through the feature complexity index formula;

[0087] Step (5) is specifically divided into the following steps:

[0088] (5.1) Obtain the image texture complexity, geometric shape complexity, and color change complexity of feature 1, feature 2, ..., feature N respectively, and perform standardization processing;

[0089] (5.2) Obtain the weight factors of the texture complexity, geometric shape complexity, and color change complexity of the standardized image to the feature complexity through the objective weighting method;

[0090] (5.3) The complexity of feature 1, feature 2, ..., feature N is obtained respectively through the feature complexity index formula.

[0091] Step (5) of obtaining the image texture complexity is as follows: extracting the texture features of the image by using the gray level co-occurrence matrix algorithm, calculating the texture features such as contrast, entropy, and energy, and then quantifying the degree of texture change to obtain the texture complexity;

[0092] The steps to obtain the geometric shape complexity of an image are as follows: perform edge detection on the image to obtain the boundaries of the objects in the image, extract the geometric features of the objects (such as boundaries, area, perimeter, aspect ratio, etc.), and calculate the shape complexity based on the geometric features (contour complexity, angle, curvature, etc.);

[0093] The steps to obtain the complexity of image color changes are as follows: convert the images in multiple sets of image data from RGB space to a suitable color space (such as HSV or Lab), calculate the color histogram, analyze the color distribution, calculate the mean, standard deviation and other statistics of the color channel, and thus obtain the complexity of image color changes.

[0094] The feature complexity index formula in step (5) is:

[0095]

[0096] Where FZD is the feature complexity index, WL is the image texture complexity, α1 is the weight factor of WL to FZD, XZ is the geometry complexity, α2 is the weight factor of XZ to FZD, YS is the color change complexity, α3 is the weight factor of YS to FZD, and e is a natural constant.

[0097] When WL = 0.8, α1 = 0.4, XZ = 0.8, α2 = 0.3, YS = 0.6, α3 = 0.3, it is found that FZD = 0.555. The higher the feature complexity index, the more complex the feature, and vice versa.

[0098] (6) Classify feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes;

[0099] Step (6) is specifically divided into the following steps:

[0100] (6.1) Compare the complexity values ​​of feature 1, feature 2, ..., feature N with a set threshold 1, and classify the features according to the comparison results;

[0101] (6.2) The features whose complexity value is greater than the set threshold value are classified as complex features;

[0102] (6.3) Features whose feature complexity values ​​are less than the set threshold value are classified as simple features.

[0103] Step (6) adopts different processing strategies for complex feature classes and simple feature classes, so that a more efficient monitoring model can be used when processing simple features, while complex features can be deeply analyzed through a more accurate monitoring model, thereby ensuring the analysis accuracy while improving the overall processing speed.

[0104] (7) Inputting the features belonging to the complex feature class into the monitoring model 1, outputting multiple groups of distribution network operation status 1, and inputting the features belonging to the simple feature class into the monitoring model 2, outputting multiple groups of distribution network operation status 2;

[0105] (8) By using the fault state prediction index formula and combining the results of distribution network operation state 1 and distribution network operation state 2, it is determined whether the distribution network operation state is normal.

[0106] Step (8) is specifically divided into the following steps:

[0107] (8.1) respectively obtaining health status assessment indexes, abnormal behavior prediction indexes and reliability indexes of multiple distribution network operation states 1 and multiple distribution network operation states 2, and performing standardization processing;

[0108] (8.2) Obtain the weight factors of the standardized health status assessment index, abnormal behavior prediction index and reliability index for the operating status through the objective weighting method;

[0109] (8.3) Obtaining the fault state prediction indexes of the plurality of distribution network operation states 1 and the plurality of distribution network operation states 2 respectively through the fault state prediction index formula;

[0110] If one of the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 has a set of fault state prediction indexes lower than the set threshold value 2, it means that the distribution network cannot operate normally; if the fault state prediction indexes of the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 are not lower than the set threshold value 2, it means that the distribution network can operate normally.

