Power transmission line insulator pollution degree detection system of hyperspectral camera

Abstract

The invention provides a power transmission line insulator pollution degree detection system based on a hyperspectral camera, and relates to the technical field of power transmission line detection, and the system comprises a server, the server carries a deep learning model, a data checking module, a chart generation module basic database and a historical database, and the server is in data connection with a monitoring center. The monitoring center is in signal connection with an unmanned aerial vehicle and an electric field detection device. The bird activity and environmental data of the target environment are sampled and investigated in advance to obtain the reference data, the deep learning model capable of simulating and predicting the pollution degree of the insulator is established according to the reference data, and the unmanned aerial vehicle detection task is formulated and adjusted according to the deep learning model. The purposiveness of the unmanned aerial vehicle detection task and the accuracy of the detection result are greatly improved, and the probability that the unmanned aerial vehicle detection result is one-sidedness due to the fact that the insulator pollution degree change rule is not met is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of power transmission line detection, and in particular to a transmission line insulator contamination degree detection system using a hyperspectral camera. Background Art

[0002] Insulators play a dual role of electrical insulation and mechanical fixation in power systems. Once dirt accumulates on their surfaces, it may cause flashover, the so-called pollution flashover. Pollution flashover is one of the important causes of power system failures. Therefore, monitoring the pollution degree of insulators is crucial to preventing such accidents.

[0003] In general, the dirt on the surface of insulators is mainly dust particles in the air, but in areas where birds are active, insulators are very prone to bird droppings pollution. Bird activities are closely related to environmental changes. This situation has caused great trouble for insulator detection tasks. Although the method of combining non-contact optical imaging based on image processing with drones can improve detection efficiency, the formulation of drone detection tasks still cannot achieve efficient and timely results.

[0004] In the prior art, the formulation of insulator inspection tasks is generally based on previous inspection results. For areas where birds gather, it is necessary to formulate them in combination with the laws of bird activity and environmental changes. In this case, the amount of data that needs to be processed is very large, and the correlation between the various data further increases the difficulty of data processing. It is difficult to formulate drone inspection tasks based on the specific environmental data of the target area, which makes the results obtained from drone inspection tasks have great one-sidedness and defects. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art. First, bird activities and environmental changes are analyzed, and the UAV detection task of the insulator contamination degree is formulated and adjusted according to the analysis results, so that the UAV detection of the insulator contamination degree is more efficient and reliable.

[0006] In order to achieve the above-mentioned purpose, the present invention adopts the following technical scheme: a transmission line insulator contamination degree detection system of a hyperspectral camera, comprising a server, the server is equipped with a deep learning model, a data verification module, a basic database of a chart generation module and a historical database, the server data is connected to a monitoring center, and the monitoring center signal is connected to a drone and an electric field detection device;

[0007] The basic database includes bird data, hyperspectral data and electric field data. The bird data includes bird population data, bird food data and bird droppings data. The hyperspectral data includes insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and insulator pollution sample hyperspectral data. The electric field data includes insulator sample electric field data and insulator pollution sample electric field data.

[0008] The historical database includes environmental data, bird change data, insulator hyperspectral detection historical data and insulator electric field detection historical data; the environmental data includes rainfall historical data, air humidity historical data, wind power historical data, temperature historical data and air pollution historical index; the bird change data includes bird activity data, bird food change data and bird droppings change data;

[0009] The UAV includes a UAV flight control system, a hyperspectral gimbal camera and a high-speed camera;

[0010] The monitoring center includes a data entry system, a display system and a data transmission system.

[0011] As a preferred embodiment, a transmission line insulator contamination degree detection system constructed by a server, a monitoring center and a drone is also included. The construction of the transmission line insulator contamination degree detection system specifically includes the following steps:

[0012] S1. First, establish the basic database;

[0013] S2, then formulate the drone inspection task;

[0014] S3, build a historical database based on the data obtained from the drone inspection mission;

[0015] S4. Analyze the impact of environmental data and bird change data on the change of insulator pollution degree, and adjust the weight of each data according to the historical database;

[0016] S5. According to the weights of environmental data and bird change data, the grey correlation method is used to analyze the correlation between the environmental data and the bird change data;

[0017] S6. According to the specific correlation between the data and the weight of the data in the historical database, a deep learning model is constructed, and the deep learning model is used to analyze and predict the change of the degree of contamination of the insulator;

[0018] S7, adjust the drone detection task according to the change of pollution level simulated by the deep learning model;

[0019] S8. Use the icon generation module to generate a line graph of the data detected by the drone and the electric field detection device, and display it in the display system.

[0020] As a preferred implementation, in step S1, the construction of the basic database specifically includes the following steps:

[0021] S1.1. First, investigate the birds in the area where the transmission tower is located and obtain the corresponding bird droppings samples, as well as insulator samples and air pollutant samples in the target area. Then, simulate the target area environment to produce multiple insulator pollution samples with different pollution levels.

