Traction network line lightning arrester health state evaluation system

Abstract

The invention belongs to the technical field of traction network line fault detection, and particularly relates to a traction network line lightning arrester health state evaluation system which comprises a data acquisition layer, a data calculation layer, a state evaluation layer and a terminal database. The data acquisition layer acquires and sends the characteristic information amount of the operating environment and the working condition of the lightning arrester to the data calculation layer; the data calculation layer is used for calculating and processing the characteristic information amount of the lightning arrester and sending related data to the state evaluation layer; and the state evaluation layer is used for evaluating the health state of the lightning arrester and comparing and predicting the future state in combination with the current influence factors. And the terminal database is used for storing data required by the training model and an evaluation prediction result. Lightning arrester leakage current fluctuation caused by traction network load change is converted into a constant component capable of being monitored, state evaluation and future prediction of the lightning arrester are achieved, the operation and maintenance efficiency of equipment is effectively improved, and the situation that faults occur to affect train and personnel safety is avoided.

Description

Technical Field

[0001] The invention belongs to the technical field of traction network line fault detection, and in particular relates to a traction network line arrester health status assessment system. Background Art

[0002] Lightning arresters are an important type of equipment in power systems, which protect other power equipment from lightning and overvoltage damage in substations and lines. Zinc oxide lightning arresters are widely used in power systems because of their good voltage characteristics and nonlinearity, which greatly improves the level of protection of power equipment. There are currently two main ways to monitor lightning arresters: the first is offline maintenance, which removes the lightning arrester from the high-voltage system and conducts a more comprehensive test of the lightning arrester's insulation performance, withstand voltage, etc., which can monitor the performance of the lightning arrester in many aspects, but there is a delay in identifying its parameter changes; the other method is online monitoring, which uses sensors to collect the electrical parameters of the lightning arrester in real time during operation and analyzes it to obtain its status information. Based on actual operating conditions and on-site accident analysis, the main external factors affecting the operating status of the lightning arrester are harmonic voltages, and the internal factors are air gaps formed by insulation structure defects and internal moisture.

[0003] In the prior art, lightning arresters widely use FFT (Fast Fourier Decomposition) decomposition to process leakage current, but this method is not effective on the traction network where electric locomotives and EMUs are located. The main reason is that the traction network contains a large number of high-frequency harmonics of 45th and above; in addition, the load fluctuation of the traction network line is large, resulting in frequent voltage changes and large changes in the leakage current of the lightning arrester; and the distance along the traction network line is long, and the environmental conditions at different locations are also different. Therefore, in the complex working conditions of the traction network line, the prior art has the problems of large calculation amount, slow response speed, and inability to directly judge the health and aging type of the lightning arrester.

[0004] Therefore, there is an urgent need for a traction network line lightning arrester health status assessment system that can significantly reduce the amount of signal calculations required for processing, extract the required characteristic signals more quickly, and improve the response speed of the status assessment system; it can more directly and effectively judge the health and aging type of the lightning arrester, while significantly improving the status assessment speed at high-frequency acquisition rates. Summary of the invention

[0005] The object of the present invention is to provide a traction network line arrester health status assessment system, characterized in that it comprises: a data acquisition layer, a data calculation layer, a status assessment layer and a terminal database;

[0006] The data acquisition layer includes: a state data acquisition unit and a state data transmission unit; the data acquisition layer sends characteristic information about the operating environment and the working status of the arrester to the data calculation layer;

[0007] The data calculation layer includes: a digital filtering module, a data decomposition module and a data reconstruction module; the data calculation layer calculates and processes the characteristic information of the arrester and sends relevant data to the state evaluation layer;

[0008] The state assessment layer includes: a state assessment preliminary review module, a multi-factor decision module, a model training module, a state review and prediction module, and a data transmission module; the state assessment layer implements the assessment of the health state of the arrester and compares and predicts the future state in combination with the current influencing factors;

[0009] The terminal database includes: a data storage unit and a database data transmission unit; the terminal database realizes the storage of data required for the training model and the evaluation prediction results.

