Method and system for evaluating the degree of contamination on the surface of an insulator

Abstract

The application relates to a kind of insulator surface contamination degree evaluation method and system, the method comprises: obtaining the leakage current data of real operation insulator under different surface contamination degree, different weather and meteorological condition and corresponding weather and meteorological condition and surface contamination degree;The obtained leakage current data is carried out discrete fourier transform, the spectral characteristic quantity of leakage current is extracted, and the multi-dimensional feature vector combined with weather and meteorological condition data is constructed in combination with spectral characteristics and weather and meteorological characteristics;The deep learning neural network model based on convolutional neural network is constructed;The multi-dimensional feature vector constructed is used as the input of deep learning neural network model, and deep learning training is carried out until the model converges;The surface contamination degree of the insulator to be evaluated is predicted and evaluated using the trained deep learning neural network model.The method and system can improve the accuracy and real-time performance of insulator surface contamination degree evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power equipment monitoring, and particularly relates to an insulator surface contamination degree evaluation method and system. BACKGROUND

[0002] The insulator is an important component in the power system, and the contamination degree on the surface of the insulator is crucial for the safe and stable operation of the power equipment. The traditional insulator surface contamination monitoring data analysis mainly uses the leakage current amplitude size discrimination method, and the harmonic spectrum diagram in the leakage current is calculated, and then the contamination degree on the surface of the insulator is observed, compared and searched by artificial periodical observation. It is difficult to accurately determine the contamination working condition on the surface of the insulator by relying on a single leakage current amplitude, and the harmonic spectrum in the leakage current of the insulator to a large extent represents the contamination degree on the surface of the insulator. However, the existing method needs artificial observation and searching, and has subjectivity and limitation, and cannot effectively process and analyze a large amount of monitoring data in real time, which limits the effect in actual application.

[0003] With the development of artificial intelligence technology, especially big data processing and pattern recognition technology, artificial intelligence can be applied to the analysis and evaluation of insulator surface contamination, so as to realize accurate and real-time evaluation of the degree of insulator surface contamination. Patent CN112884720A discloses a power distribution line pollution flashover insulator detection method and system, which comprises: acquiring a to-be-detected image; extracting insulators in the to-be-detected image to obtain a to-be-detected insulator image; calculating the Fourier spectrum of each to-be-detected insulator image to obtain a detection spectrum feature; calculating the detection average Fourier spectrum density of each to-be-detected insulator image according to the detection spectrum feature of each to-be-detected insulator image; inputting the detection average Fourier spectrum density of each to-be-detected insulator image into a trained single classifier model to obtain a detection result of whether the insulator flashes; the method and system can realize accurate and efficient detection of pollution flashover insulators. Patent CN104237757A discloses an insulator contamination discharge pattern recognition method based on EEMD and marginal spectrum entropy. The method first performs artificial contamination experiments, collects a large number of acoustic emission signals of different discharge modes of insulators, then decomposes the acoustic emission signals of different discharge stages using the ensemble empirical mode decomposition method to obtain their intrinsic mode components, and then performs Hilbert transform on the intrinsic mode components to obtain the time-frequency spectrum of the acoustic emission signals. On this basis, the marginal spectrum of the acoustic emission signal is calculated, and the marginal spectrum entropy and the center of gravity frequency of the acoustic emission signal are used as the characteristic values of the acoustic emission signal. Finally, the neural network is used to realize pattern recognition of different discharge stages of the insulator. Combined with a large number of insulator contamination discharge tests, the results of the method show that the invention can effectively distinguish the three different discharge stages of the contamination discharge of the insulator, and provides technical support for judging the external insulation state of the insulator and realizing pollution flashover warning. The above patents disclose methods for detecting and identifying the contamination of insulators by obtaining the frequency spectrum features of the insulator images, the time-frequency spectrum of the acoustic emission signals during insulator discharge, etc. Although the automation degree and efficiency of insulator contamination detection and identification are improved, these methods do not directly use the harmonic spectrum features of the insulator leakage current, and do not consider the influence of weather changes on the detection results, so the accuracy of insulator surface contamination detection and evaluation needs to be further improved. SUMMARY

[0004] The purpose of the present application is to provide an insulator surface contamination degree evaluation method and system which can improve the accuracy and real-time performance of insulator surface contamination degree evaluation.