[0022] S1.2, then photographing the insulator samples, bird droppings samples, air pollutant samples and insulator pollution samples through the experimental hyperspectral camera, and obtaining the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data, and then classifying the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data as hyperspectral data through the data entry system, and uploading them to the basic database in the server;

[0023] S1.3, then installing the insulator samples and the insulator pollution samples on the high-voltage power transmission simulation equipment, and then performing electric field detection on the insulator samples and the multiple groups of insulator pollution samples through the experimental electric field detection equipment to obtain the insulator sample electric field data and the multiple groups of insulator pollution sample electric field data, and then classifying the insulator sample electric field data and the insulator pollution sample electric field data as electric field data, and uploading them to the basic database in the server through the data entry system;

[0024] S1.4. Obtain bird population data in the target area through surveys, and obtain corresponding bird food data and bird droppings data based on the bird population data, and classify the obtained bird population data, bird food data and bird droppings data as bird data, and then upload them to the basic database of the server through the data entry system.

[0025] As a preferred implementation, in step S3, the construction of the historical database specifically includes the following steps:

[0026] S3.1. First, obtain data such as rainfall, air humidity, wind speed, temperature and air pollution index through the meteorological center, and arrange the same data by date over time, divide them into historical rainfall data, historical air humidity data, historical wind speed data, historical temperature data and historical air pollution index, and classify them as environmental data;

[0027] S3.2. Then, when performing the insulator inspection task, the high-speed camera of the drone is used to investigate the bird activities and plant distribution in the target area, and to sample bird droppings to obtain bird activity data and bird food change data. Then, based on the bird food change data and the bird droppings samples, the bird droppings composition changes are analyzed to obtain the bird droppings change data. Then, the bird activity data, bird food change data, and bird droppings change data at different times are sorted according to the timeline and classified as bird change data.

[0028] S3.3, sorting the insulator hyperspectral detection data obtained from different UAV detection tasks according to the time axis and classifying them as insulator hyperspectral detection historical data;

[0029] S3.4. Arrange the insulator electric field detection data of different time periods according to the time axis and classify them into insulator electric field detection historical data;

[0030] S3.5. Upload the environmental data and bird change data to the historical database of the server through the data entry system, and upload the insulator hyperspectral detection historical data and insulator electric field detection historical data to the historical database of the server through the data transmission system.

[0031] As a preferred implementation, in step S4, the pollution degree is formulated based on the upper limit of the flashover occurrence of the insulator, and the pollution degree of the insulator is divided into specific levels according to the historical data of the insulator hyperspectral detection and the historical data of the insulator electric field detection, and then the rainfall, air humidity value, wind speed, wind direction, temperature value, air pollution index, number of birds, and bird food change ratio are given weights. The specific formula is:

[0032] PI=W1×R+W2×H+W3×S+W4×Y+W5×T+W6×M+W7×N+W8×B

[0033] Among them, let PI be the degree of pollution, R be the rainfall, H be the air humidity value, S be the wind speed, Y be the wind direction, T be the temperature value, M be the air pollution index, N be the number of birds, B be the bird food change ratio, W1 be the weight value of rainfall, W2 be the weight value of air humidity value, W3 be the weight value of wind speed, W4 be the weight value of wind direction, W5 be the weight value of temperature value, W6 be the weight value of air pollution index, W7 be the weight value of the number of birds, W8 be the weight value of the bird food change ratio;

[0034] According to different levels of pollution, the specific weights of rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population, and bird food change ratio are calculated using the historical database.

[0035] As a preferred implementation, in step S5, according to the weights of the environmental data and the bird change data in step S4, the grey correlation method is used to analyze the correlation between the environmental data and the bird change data, specifically:

[0036] S5.1. The pollution level PI set in step S4 is used as the parent sequence, which is recorded as:

[0037] PI={PI(k)k=1,2,.....,n}

[0038] S5.2, the rainfall R, air humidity H, wind speed S, wind direction Y, temperature T, air pollution index M, bird number N, and bird food change ratio B set in step S4 are taken as subsequences and recorded as:

[0039] x i ={x i kk=1,2,......,n}

[0040] Where i is the algebraic symbol of R, H, S, Y, T, M, N, B;

[0041] S5.3. Then divide each value in the rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the first value of the corresponding item, and replace the original corresponding value with the obtained value. The formula is:

[0042]

[0043] S5.4. Further narrow the range of variables and simplify the calculation. Divide each value of rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the mean of the corresponding item, and replace the original corresponding value with the obtained value. The formula is:

[0044]

[0045] S5.5, then calculate the absolute difference between the pollution degree and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population, and bird food change ratio, i.e. Δ i (k) = PI(k) - x i (k), obtain the difference matrix;

[0046] S5.6. Find the minimum difference between the two levels and the maximum difference between the two levels according to the difference matrix, represent them by a and b respectively, and then calculate the correlation coefficient:

[0047]

[0048] Among them, ρ is the resolution coefficient;

[0049] S5.7. For each subsequence, calculate the average value of all its correlation coefficients to obtain the correlation between the subsequence and the parent sequence. The formula is:

[0050]

[0051] Among them, ri is the correlation between the degree of pollution and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, number of birds, and bird food change ratio.