[0111] The health status assessment index in step (8) is obtained by collecting real-time monitoring data of each device in the distribution network, such as temperature, humidity, current, voltage, vibration, etc., as well as historical maintenance records, fault history and other data of the equipment. Extract operating characteristics from the collected real-time monitoring data, including the operating status of the equipment (such as current overload, temperature anomaly, etc.), failure rate (frequency of historical equipment failures), etc. Standardize data from different sources and units so that each data has a unified dimension. According to the health impact of each device, assign weights to each feature (through objective weighting method or weighting based on expert experience), and sum all features weighted to obtain the health status index;

[0112] The abnormal behavior prediction index is obtained by collecting historical data on equipment operation, including current, voltage fluctuations, changes in the state of equipment switches, etc., and identifying past abnormal behavior patterns (such as sudden voltage drops, intermittent failures during equipment operation, etc.). Extract abnormal behavior features, such as the frequency of equipment state changes, load fluctuation rate, abnormal fluctuations in current and voltage, etc. Use time series analysis (such as autoregressive moving average model) to model historical data and predict the probability of abnormal behavior of equipment in the future. Output the abnormal behavior prediction index based on the model;

[0113] The reliability index is obtained by collecting the historical fault data of each device in the distribution network, including the number of faults, fault types, time intervals between faults, and maintenance time. The failure rate of each device is calculated based on the historical fault data records. The average repair time of all faults is calculated by analyzing the fault repair history records. The weights of the failure rate and the average repair time are obtained through the objective weighting method. The reliability index is obtained by weighted summation of the failure rate and the average repair time.

[0114] The fault state prediction index formula in step (8) is:

[0115] YXZT=sinh -1 (JK*β1-YC*β2+KX*β3)

[0116] Where, YXZT is the fault state prediction index, JK is the health state assessment index, β1 is the weight factor of JK to YXZT, YC is the abnormal behavior prediction index, β2 is the weight factor of YC to YXZT, KX is the color change complexity, and β3 is the weight factor of KX to YXZT;

[0117] When JK = 0.9, β1 = 0.5, YC = 0.8, β2 = 0.2, KX = 0.8, β3 = 0.3, it is obtained that YXZT = 0.507. The higher the fault state prediction index, the better the current operation state of the distribution network. Conversely, if the fault state prediction index is lower, it means that the current operation state of the distribution network is worse.

[0118] like Figure 2 As shown, the present invention also provides a distribution network operation status monitoring system based on visual images, characterized in that it includes a data acquisition module, a data transmission and preprocessing module, a feature extraction module, a feature classification module, and an operation status monitoring module;

[0119] The data acquisition module collects multiple sets of image data of the power distribution network through the drone, which is used as a data acquisition tool to obtain high-resolution image data through aerial photography equipment;

[0120] It needs to be explained that because drones can cover a large area and are suitable for complex environments, especially difficult-to-reach distribution network facilities, and their image acquisition speed is fast, they can obtain a large amount of data in a short period of time, thereby achieving efficient and automated collection, reducing human intervention, and adapting to various environmental conditions, such as high-altitude areas away from human operations, thereby improving monitoring efficiency.

[0121] The data transmission and preprocessing module transmits the collected image data to the ground monitoring center in real time through the wireless network and performs preprocessing;

[0122] It needs to be explained that through wireless networks, data can be transmitted quickly to monitor the status of the distribution network in real time.

[0123] The feature extraction module extracts features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers;

[0124] Among them, automated feature extraction is used to reduce manual intervention, and key information is efficiently extracted through automated methods, thereby improving the accuracy of the monitoring system.

[0125] The feature classification module constructs monitoring model 1 through a deep learning algorithm, constructs monitoring model 2 through a support vector machine algorithm, obtains the complexity of feature 1, feature 2, ..., feature N through a feature complexity index formula, and classifies feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes;

[0126] By classifying feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes, the features of the simple feature class are monitored and predicted through monitoring model 2, and the features of the complex feature class are monitored and predicted through monitoring model 1, thereby avoiding the phenomenon of simple features being predicted by complex models and effectively improving the efficiency of monitoring and prediction.