[0005] To achieve the above purpose, the technical solution adopted by the present application is as follows: an insulator surface contamination degree evaluation method, comprising:

[0006] Obtaining leakage current data of real running insulators under different surface contamination degrees and different weather conditions, and corresponding weather conditions and surface contamination degrees;

[0007] Discrete Fourier transform is performed on the obtained leakage current data, the spectral feature quantity of the leakage current is extracted, and a multi-dimensional feature vector combining the spectral feature and the weather condition feature is constructed;

[0008] A deep learning neural network model based on a convolutional neural network is constructed; the constructed multi-dimensional feature vector is used as the input of the deep learning neural network model, and deep learning training is performed until the model converges;

[0009] The trained deep learning neural network model is used to predict and evaluate the surface contamination degree of the insulator to be evaluated.

[0010] Further, the surface contamination degree of the insulator is divided into normal, warning and serious, and the output Y of the deep learning neural network model is [y0, y1, y2], wherein y0 represents the probability value of the contamination degree being normal, y1 represents the probability value of the contamination degree being warning, and y2 represents the probability value of the contamination degree being serious.

[0011] Further, the leakage current signal of the insulator is collected by the current sensor; the leakage current signal is represented as a linear superposition of a plurality of signal components, and current sampling is performed by the following current sampling formula to obtain the leakage current data of the insulator.

[0012]

[0013] In the formula, x(nΔt) is a discrete time signal obtained by sampling, nΔt is a discrete time, n is a time sequence, and Δt is a sampling interval; N is the total number of signal components, i is a signal component, ω i 、A i 、p i The angular frequency, amplitude and initial phase of the i signal component, respectively.

[0014] Further, discrete Fourier transform is performed on the discrete time signal obtained by sampling by formula (1):

[0015]

[0016] In the formula, X[k] is the spectrum of the discrete time signal, N is the length of the time sequence, and k is the frequency index.

[0017] Further, the spectral amplitude is extracted as the spectral feature quantity of the leakage current; the amplitude corresponding to the k frequency component is:

[0018]

[0019] In the formula, Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k].

[0020] Further, a (p+q)-dimensional multi-dimensional feature vector is formed by the first p harmonic amplitudes of the leakage current and q weather and meteorological features.

[0021] Further, the weather and meteorological state data includes humidity data and temperature data, thereby forming 2 weather and meteorological features including humidity features and temperature features.

[0022] Further, a deep learning neural network model based on Softmax regression is constructed, and the deep learning neural network model includes 3 convolutional layers and 2 fully connected layers.

[0023] Further, the deep learning neural network model based on Softmax regression normalizes the data of the previous layer by a Softmax function, converts it into a value between 0 and 1, and uses it as a probability distribution as a multi-classification target prediction value.

[0024]

[0025] In the formula, x i is an element in the vector, Softmax() is a Softmax function, and Y(x i ) is the output of the Softmax function.

[0026] The application also provides an insulator surface contamination degree evaluation system, which comprises a memory, a processor and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the above-mentioned method steps can be realized.

[0027] Compared with the prior art, the application has the following beneficial effects: the application provides an insulator surface contamination degree evaluation method and system, which analyzes and identifies the current signal of the insulator surface contamination by using artificial intelligence technology, adopts a deep learning method based on spectral features, and realizes accurate and real-time evaluation of the insulator surface contamination degree by deep learning of the spectral feature quantities under different contamination degrees of the insulator surface and different weather and meteorological states. The method can fully mine the implicit information in the insulator leakage current data, improve the accuracy and objectivity of the evaluation, and also improve the processing efficiency of a large amount of monitoring data. Therefore, the application has strong practicability and broad application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0028] Figure 1 is a method implementation flowchart of an embodiment of the application.