[0010] The state data acquisition unit collects device state information based on the edge sensing device; the edge sensing device includes: a leakage current sensor and a temperature and humidity sensor; the leakage current sensor is used to collect leakage current signals; the temperature and humidity sensor is used to collect the temperature and humidity of the device operating environment;

[0011] The state data transmission unit is used to upload the collected real-time state data to the data calculation layer and the state evaluation layer through the data transmission line for processing.

[0012] The digital filtering module uses an SG filter to remove abnormalities and reduce noise from the device status data uploaded by the data acquisition layer, and then stores the normal data after noise reduction and transmits it to the data decomposition module;

[0013] The data decomposition module uses the local mean decomposition method to perform modal decomposition on the data transmitted by the digital filter module, decomposes the mixed signal with variable frequency and amplitude into signals of fixed frequency bands, removes the interference components with extremely low content, stores the components with higher content and transmits the signal components of each frequency band to the data reconstruction module;

[0014] The data reconstruction module uses fast Fourier decomposition to reconstruct the variable amplitude signal into a sinusoidal signal with constant amplitude and frequency, thereby reducing the difficulty of leakage current calculation, and stores the reconstructed signal and uploads it to the status evaluation layer through the data transmission module to perform arrester status evaluation.

[0015] The state assessment preliminary review module uses a high- and low-frequency current amplitude weighted comparison method, including: weighting the current effect by the content size of different frequencies, and then calculating the ratio with the current size under normal operation and comparing it with the judgment threshold, so as to conduct a preliminary review of the health status of the lightning arrester and obtain the preliminary review result level; the initial result level includes: normal level, warning level and fault level;

[0016] The multi-factor decision module pairs the external conditions with the preliminary review result levels of the corresponding time periods of the external conditions and establishes a multi-factor decision tree model; the external conditions include: temperature, humidity, running time, external pollution index and pre-set factor information that may cause an impact;

[0017] The model training module performs preliminary training and calculation on the established multi-factor decision tree model. By establishing an XGBoost model, the multi-factor decision tree model training is completed quickly and accurately, the current state of the equipment is judged, the future state of the equipment is predicted, and possible equipment failures are diagnosed and warned in advance.

[0018] The state review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review result under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

[0019] The data transmission module includes: a data receiving unit and an evaluation data transmission unit, which are used to upload and send relevant data through a wireless communication transmission method; the wireless communication transmission method includes: LoRa or 5G;

[0020] The data receiving unit is used to receive the device status information of the data acquisition layer, the prediction results of the data reconstruction module and the status review and prediction module;

[0021] The evaluation data transmission unit is used to transmit the reconstruction signal of the data reconstruction module to the state evaluation layer and the terminal database; and to send the updated state evaluation information and prediction results of the state review and prediction module to the model training module.

[0022] The data storage unit is used to store real-time original status data of the device, historical data and historical models, and newly trained and updated model data.

[0023] Another object of the present invention is to provide a method for assessing the health status of a lightning arrester, characterized in that it comprises the following steps:

[0024] Step S1, using the state data acquisition unit of the data acquisition layer to collect device state information; using the state data transmission unit of the data acquisition layer to send the operating environment of the lightning arrester and the characteristic information reflecting the working status to the data calculation layer;

[0025] Step S2, using the digital filtering module, data decomposition module and data reconstruction module of the data calculation layer to calculate and process the characteristic information of the arrester, and send the relevant data to the state assessment layer;

[0026] Step S3, using the state assessment preliminary review module, multi-factor decision-making module, model training module, state review and prediction module and data transmission module of the state assessment layer to assess the health status of the arrester and predict the future status in combination with the current influencing factors;

[0027] Step S4: Use the data storage unit and database data transmission unit of the terminal database to store the data required for training the model and evaluating the prediction results.

[0028] The calculation and processing of the characteristic information amount of the lightning arrester in step S2 includes:

[0029] Step S21, running the digital filtering module, using the convolution sliding window weighted average algorithm to filter, to improve the smoothness of the signal and reduce the interference of noise;

[0030] Step S22, running the data decomposition module, processing the leakage current using the local mean decomposition method, and decomposing the alternating current signal into signals of different frequency bands;

[0031] Step S23, running the data reconstruction module, using FFT to reconstruct the alternating current signal into a sinusoidal signal with a constant amplitude at different frequencies.