[0052] As a preferred implementation, in step S6, the process of building the deep learning model is specifically as follows:

[0053] S6.1. Use the Keras library to build an LSTM model as the core of the deep learning model, and use the basic database as the basic data set to build a deep learning model;

[0054] S6.2, use the historical database as a training set, and combine the weights of each data in S4 {W1, W2, W3, W4, W5, W6, W7, W8} and the specific correlation between each data in S5 r i , train the deep learning model;

[0055] S6.3. Use a deep learning model to simulate the change of pollution level over time and obtain simulated pollution level simulation data;

[0056] S6.4. Use the data verification module to verify the pollution degree simulation data at the corresponding time point with the newly detected data of the drone and the newly detected electric field data of the electric field detection equipment, and optimize the deep learning model according to the verification results;

[0057] S6.5. Use the optimized deep learning model to adjust the UAV inspection task, and continuously optimize the deep learning model with the newly added inspection data. Continuously adjust the UAV inspection task according to the continuously optimized deep learning model, and repeat the cycle until the difference between the insulator contamination degree detected by the UAV and the insulator contamination degree simulated by the deep learning model in the same time period is within the threshold.

[0058] As a preferred implementation, in step S8, the horizontal axis of the line graph generated by the chart generation module is time, and the vertical axis of the line graph is pollution degree. A data set is established on the pollution degree coordinate corresponding to each detection time point, and the items included in the data set are: pollution degree, electric field distribution data, air humidity data within the detection period, rainfall data within the detection period, wind speed data within the detection period, temperature data within the detection period, air pollution data within the detection period, insulator hyperspectral detection data, bird activity change data within the detection period, bird food change data within the detection period, and bird droppings change data within the detection period.

[0059] As a preferred implementation, the data set includes: air humidity data within a detection period, rainfall data within a detection period, wind data within a detection period, temperature data within a detection period, air pollution data within a detection period, insulator hyperspectral detection data, bird activity change data within a detection period, bird food change data within a detection period, and bird droppings change data within a detection period. The detection period is the time between the current detection time point and the previous detection time point. According to the different time intervals of the drone detection tasks, the length of the detection period varies with the time interval of the drone detection tasks.

[0060] Compared with the prior art, the advantages and positive effects of the present invention are:

[0061] 1. The present invention obtains benchmark data by sampling and investigating bird activities and environmental data of the target environment in advance, establishes a deep learning model that can simulate and predict the contamination degree of insulators based on the benchmark data, formulates and adjusts UAV inspection tasks based on the deep learning model, greatly improves the purposefulness of UAV inspection tasks and the accuracy of inspection results, and reduces the probability of one-sided UAV inspection results caused by failure to conform to the law of changes in insulator contamination degrees.

[0062] 2. The present invention first establishes a basic database through testing, sampling and investigation, and obtains various data in the historical database through the UAV inspection tasks formulated in advance, and then uses the basic database and the historical database to build a deep learning model, and uses the deep learning model to optimize the UAV inspection tasks. The deep learning model is reversely optimized according to the new inspection data obtained by the UAV, and the UAV inspection tasks are adjusted according to the optimized deep learning model to form a virtuous circle, so that the entire UAV inspection task is more in line with the environment of the target area, and the detection of insulator contamination is more efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 The module diagram of a transmission line insulator contamination degree detection system using a hyperspectral camera is proposed for the present invention;

[0064] Figure 2 A schematic diagram of the historical database structure of a transmission line insulator contamination degree detection system using a hyperspectral camera is provided for the present invention;

[0065] Figure 3 A schematic diagram of test items of a transmission line insulator contamination degree detection system using a hyperspectral camera is proposed for the present invention;

[0066] Figure 4A schematic diagram of the basic database structure of a transmission line insulator contamination degree detection system using a hyperspectral camera is provided for the present invention;

[0067] Figure 5 A construction flow chart of a transmission line insulator contamination degree detection system using a hyperspectral camera is proposed for the present invention;

[0068] Figure 6 A diagram showing a contamination degree detection system for transmission line insulators using a hyperspectral camera is provided for the present invention;

[0069] Figure 7 A schematic diagram of the data set composition of a transmission line insulator contamination degree detection system using a hyperspectral camera is proposed for the present invention. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0071] Example

[0072] like Figure 1 , Figure 2 and Figure 4 As shown, the present invention provides a technical solution: a transmission line insulator contamination degree detection system of a hyperspectral camera, comprising a server, wherein the server is equipped with a deep learning model, a data verification module, a basic database of a chart generation module and a historical database, the server data is connected to a monitoring center, and the monitoring center signal is connected to a drone and an electric field detection device;

[0073] The basic database includes bird data, hyperspectral data and electric field data. The bird data includes bird population data, bird food data and bird droppings data. The hyperspectral data includes insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and insulator pollution sample hyperspectral data. The electric field data includes insulator sample electric field data and insulator pollution sample electric field data.

[0074] The historical database includes environmental data, bird change data, insulator hyperspectral detection historical data and insulator electric field detection historical data; the environmental data includes rainfall historical data, air humidity historical data, wind power historical data, temperature historical data and air pollution historical index; the bird change data includes bird activity data, bird food change data and bird droppings change data;

[0075] The UAV includes a UAV flight control system, a hyperspectral gimbal camera and a high-speed camera;

[0076] The monitoring center includes a data entry system, a display system and a data transmission system.