[0029] Figure 2 is the current waveform collected when the contamination level is normal in the embodiment of the present application;

[0030] Figure 3 is the current waveform collected when the contamination level is warning in the embodiment of the present application;

[0031] Figure 4 is the current waveform collected when the contamination level is serious in the embodiment of the present application;

[0032] Figure 5 is the frequency amplitude component calculated when the contamination level is normal in the embodiment of the present application;

[0033] Figure 6 is the frequency amplitude component calculated when the contamination level is warning in the embodiment of the present application;

[0034] Figure 7 is the frequency amplitude component calculated when the contamination level is serious in the embodiment of the present application;

[0035] Figure 8 is the architecture diagram of the deep learning neural network model based on Softmax regression in the embodiment of the present application;

[0036] Figure 9 is the loss curve diagram of the deep learning neural network model in the embodiment of the present application. DETAILED DESCRIPTION

[0037] The present application will be further described below in conjunction with the accompanying drawings and embodiments.

[0038] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise indicated, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application pertains.

[0039] It should be noted that the terms used herein are only for the purpose of describing the specific embodiments, and are not intended to limit the exemplary embodiments according to the present application. As used herein, the singular form is intended to include the plural form unless the context clearly indicates otherwise, and furthermore, it should be understood that when the terms "comprise" and / or "include" are used in the specification, there is a presence of the features, steps, operations, devices, components and / or combinations thereof.

[0040] As shown in Figure 1 The present embodiment provides an insulator surface contamination level evaluation method, comprising the following steps:

[0041] 1) Obtain the leakage current data of the real running insulator under different surface contamination levels and different weather conditions, and the corresponding weather conditions and surface contamination levels.

[0042] 2) Discrete Fourier transform is performed on the obtained leakage current data to extract the spectral feature quantity of the leakage current, and the weather condition data is combined to construct a multi-dimensional feature vector combining the spectral feature and the weather condition feature.

[0043] 3) A deep learning neural network model based on a convolutional neural network is constructed. The constructed multi-dimensional feature vector is used as the input of the deep learning neural network model, and the deep learning training is performed until the model converges.

[0044] 4) The trained deep learning neural network model is used to predict and evaluate the surface contamination level of the insulator to be evaluated.

[0045] In the method, the surface contamination level of the insulator is divided into normal, warning and serious, and the output Y of the deep learning neural network model is [y0, y1, y2], wherein y0 represents the probability value of the contamination level being normal, y1 represents the probability value of the contamination level being warning, and y2 represents the probability value of the contamination level being serious.

[0046] The contamination level of the insulator surface will cause the change of the surface resistance, thereby affecting the electrical performance of the insulator. When the surface contamination level of the insulator is high, the contamination may form a conductive path, increasing the path of the leakage current, resulting in the increase of the high-order harmonic of the leakage current. In addition, the contamination may also cause partial discharge on the surface of the insulator, further increasing the generation of the high-order harmonic of the leakage current. In addition, weather changes also have an impact on the high-order harmonic of the leakage current. For example, humidity, temperature, wind speed and other factors also have an impact on the electrical performance of the insulator. In a high humidity environment, the contamination on the surface of the insulator is easy to absorb moisture, increasing the formation of the conductive path, thereby increasing the generation of the high-order harmonic of the leakage current. The change of temperature also affects the conductivity of the contamination on the surface of the insulator, thereby affecting the size of the high-order harmonic of the leakage current. The present application adopts the spectral feature formation mechanism of the leakage current signal as the starting point, collects a large amount of leakage current data of the real running insulator, then classifies and labels the data and sends it into the constructed deep learning neural network model based on the convolutional neural network for training until the model converges, thereby realizing the effective prediction of the surface contamination level of the insulator by using the nonlinear, adaptive and fault-tolerant capabilities of the neural network.

[0047] In step 1) of the embodiment, the current waveform of the leakage current signal of the insulator of the 110kV and above voltage level overhead transmission line is collected by a current sensor.

[0048] The leakage current signal is expressed as a linear superposition of a plurality of signal components, and current sampling is performed by a current sampling formula as follows to obtain the leakage current data of the insulator.

[0049]

[0050] wherein x(nΔt) is a discrete-time signal obtained by sampling, nΔt is a discrete time, n is a time sequence, and Δt is a sampling interval; N is the total number of signal components, i is a signal component, ω i , A i , p i are the angular frequency, amplitude, and initial phase of the i signal component, respectively.