[0032] The step S3 comprises:

[0033] Step S31, the operation status evaluation module performs status evaluation, which is used to evaluate the signals of each frequency band reconstructed in S23, weight the current effect by the content of different frequencies, and then compare the ratio with the current size under normal operation with the judgment threshold, so as to conduct a preliminary evaluation of the health status of the lightning arrester, and locate the preliminary evaluation results into three levels: normal, warning and fault;

[0034] Step S32, running a multi-factor decision-making module, including temperature, humidity, operating time, external pollution index and other factors that are artificially determined to have an impact, pairing each external condition with the preliminary review result of its corresponding time period and establishing a multi-factor decision tree model;

[0035] Step S33, the operation status review and prediction module is used to perform preliminary training and calculation on the established multi-factor decision tree model, and to establish an XGBoost model to achieve fast and high-precision parallel training of the multi-factor decision tree, judge the current status of the equipment, predict the future status of the equipment, diagnose possible equipment failures and provide early warning;

[0036] Step S34, the operating status review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review results under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

[0037] The beneficial effects of the present invention are:

[0038] The health status assessment system of the traction network line arrester proposed in the present invention is applied. After the noise part of the collected arrester leakage current signal is subjected to Savitzky-Golay (SG) filtering, the signal is processed by the local mean decomposition method (LMD) and then combined with FFT to reconstruct the signal and the leakage current signal characteristics are used to establish the XGBoost (eXtreme Gradient Boosting) prediction model. The measurement signal is iteratively trained to realize the multiple evaluation models of temperature, humidity, and pollution index, and the arrester status assessment is realized. The amount of signal calculation required for processing can be greatly reduced, the required characteristic signals can be extracted more quickly, and the response speed of the status assessment system can be improved; a quantitative determination calculation method for the health degree is proposed, which can more directly and effectively judge the health degree and aging type of the arrester, and at the same time greatly improve the status assessment speed at a high-frequency acquisition rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0040] Figure 1 It is a structural schematic diagram of the traction network line arrester health status assessment system of the present invention;

[0041] Figure 2 A schematic diagram of the data computing layer process of an embodiment of the present invention;

[0042] Figure 3 A schematic diagram of a state assessment layer process according to an embodiment of the present invention;

[0043] Figure 4 The figure is a flow chart of the status assessment preliminary review module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The present invention provides a traction network line arrester health status assessment system, which is further described in detail below in conjunction with the accompanying drawings.

[0045] like Figure 1 The embodiment of the present invention shown discloses a traction network line arrester health status assessment system, including: a data acquisition layer, a data calculation layer, a status assessment layer and a terminal database;

[0046] The data acquisition layer includes: a state data acquisition unit and a state data transmission unit; the data acquisition layer sends characteristic information about the operating environment and the working status of the arrester to the data calculation layer;

[0047] The data calculation layer includes: a digital filtering module, a data decomposition module and a data reconstruction module; the data calculation layer calculates and processes the characteristic information of the arrester and sends relevant data to the state evaluation layer;

[0048] The state assessment layer includes: a state assessment preliminary review module, a multi-factor decision module, a model training module, a state review and prediction module, and a data transmission module; the state assessment layer implements the assessment of the health state of the arrester and compares and predicts the future state in combination with the current influencing factors;

[0049] The terminal database includes: a data storage unit and a database data transmission unit; the terminal database implements the storage of data required for the training model and the evaluation prediction results.

[0050] In this embodiment, after the noise part of the collected arrester leakage current signal is SG filtered, the local mean decomposition method LMD is used for processing, and then the signal is reconstructed in combination with FFT and the leakage current signal characteristics are used to establish an XGBoost prediction model. The measurement signal is iteratively trained to realize multiple evaluation models of temperature, humidity, and pollution index, and to realize the state evaluation of the arrester. It can greatly reduce the amount of signal calculation required for processing, extract the required characteristic signal more quickly, and improve the response speed of the state evaluation system; a quantitative determination calculation method for health status is proposed, which can more directly and effectively judge the health status and aging type of the arrester, and greatly improve the state evaluation speed at high-frequency acquisition rate.

[0051] Each layer is explained in detail below.