[0077] In this embodiment:

[0078] The deep learning model is used to analyze and process the data in the basic database and historical database, and simulate and predict the changes in the degree of insulator contamination based on the results of the analysis and processing. The data verification module can verify the simulated and predicted changes in the degree of insulator contamination with the basic data and historical data.

[0079] The data in the basic database and the historical database are used to provide data support for the construction and training of deep learning models;

[0080] Drones are used to perform the task of detecting the degree of contamination of insulators of outdoor transmission lines, obtaining hyperspectral images of insulators and bird monitoring images in the target area;

[0081] The monitoring center is mainly used for data transmission and display of test results.

[0082] Example 2

[0083] like Figure 1-7 As shown, it also includes a transmission line insulator contamination degree detection system constructed by a server, a monitoring center and a drone. The construction of the transmission line insulator contamination degree detection system specifically includes the following steps:

[0084] S1. First, establish a basic database, which specifically includes the following steps:

[0085] S1.1. First, investigate the birds in the area where the transmission tower is located and obtain the corresponding bird droppings samples, as well as insulator samples and air pollutant samples in the target area. Then, simulate the target area environment to produce multiple insulator pollution samples with different pollution levels.

[0086] The survey of bird population and quantity can directly obtain data from biological protection agencies in the target area, and can also be carried out by dispatching drones equipped with infrared cameras, high-speed cameras and other image acquisition equipment to inspect the target area in advance, and obtain bird-related data in the target area by sampling survey, and then obtain bird droppings data that may contaminate insulators in the target area according to the habits of the corresponding birds and the method of collecting samples;

[0087] Moreover, air pollutant samples in the target area can be directly obtained through the meteorological department of the corresponding area;

[0088] Furthermore, the pollution of insulators is simulated based on the bird droppings data, bird data, air pollutant samples and meteorological data of the target area, so as to produce insulator pollution samples with different pollution degrees;

[0089] S1.2, then photographing the insulator samples, bird droppings samples, air pollutant samples and insulator pollution samples through the experimental hyperspectral camera, and obtaining the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data, and then classifying the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data as hyperspectral data through the data entry system, and uploading them to the basic database in the server;

[0090] Among them, the different manifestations of bird droppings, air pollutants and insulators under the hyperspectral spectrum are used to obtain hyperspectral images under different pollution levels as supporting data for analyzing the pollution of bird droppings and air pollutants on insulators;

[0091] S1.3, then installing the insulator samples and the insulator pollution samples on the high-voltage power transmission simulation equipment, and then performing electric field detection on the insulator samples and the multiple groups of insulator pollution samples through the experimental electric field detection equipment to obtain the insulator sample electric field data and the multiple groups of insulator pollution sample electric field data, and then classifying the insulator sample electric field data and the insulator pollution sample electric field data as electric field data, and uploading them to the basic database in the server through the data entry system;

[0092] Among them, the electric field data of insulators at different pollution levels are detected and recorded by taking advantage of the different degrees of disturbance of the electric field of the transmission line under different pollution levels, so as to provide data support for the analysis of the impact of pollution levels on the insulation effect of insulators;

[0093] S1.4. Obtain bird population data in the target area through surveys, and obtain corresponding bird food data and bird droppings data based on the bird population data, and classify the obtained bird population data, bird food data and bird droppings data as bird data, and then upload them to the basic database of the server through the data entry system;

[0094] The amount of bird droppings produced is closely related to the habits and numbers of birds. By analyzing the food data and population data of birds, the amount of bird droppings produced within a certain period of time can be determined, thereby providing data support for the degree of pollution that bird droppings may cause to insulators over a period of time.

[0095] S2, then formulate the drone inspection task;

[0096] Among them, the drone inspection mission in this stage is mainly to obtain environmental data and bird change data in the target area. The division of insulator hyperspectral detection data and insulator electric field detection data obtained multiple times facilitates the construction of insulator hyperspectral detection historical data projects and insulator electric field detection historical data projects, providing data support for subsequent data analysis and model construction.

[0097] S3. Build a historical database based on the data obtained from the drone inspection mission, which specifically includes the following steps:

[0098] S3.1. First, obtain data such as rainfall, air humidity, wind speed, temperature and air pollution index through the meteorological center, and arrange the same data by date over time, divide them into historical rainfall data, historical air humidity data, historical wind speed data, historical temperature data and historical air pollution index, and classify them as environmental data;

[0099] Environmental data not only affects the attachment of air pollutants to insulators, but also affects the activities of birds. Obtaining accurate environmental data can determine the amount of air pollutants attached to insulators and the frequency of bird droppings on insulators, providing data support for determining changes in the degree of pollution of insulators.

[0100] S3.2. Then, when performing the insulator inspection task, the high-speed camera of the drone is used to investigate the bird activities and plant distribution in the target area, and to sample bird droppings to obtain bird activity data and bird food change data. Then, based on the bird food change data and the bird droppings samples, the bird droppings composition changes are analyzed to obtain the bird droppings change data. Then, the bird activity data, bird food change data, and bird droppings change data at different times are sorted according to the timeline and classified as bird change data.