[0051] The current waveforms collected when the contamination degree is normal, warning, and severe are shown in FIGS. 1, 2, and 3, respectively. Figures 2-4

[0052] In step 2), a discrete Fourier transform is performed on the discrete-time signal obtained by sampling according to formula (1):

[0053]

[0054] wherein X[k] is the frequency spectrum of the discrete-time signal, N is the length of the time sequence, and k is a frequency index.

[0055] The spectral amplitude is extracted as the spectral characteristic quantity of the leakage current; the amplitude corresponding to the k frequency component is:

[0056]

[0057] wherein Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k].

[0058] The frequency amplitude components calculated when the contamination degree is normal, warning, and severe are shown in FIGS. 4, 5, and 6, respectively. Figures 5-7

[0059] Then, a (p+q)-dimensional multi-dimensional feature vector is formed using the first p harmonic amplitudes of the leakage current and q weather and meteorological characteristics.

[0060] In this embodiment, the weather and meteorological state data includes humidity data and temperature data, thereby forming 2 weather and meteorological characteristics including a humidity characteristic and a temperature characteristic. A 400-dimensional feature vector X=[t, h, A1, …, A398] is formed using the first 398 harmonic amplitudes of the leakage current, the temperature, and the humidity, wherein t is the temperature characteristic, h is the humidity characteristic, and A1-A398 are the harmonic amplitudes. The 400-dimensional feature vector is converted into a 20*20 matrix vector as the input of the neural network.

[0061] ​​In step 3) of the present embodiment, a deep learning neural network model based on Softmax regression is constructed, and the architecture thereof is as shown in Figure 8 The advantage of this model is that the Softmax regression function can convert the output into a probability distribution of multiple categories, so that the results of the model output are more intuitive and easy to interpret. The Softmax regression function is suitable for multi-classification problems, can handle multi-classification tasks, and has good convexity in optimization algorithms such as gradient descent, which is beneficial to the optimization and convergence of model parameters.

[0062] For discriminating insulators of different pollution levels, the present method selects the first 398 harmonics of leakage current and temperature and humidity data to jointly form a 400-dimensional feature vector and convert it into a matrix vector as the input parameter of the deep learning neural network model. The deep learning neural network model includes 3 convolutional layers and 2 fully connected layers, and the output response variable Y = [y0, y1, y2], y0 represents the probability value of the pollution level being normal, y1 represents the probability value of the pollution level being warning, and y2 represents the probability value of the pollution level being severe.

[0063] The deep learning neural network model normalizes the data of the previous layer by the Softmax function, converts it into a value between 0 and 1, and uses it as a probability distribution as the target prediction value of multi-classification; the calculation formula is as follows:

[0064]

[0065] In the formula, x i is an element in the vector, Softmax() is the Softmax function, and Y(x i ) is the output of the Softmax function.

[0066] In the present embodiment, for the neural network training process in the deep learning neural network model based on Softmax regression, a total of 5000 insulator leakage current waveform data samples are collected, including 1600 data samples with normal pollution level, 1400 data samples with warning pollution level, and 2000 data samples with severe pollution level. All samples are manually labeled and input to the neural network for training, and the network learning accuracy reaches 100%.

[0067] After 15 rounds (Epoch) of training of 5000 images per Epoch, the loss curve obtained is as shown in Figure 9

[0068] 2000 data samples containing three pollution levels are selected from the collected waveforms to test and verify the accuracy of the present method, and the accuracy reaches 96%, as shown in Table 1.

[0069] ​Table 1 Test results of the method for evaluating the degree of pollution on the surface of insulators

[0070] Degree of contamination Number of test data Number of correct recognitions Accuracy rate Normal 700 677 96.7％ Warning 700 672 96％ Serious 600 579 96.5％

[0071] The embodiment also provides an evaluation system for the degree of pollution on the surface of insulators, comprising a memory, a processor, and computer program instructions stored in the memory and capable of being executed by the processor, and when the processor executes the computer program instructions, the method steps described above can be implemented.

[0072] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can 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-ROMs, optical storage, etc.) containing computer-usable program code.