[0127] The operation status monitoring module inputs the features belonging to the complex feature class into the monitoring model 1, outputs multiple groups of distribution network operation status 1, inputs the features belonging to the simple feature class into the monitoring model 2, outputs multiple groups of distribution network operation status 2, and judges whether the distribution network operation status is normal by combining the results of the distribution network operation status 1 and the distribution network operation status 2 through the fault state prediction index formula;

[0128] If both the distribution network operation status 1 and the distribution network operation status 2 are normal, the distribution network is recorded as normal. If the distribution network operation status 1 is normal and the distribution network operation status 2 is abnormal, the distribution network is recorded as abnormal and the alarm system is activated. If the distribution network operation status 1 is abnormal and the distribution network operation status 2 is normal, or if the distribution network operation status 1 and the distribution network operation status 2 are both abnormal, the distribution network is recorded as abnormal and the alarm system is activated, and the alarm information is automatically sent to the monitoring center to notify the staff. The staff further investigates the problem according to the fault type of the distribution network.

[0129] A comprehensive judgment is made by combining the results of the two models. If any model detects an abnormal state, the alarm mechanism is activated to automatically notify the staff, thereby effectively reducing the false alarm rate and improving the response speed. Through multi-dimensional evaluation, it can provide the staff with a basis for taking different emergency measures according to different fault types.

[0130] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0131] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0132] These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0133] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] Although the preferred embodiments of the present invention have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention.

[0135] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.

Claims

1. A distribution network operation status monitoring method based on visual images, characterized in that: The following steps are involved: (1) photographing the distribution network through a drone, collecting multiple sets of photos of multiple parts of the distribution network, and obtaining multiple sets of image data, wherein the multiple parts of the distribution network include cables, transformers, distribution boxes, and power metering devices; (2) transmitting the multiple sets of image data to a ground monitoring center in real time via a wireless network, and then performing denoising and image enhancement preprocessing on the multiple sets of image data to obtain the preprocessed multiple sets of image data; (3) extracting features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature types; (4) construct monitoring model 1 through deep learning algorithm, and construct monitoring model 2 through support vector machine algorithm; (5) The complexity of feature 1, feature 2, ..., feature N is obtained through the feature complexity index formula; (6) Classify feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes; (7) Inputting the features belonging to the complex feature class into the monitoring model 1, outputting multiple groups of distribution network operation status 1, and inputting the features belonging to the simple feature class into the monitoring model 2, outputting multiple groups of distribution network operation status 2; (8) By using the fault state prediction index formula and combining the results of distribution network operation state 1 and distribution network operation state 2, it is determined whether the distribution network operation state is normal.

2. According to claim 1, a method for monitoring the operation status of a distribution network based on visual images is characterized in that: The step (3) is specifically divided into the following steps: (3.1) performing image segmentation on the preprocessed multiple sets of image data, extracting each component in the distribution network, and obtaining the segmented area of ​​each component; (3.2) Extract the geometric features, color features, texture features and structural features of the segmented area respectively; (3.3) Based on the extracted geometric features, color features, texture features and structural features, multiple feature vectors are generated to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers.

3. According to the method for monitoring the operation status of a distribution network based on visual images according to claim 1, it is characterized in that: The step (4) is specifically divided into the following steps: (4.1) obtaining historical information of the distribution network, wherein the historical information of the distribution network includes historical photo information of multiple parts of the distribution network and historical operation status information of the distribution network, and the historical photo information of multiple parts of the distribution network and the historical operation status information of the distribution network are matched one by one; (4.2) Preprocessing the historical information of the distribution network; (4.3) Based on the preprocessed distribution network historical information, divide the training set and the test set; (4.4) Inputting the training set into the deep learning algorithm model and the support vector machine algorithm model for training respectively, and obtaining preliminary monitoring model 1 and preliminary monitoring model 2 respectively; (4.5) The test set is input into preliminary monitoring model 1 and preliminary monitoring model 2 for testing respectively to obtain monitoring model 1 and monitoring model 2.