[0101] Among them, bird change data and bird droppings data are obtained by observation and sampling, and combined with the design in step S3.1, the bird change data is made more accurate, providing more accurate data support for judging the change of the pollution degree of the insulator;

[0102] S3.3, sorting the insulator hyperspectral detection data obtained from different UAV detection tasks according to the time axis and classifying them as insulator hyperspectral detection historical data;

[0103] Among them, the insulator hyperspectral detection data obtained by the drone detection mission can directly obtain the insulator contamination data, and the insulator hyperspectral detection historical data can also be used as the basis for building a deep learning model;

[0104] S3.4. Arrange the insulator electric field detection data of different time periods according to the time axis and classify them into insulator electric field detection historical data;

[0105] Among them, the insulator electric field detection data combined with the insulator hyperspectral detection data can accurately determine the degree of insulator contamination and provide a theoretical basis for the construction of a deep learning model;

[0106] S3.5. Upload the environmental data and bird change data to the historical database of the server through the data entry system, and upload the insulator hyperspectral detection historical data and insulator electric field detection historical data to the historical database of the server through the data transmission system.

[0107] S4. Analyze the impact of environmental data and bird change data on the change of insulator pollution degree, formulate the pollution degree based on the upper limit of the flashover of the insulator, divide the pollution degree of the insulator into specific levels according to the historical data of insulator hyperspectral detection and insulator electric field detection, and then assign weights to rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, number of birds, and bird food change ratio. The specific formula is:

[0108] PI=W1×R+W2×H+W3×S+W4×Y+W5×T+W6×M+W7×N+W8×B

[0109] Wherein, PI is set as pollution degree, R is rainfall, H is air humidity value, S is wind speed, Y is wind direction, T is temperature value, M is air pollution index, N is number of birds, B is bird food change ratio, W1 is weight value of rainfall, W2 is weight value of air humidity value, W3 is weight value of wind speed, W4 is weight value of wind direction, W5 is weight value of temperature value, W6 is weight value of air pollution index, W7 is weight value of number of birds, W8 is weight value of bird food change ratio, according to different levels of pollution degree, use historical database to calculate and determine specific weights of rainfall, air humidity value, wind speed, wind direction, temperature value, air pollution index, number of birds, and bird food change ratio;

[0110] Among them, due to the different bird populations and meteorological influences in the target areas under different meteorological environments, the contamination levels of insulators in different areas will also be different in the same time period. The weights of each data are adjusted according to the actual detection data at specific and multiple time points, making the weight distribution of each data more reasonable and accurate.

[0111] S5. According to the weights of the environmental data and the bird change data, the grey correlation method is used to analyze the correlation between the environmental data and the bird change data, which specifically includes the following steps:

[0112] S5.1. The pollution level PI set in step S4 is used as the parent sequence, which is recorded as:

[0113] PI={PI(k)k=1,2,.....,n}

[0114] S5.2, the rainfall R, air humidity H, wind speed S, wind direction Y, temperature T, air pollution index M, bird number N, and bird food change ratio B set in step S4 are taken as subsequences and recorded as:

[0115] x i ={x i kk=1,2,......,n}

[0116] Where i is the algebraic symbol of R, H, S, Y, T, M, N, B;

[0117] S5.3. Then divide each value in the rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the first value of the corresponding item, and replace the original corresponding value with the obtained value. The formula is:

[0118]

[0119] Among them, since each data needs to emphasize the change relative to the initial state, the initial value transformation method is used to divide each data point by the value of the first data point, so that the data series has the same starting point;

[0120] S5.4. Further narrow the range of variables and simplify the calculation. Divide each value of rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the mean of the corresponding item, and replace the original corresponding value with the obtained value. The formula is:

[0121]

[0122] Among them, due to the different units of various data, the mean transformation method is used to divide each data point by the average value of the entire data series to achieve de-dimensionalization, eliminate the impact of different units of different indicators, and make the data have a unified scale;

[0123] S5.5, then calculate the absolute difference between the pollution degree and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population, and bird food change ratio, i.e. Δ i (k) = PI(k) - x i (k), obtain the difference matrix;

[0124] Among them, by taking the absolute value of each data in the difference matrix, the data processing process can be greatly simplified and the accuracy and reliability of the analysis can be improved;

[0125] S5.6. Find the minimum difference between the two levels and the maximum difference between the two levels according to the difference matrix, represent them by a and b respectively, and then calculate the correlation coefficient:

[0126]

[0127] Among them, ρ is the resolution coefficient, which is generally taken as 0.5;

[0128] S5.7. For each subsequence, calculate the average value of all its correlation coefficients to obtain the correlation between the subsequence and the parent sequence. The formula is:

[0129]

[0130] Among them, r i That is, the correlation between pollution degree and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, number of birds, and change ratio of bird food.

[0131] S6. According to the specific correlation between the data and the weight of the data in the historical database, a deep learning model is constructed, and the deep learning model is used to analyze and predict the change of the degree of contamination of the insulator, which specifically includes the following steps:

[0132] S6.1. Use the Keras library to build an LSTM model as the core of the deep learning model, and use the basic database as the basic data set to build a deep learning model;

[0133] Among them, Keras is a high-level neural network API that can easily and quickly build and train deep learning models. LSTM, through its unique structure and gating mechanism, can capture long-term dependencies in sequence data and can specifically analyze the correlation changes between various data in the historical database.