[0073] The present application is described with reference to flowcharts and / or block diagrams of the method, device (system), and computer program product according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows and / or blocks in the flowcharts and / or block diagrams 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 apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus produce a device implemented in the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that performs the functions specified in the flow(s) or block(s).

[0074] These computer program instructions can also be stored in a computer-readable memory that can direct the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction apparatus, which implements the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that performs the functions specified in the flow(s) or block(s).

[0075] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, so that a series of operation steps are performed on the computer or other programmable data processing apparatus to produce a computer-implemented process, so that the instructions executed on the computer or other programmable data processing apparatus provide a process for implementing the flowcharts and / or block diagrams. Figure 1 one or more flows and / or blocks Figure 1 an apparatus that performs the functions specified in the flow(s) or block(s).

[0076] The above descriptions are only the preferred embodiments of the present application, not intended to limit the present application in other forms. Any person skilled in the art can make changes or modifications to the equivalent embodiments with the disclosed technical contents. However, any simple modification, equivalent change and modification made to the above embodiments without departing from the technical solution of the present application and according to the technical essence of the present application still belong to the protection scope of the present application.

Claims

1. A method for evaluating the degree of contamination on the surface of an insulator, characterized by, The method comprises: obtaining leakage current data of a real running insulator under different surface contamination degrees and different weather conditions, and corresponding weather conditions and surface contamination degrees; performing discrete Fourier transform on the obtained leakage current data, extracting spectral feature quantities of the leakage current, and constructing a multi-dimensional feature vector combining the spectral features and weather conditions; constructing a deep learning neural network model based on a convolutional neural network; taking the constructed multi-dimensional feature vector as the input of the deep learning neural network model, and performing deep learning training until the model converges; using the trained deep learning neural network model to predict and evaluate the surface contamination degree of an insulator to be evaluated; acquiring the leakage current signal waveform of the insulator through a current sensor; representing the leakage current signal as a linear superposition of multiple signal components, and performing current sampling through the following current sampling formula to obtain the leakage current data of the insulator; (1) In the formula, x(nΔt) is a discrete-time signal obtained by sampling, nΔt is a discrete time, n is a time sequence, Δt is a sampling interval; N is the total number of signal components, i is a signal component, ω i , A i , p i are the angular frequency, amplitude and initial phase of the i signal component respectively. performing discrete Fourier transform on the discrete-time signal sampled through formula (1): (2) wherein X[k] is the frequency spectrum of the discrete-time signal, N is the length of the time series, and k is the frequency index; extracting the spectral amplitude as the spectral feature quantity of the leakage current; the amplitude corresponding to the k frequency component is: (3) wherein Re(X[k]) represents the real part of X[k], and Im(X[k]) represents the imaginary part of X[k]; forming a (p+q)-dimensional multi-dimensional feature vector with the first p harmonic amplitudes of the leakage current and q weather conditions. The weather condition data includes humidity data and temperature data, thereby forming two weather conditions including humidity features and temperature features.

2. The insulator surface contamination degree evaluation method according to claim 1, characterized by, The surface contamination degree of the insulator is divided into normal, warning and severe, and the output Y of the deep learning neural network model is [y0, y1, y2], wherein y0 represents the probability value of the contamination degree being normal, y1 represents the probability value of the contamination degree being warning, and y2 represents the probability value of the contamination degree being severe.

3. The insulator surface contamination degree evaluation method according to claim 1, characterized by, The deep learning neural network model based on Softmax regression comprises three convolutional layers and two fully connected layers.

4. The insulator surface contamination degree evaluation method according to claim 3, characterized by, The deep learning neural network model based on Softmax regression normalizes the data of the previous layer through a Softmax function, converts it into a value between 0 and 1, and uses it as a probability distribution as the target prediction value of multi-classification. (3)； In the formula, x i is an element in a vector, Softmax( ) is a Softmax function, and Y(x i ) is an output of the Softmax function.

5. An insulator surface contamination level assessment system, characterized by, The computer program instructions stored in the memory and capable of being executed by the processor can implement the method of any one of claims 1-4 when the processor executes the computer program instructions.

Images