4. According to claim 1, a method for monitoring the operation status of a distribution network based on visual images is characterized in that: The step (5) is specifically divided into the following steps: (5.1) Obtain the image texture complexity, geometric shape complexity, and color change complexity of feature 1, feature 2, ..., feature N respectively, and perform standardization processing; (5.2) Obtain the weight factors of the texture complexity, geometric shape complexity, and color change complexity of the standardized image to the feature complexity through the objective weighting method; (5.3) The complexity of feature 1, feature 2, ..., feature N is obtained respectively through the feature complexity index formula.

5. A method for monitoring the operation status of a distribution network based on visual images according to claim 4, characterized in that: The feature complexity index formula is: Where FZD is the feature complexity index, WL is the image texture complexity, α1 is the weight factor of WL to FZD, XZ is the geometry complexity, α2 is the weight factor of XZ to FZD, YS is the color change complexity, α3 is the weight factor of YS to FZD, and e is a natural constant.

6. The method for monitoring the operation status of a distribution network based on visual images according to claim 1, characterized in that: The step (6) is specifically divided into the following steps: (6.1) Compare the complexity values ​​of feature 1, feature 2, ..., feature N with a set threshold 1, and classify the features according to the comparison results; (6.2) The features whose complexity value is greater than the set threshold value are classified as complex features; (6.3) Features whose feature complexity values ​​are less than the set threshold value are classified as simple features.

7. The method for monitoring the operation status of a distribution network based on visual images according to claim 1, characterized in that: The step (8) is specifically divided into the following steps: (8.1) respectively obtaining health status assessment indexes, abnormal behavior prediction indexes and reliability indexes of multiple distribution network operation states 1 and multiple distribution network operation states 2, and performing standardization processing; (8.2) Obtain the weight factors of the standardized health status assessment index, abnormal behavior prediction index and reliability index for the operating status through the objective weighting method; (8.3) The fault state prediction index formula is used to obtain the fault state prediction index of multiple distribution network operation states 1 and multiple distribution network operation states 2 respectively.

8. A method for monitoring the operation status of a distribution network based on visual images according to claim 7, characterized in that: The fault state prediction index formula is: YXZT=sinh -1 (JK*β1-YC*β2+KX*β3) Where YXZT is the fault state prediction index, JK is the health state assessment index, β1 is the weight factor of JK to YXZT, YC is the abnormal behavior prediction index, β2 is the weight factor of YC to YXZT, KX is the color change complexity, and β3 is the weight factor of KX to YXZT.

9. A method for monitoring the operation status of a distribution network based on visual images according to claim 7, characterized in that: If one of the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 has a set of fault state prediction indexes lower than the set threshold value two, it means that the distribution network cannot operate normally; if the fault state prediction indexes of the multiple groups of distribution network operating states 1 and the multiple groups of distribution network operating states 2 are not lower than the set threshold value two, it means that the distribution network can operate normally.

10. A distribution network operation status monitoring system based on visual images, characterized in that: It includes data acquisition module, data transmission and preprocessing module, feature extraction module, feature classification module and operation status monitoring module; The data acquisition module collects multiple sets of image data of the power distribution network through the drone; The data transmission and preprocessing module transmits the collected image data to the ground monitoring center in real time through the wireless network and performs preprocessing; The feature extraction module extracts features from the preprocessed image data to obtain feature 1, feature 2, ..., feature N, where N is the total number of feature type numbers; The feature classification module constructs monitoring model 1 through a deep learning algorithm, constructs monitoring model 2 through a support vector machine algorithm, obtains the complexity of feature 1, feature 2, ..., feature N through a feature complexity index formula, and classifies feature 1, feature 2, ..., feature N into complex feature classes and simple feature classes; The operation status monitoring module inputs the features belonging to the complex feature class into the monitoring model 1, outputs multiple groups of distribution network operation status 1, inputs the features belonging to the simple feature class into the monitoring model 2, outputs multiple groups of distribution network operation status 2, and uses the fault state prediction index formula to combine the results of the distribution network operation status 1 and the distribution network operation status 2 to determine whether the distribution network operation status is normal.

Images