[0134] S6.2, using the historical database as a training set, and combining the weights of the data in step S4 and the specific correlations between the data in step S5, training the deep learning model;

[0135] Among them, based on the weight relationship between the rainfall R, air humidity value H, wind speed S, wind direction Y, temperature value T, air pollution index M, number of birds N, bird food change ratio B and pollution degree PI in step S4, the correlation between the rainfall R, air humidity value H, wind speed S, wind direction Y, temperature value T, air pollution index M, number of birds N, bird food change ratio B and pollution degree PI in step S5 is further added, and the weight relationship and correlation are used as the basic reference logic to guide the training of the deep learning model;

[0136] S6.3. Use a deep learning model to simulate the change of pollution level over time and obtain simulated pollution level simulation data;

[0137] Among them, the task of the deep learning model is to predict the change of the pollution degree of the insulators in the simulated target area over time. The drone inspection task formulated according to the simulated predicted pollution degree has a scientific basis, which also makes the safety inspection of insulators more efficient and reliable.

[0138] S6.4. Use the data verification module to verify the pollution degree simulation data at the corresponding time point with the newly detected data of the drone and the newly detected electric field data of the electric field detection equipment, and optimize the deep learning model according to the verification results;

[0139] Among them, as the amount of data in the historical database increases, the data used for the deep learning model's operation also increases accordingly, and the simulation prediction results of the deep learning model are verified in combination with the specific results of drone detection, which makes it more convenient to optimize and adjust the deep learning model, and further increases the simulation prediction accuracy of the deep learning model with the increase of time and data volume;

[0140] S6.5. Use the optimized deep learning model to adjust the drone inspection task, and continuously optimize the deep learning model with the newly added inspection data. Continuously adjust the drone inspection task according to the continuously optimized deep learning model, and repeat the process until the difference between the insulator contamination degree detected by the drone and the insulator contamination degree simulated by the deep learning model in the same time period is within the threshold;

[0141] Among them, optimizing the deep learning model according to the data detected by the drone inspection task, and then adjusting the drone inspection task through the optimized deep learning model can form a virtuous circle, making the simulation prediction results of the deep learning model more and more accurate over time, and also making the results detected by the drone inspection task more helpful in the formulation of the insulator cleaning plan;

[0142] In addition, the setting of the insulator pollution difference threshold is based on the specific meteorological environment and ecological environment of the target area. The specific reference data items of the interference factors include but are not limited to: temperature data, rainfall data, wind data, air pollution index, bird-related data, etc. When the weight values ​​of two or more data in the above reference data items differ too much, the correlation of the reference data items will change, thereby causing the final difference threshold range to fluctuate. Therefore, when setting the insulator pollution difference threshold for different target areas, the staff can analyze the specific situation to formulate the threshold.

[0143] S7, adjust the drone detection task according to the change of pollution level simulated by the deep learning model;

[0144] Among them, when the deep learning model is optimized to the point where the simulated insulator contamination level matches the actual detection results of the UAV inspection task, it can be determined that the currently formulated UAV inspection task can meet the insulator contamination detection needs within a certain period of time.

[0145] S8. Using an icon generation module to generate a line graph of the data detected by the drone and the electric field detection device, and displaying it in a display system;

[0146] Among them, the horizontal axis of the line graph generated by the chart generation module is time, and the vertical axis of the line graph is the degree of pollution. A data set is established on the degree of pollution coordinate corresponding to each detection time point. The items included in the data set are: degree of pollution, electric field distribution data, air humidity data within the detection period, rainfall data within the detection period, wind data within the detection period, temperature data within the detection period, air pollution data within the detection period, insulator hyperspectral detection data, bird activity change data within the detection period, bird food change data within the detection period, and bird droppings change data within the detection period. Since the degree of pollution is calculated by combining the electric field distribution data and the insulator hyperspectral detection data actually detected, the obtained degree of pollution can be directly used as the formulation standard for the insulator cleaning task, and the air humidity data within the detection period, the rainfall data within the detection period, the wind data within the detection period, the temperature data within the detection period, the air pollution data within the detection period, the bird activity change data within the detection period, the bird food change data within the detection period, and the bird droppings change data within the detection period facilitate the staff to quickly screen the data according to the data changes and judge the changes in the degree of pollution of the insulators in the target area by themselves based on their personal experience;

[0147] Furthermore, in the data set: the detection period of air humidity data within the detection period, rainfall data within the detection period, wind data within the detection period, temperature data within the detection period, air pollution data within the detection period, bird activity change data within the detection period, bird food change data within the detection period and bird droppings change data within the detection period is a detection period, and the time between the current detection time point and the previous detection time point is a detection period. According to the different time intervals of the drone detection tasks, the length of the detection period varies with the time intervals of the drone detection tasks.

[0148] Working principle:

[0149] like Figure 1-7As shown, the present invention first establishes a basic database through testing, sampling and investigation, and obtains various data in the historical database through the UAV detection tasks formulated in advance, and then uses the basic database and the historical database to build a deep learning model, uses the deep learning model to optimize the UAV detection task, and reversely optimizes the deep learning model according to the new detection data obtained by the UAV, and adjusts the UAV detection task according to the optimized deep learning model, forming a virtuous circle, so that the entire UAV detection task is more in line with the environment of the target area, and the detection of insulator contamination is more efficient.

[0150] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any technician familiar with the profession may use the technical content disclosed above to change or modify it into an equivalent embodiment with equivalent changes and apply it to other fields. However, any simple modification, equivalent change and modification made to the above embodiment based on the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A transmission line insulator contamination detection system based on a hyperspectral camera. Features: It includes a server, which is equipped with a deep learning model, a data verification module, a basic database of a chart generation module, and a historical database. The server is data-connected to a monitoring center, and the monitoring center is signal-connected to a drone and an electric field detection device; The basic database includes bird data, hyperspectral data and electric field data. The bird data includes bird population data, bird food data and bird droppings data. The hyperspectral data includes insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and insulator pollution sample hyperspectral data. The electric field data includes insulator sample electric field data and insulator pollution sample electric field data. The historical database includes environmental data, bird change data, insulator hyperspectral detection historical data and insulator electric field detection historical data; the environmental data includes rainfall historical data, air humidity historical data, wind power historical data, temperature historical data and air pollution historical index; the bird change data includes bird activity data, bird food change data and bird droppings change data; The UAV includes a UAV flight control system, a hyperspectral gimbal camera and a high-speed camera; The monitoring center includes a data entry system, a display system and a data transmission system.

2. A transmission line insulator contamination detection system using a hyperspectral camera according to claim 1, Features: It also includes a transmission line insulator contamination detection system constructed by a server, a monitoring center and a drone. The construction of the transmission line insulator contamination detection system specifically includes the following steps: S1. First, establish the basic database; S2, then formulate the drone inspection task; S3, build a historical database based on the data obtained from the drone inspection mission; S4. Analyze the impact of environmental data and bird change data on the change of insulator pollution degree, and adjust the weight of each data according to the historical database; S5. According to the weights of environmental data and bird change data, the grey correlation method is used to analyze the correlation between the environmental data and the bird change data; S6. According to the specific correlation between the data and the weight of the data in the historical database, a deep learning model is constructed, and the deep learning model is used to analyze and predict the change of the degree of contamination of the insulator; S7, adjust the drone detection task according to the change of pollution level simulated by the deep learning model; S8. Use the icon generation module to generate a line graph of the data detected by the drone and the electric field detection device, and display it in the display system.

3. A transmission line insulator contamination detection system using a hyperspectral camera according to claim 2, Features: In step S1, the construction of the basic database specifically includes the following steps: S1.

1. First, investigate the birds in the area where the transmission tower is located and obtain the corresponding bird droppings samples, as well as insulator samples and air pollutant samples in the target area. Then, simulate the target area environment to produce multiple insulator pollution samples with different pollution levels. S1.2, then photographing the insulator samples, bird droppings samples, air pollutant samples and insulator pollution samples through the experimental hyperspectral camera, and obtaining the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data, and then classifying the insulator sample hyperspectral data, bird droppings sample hyperspectral data, air pollutant sample hyperspectral data and multiple groups of insulator pollution sample hyperspectral data as hyperspectral data through the data entry system, and uploading them to the basic database in the server; S1.3, then installing the insulator samples and the insulator pollution samples on the high-voltage power transmission simulation equipment, and then performing electric field detection on the insulator samples and the multiple groups of insulator pollution samples through the experimental electric field detection equipment to obtain the insulator sample electric field data and the multiple groups of insulator pollution sample electric field data, and then classifying the insulator sample electric field data and the insulator pollution sample electric field data as electric field data, and uploading them to the basic database in the server through the data entry system; S1.

4. Obtain bird population data in the target area through surveys, and obtain corresponding bird food data and bird droppings data based on the bird population data, and classify the obtained bird population data, bird food data and bird droppings data as bird data, and then upload them to the basic database of the server through the data entry system.

4. According to claim 2, a transmission line insulator contamination detection system using a hyperspectral camera, Features: In step S3, the construction of the historical database specifically includes the following steps: S3.

1. First, obtain data such as rainfall, air humidity, wind speed, temperature and air pollution index through the meteorological center, and arrange the same data by date over time, divide them into historical rainfall data, historical air humidity data, historical wind speed data, historical temperature data and historical air pollution index, and classify them as environmental data; S3.

2. Then, when performing the insulator inspection task, the high-speed camera of the drone is used to investigate the bird activities and plant distribution in the target area, and to sample bird droppings to obtain bird activity data and bird food change data. Then, based on the bird food change data and the bird droppings samples, the bird droppings composition changes are analyzed to obtain the bird droppings change data. Then, the bird activity data, bird food change data, and bird droppings change data at different times are sorted according to the timeline and classified as bird change data. S3.3, sorting the insulator hyperspectral detection data obtained from different UAV detection tasks according to the time axis and classifying them as insulator hyperspectral detection historical data; S3.

4. Arrange the insulator electric field detection data of different time periods according to the time axis and classify them into insulator electric field detection historical data; S3.

5. Upload the environmental data and bird change data to the historical database of the server through the data entry system, and upload the insulator hyperspectral detection historical data and insulator electric field detection historical data to the historical database of the server through the data transmission system.

5. According to claim 1, a transmission line insulator contamination detection system using a hyperspectral camera, Features: In step S4, the pollution degree is determined based on the upper limit of the flashover occurrence of the insulator, and the pollution degree of the insulator is divided into specific levels according to the historical data of the insulator hyperspectral detection and the historical data of the insulator electric field detection. Then, weights are assigned to the rainfall, air humidity value, wind speed, wind direction, temperature value, air pollution index, number of birds, and bird food change ratio. The specific formula is: PI=W 1 ×R+W 2 ×H+W 3 ×S+W 4 ×Y+W 5 ×T+W 6 ×M+W 7 ×N+W 8 ×B Among them, let PI be the pollution degree, R be the rainfall, H be the air humidity value, S be the wind speed, Y be the wind direction, T be the temperature value, M be the air pollution index, N be the number of birds, B be the change ratio of bird food, W be 1 is the weight value of rainfall, W 2 is the weight value of the air humidity value, W 3 is the weight value of wind speed, W 4 is the weight value of wind direction, W 5 is the weight value of the temperature value, W 6 is the weight value of the air pollution index, W 7 is the weight value of the number of birds, W 8 is the weight value of the bird food change ratio; According to different levels of pollution, the specific weights of rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population, and bird food change ratio are calculated using the historical database.

6. A transmission line insulator contamination detection system using a hyperspectral camera according to claim 5, Features: In step S5, according to the weights of the environmental data and the bird change data in step S4, the grey correlation method is used to analyze the correlation between the environmental data and the bird change data, specifically: S5.

1. The pollution level PI set in step S4 is used as the parent sequence, which is recorded as: PI={PI(k)k=1,2,.....,n} S5.2, the rainfall R, air humidity H, wind speed S, wind direction Y, temperature T, air pollution index M, bird number N, and bird food change ratio B set in step S4 are taken as subsequences and recorded as: x i ={x i kk=1,2,......,n} Where i is the algebraic symbol of R, H, S, Y, T, M, N, B; S5.

3. Then divide each value in the rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the first value of the corresponding item, and replace the original corresponding value with the obtained value. The formula is: S5.

4. Further narrow the range of variables and simplify the calculation. Divide each value of rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population and bird food change ratio by the mean of the corresponding item, and replace the original corresponding value with the obtained value. The formula is: S5.5, then calculate the absolute difference between the pollution degree and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, bird population, and bird food change ratio, i.e. Δ i (k)=|PI(k)-x i (k)|, and obtain the difference matrix; S5.

6. Find the minimum difference between the two levels and the maximum difference between the two levels according to the difference matrix, represent them by a and b respectively, and then calculate the correlation coefficient: Among them, ρ is the resolution coefficient; S5.

7. For each subsequence, calculate the average value of all its correlation coefficients to obtain the correlation between the subsequence and the parent sequence. The formula is: Among them, r i That is, the correlation between pollution degree and rainfall, air humidity, wind speed, wind direction, temperature, air pollution index, number of birds, and change ratio of bird food.

7. A transmission line insulator contamination detection system using a hyperspectral camera according to claim 6, Features: In step S6, the construction process of the deep learning model is specifically as follows: S6.

1. Use the Keras library to build an LSTM model as the core of the deep learning model, and use the basic database as the basic data set to build a deep learning model; S6.2, using the historical database as a training set, and combining the weights of each data in S4 {W 1 ,W 2 ,W 3 ,W 4 ,W 5 ,W 6 ,W 7 ,W 8 } and the specific correlation between the data in S5 i , train the deep learning model; S6.

3. Use a deep learning model to simulate the change of pollution level over time and obtain simulated pollution level simulation data; S6.

4. Use the data verification module to verify the pollution degree simulation data at the corresponding time point with the newly detected data of the drone and the newly detected electric field data of the electric field detection equipment, and optimize the deep learning model according to the verification results; S6.

5. Use the optimized deep learning model to adjust the UAV inspection task, and continuously optimize the deep learning model with the newly added inspection data. Continuously adjust the UAV inspection task according to the continuously optimized deep learning model, and repeat the cycle until the difference between the insulator contamination degree detected by the UAV and the insulator contamination degree simulated by the deep learning model in the same time period is within the threshold.

8. The transmission line insulator contamination detection system using a hyperspectral camera according to claim 2, Features: In the step S8, the horizontal axis of the line graph generated by the chart generation module is time, and the vertical axis of the line graph is the pollution degree. A data set is established on the pollution degree coordinate corresponding to each detection time point, and the items included in the data set are: pollution degree, electric field distribution data, air humidity data within the detection period, rainfall data within the detection period, wind speed data within the detection period, temperature data within the detection period, air pollution data within the detection period, insulator hyperspectral detection data, bird activity change data within the detection period, bird food change data within the detection period, and bird droppings change data within the detection period.

9. A transmission line insulator contamination detection system using a hyperspectral camera according to claim 8, Features: The data set includes: air humidity data within a detection period, rainfall data within a detection period, wind data within a detection period, temperature data within a detection period, air pollution data within a detection period, insulator hyperspectral detection data, bird activity change data within a detection period, bird food change data within a detection period, and bird droppings change data within a detection period. The detection period is the time between the current detection time point and the previous detection time point. According to the different time intervals of the drone detection tasks, the length of the detection period varies with the time interval of the drone detection tasks.

Images