[0052] In this embodiment, the data collection layer collects and sends to the data calculation layer the characteristic information of the operating environment and working condition of the arrester. The data collection layer includes: a state data collection unit and a state data transmission unit;

[0053] The state data acquisition unit collects device state information based on the edge sensing device; the edge sensing device includes: a leakage current sensor and a temperature and humidity sensor; the leakage current sensor is used to collect leakage current signals; the temperature and humidity sensor is used to collect the temperature and humidity of the device operating environment;

[0054] The state data transmission unit is used to upload the collected real-time state data to the data calculation layer and the state evaluation layer through the data transmission line for processing.

[0055] In this embodiment, the data calculation layer calculates and processes the characteristic information of the lightning arrester and sends the relevant data to the state evaluation layer. Figure 2 As shown, the data calculation layer includes: a digital filtering module, a data decomposition module and a data reconstruction module; the data calculation layer calculates and processes the characteristic information of the lightning arrester and sends relevant data to the state evaluation layer;

[0056] The digital filtering module uses an SG filter to remove abnormalities and reduce noise from the device status data uploaded by the data acquisition layer, and then stores the normal data after noise reduction and transmits it to the data decomposition module;

[0057] The data decomposition module uses the local mean decomposition method to perform modal decomposition on the data transmitted by the digital filter module, decomposes the mixed signal of variable frequency and variable amplitude into signals of fixed frequency bands, removes the interference components with extremely low content, stores the components with higher content and transmits the signal components of each frequency band to the data reconstruction module;

[0058] The data reconstruction module uses fast Fourier decomposition to reconstruct the variable amplitude signal into a sinusoidal signal with constant amplitude and frequency, thereby reducing the difficulty of leakage current calculation, and stores the reconstructed signal and uploads it to the status evaluation layer through the data transmission module to perform arrester status evaluation.

[0059] In this embodiment, the state assessment layer implements the assessment of the health state of the arrester and predicts the future state in combination with the current influencing factors. The state assessment layer includes: a state assessment preliminary review module, a multi-factor decision module, a model training module, a state review and prediction module, and a data transmission module; the state assessment layer implements the assessment of the health state of the arrester and predicts the future state in combination with the current influencing factors.

[0060] The state assessment preliminary review module uses a high- and low-frequency current amplitude weighted comparison method, including: weighting the current effect by the content size of different frequencies, and then calculating the ratio with the current size under normal operation and comparing it with the judgment threshold, so as to conduct a preliminary review of the health status of the lightning arrester and obtain the preliminary review result level; the initial result level includes: normal level, warning level and fault level;

[0061] The multi-factor decision module pairs the external conditions with the preliminary review result levels of the corresponding time periods of the external conditions and establishes a multi-factor decision tree model; the external conditions include: temperature, humidity, running time, external pollution index and pre-set factor information that may cause influence;

[0062] The model training module performs preliminary training and calculation on the established multi-factor decision tree model. By establishing an XGBoost model, the multi-factor decision tree model training is completed quickly and accurately, the current state of the equipment is judged, the future state of the equipment is predicted, and possible equipment failures are diagnosed and warned in advance.

[0063] The state review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review result under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

[0064] The data transmission module includes: a data receiving unit and an evaluation data transmission unit, which are used to upload and send relevant data through a wireless communication transmission method; the wireless communication transmission method includes: LoRa or 5G;

[0065] The data receiving unit is used to receive the device status information of the data acquisition layer, the prediction results of the data reconstruction module and the status review and prediction module;

[0066] The evaluation data transmission unit is used to transmit the reconstruction signal of the data reconstruction module to the state evaluation layer and the terminal database; and to send the updated state evaluation information and prediction results of the state review and prediction module to the model training module.

[0067] In this embodiment, the terminal database implements the storage of the data required for the training model and the evaluation prediction results. The terminal database includes: a data storage unit and a database data transmission unit; the terminal database implements the storage of the data required for the training model and the evaluation prediction results.

[0068] The data storage unit is used to store real-time original status data of the device, historical data and historical models, and newly trained and updated model data.

[0069] By using a traction network line arrester health status assessment system disclosed in the present invention, the arrester leakage current fluctuations caused by traction network load changes can be converted into a constant component that can be monitored, and the arrester status assessment and future prediction can be realized, thereby effectively improving the operation and maintenance efficiency of the equipment and avoiding failures that affect the safety of trains and personnel.

[0070] Another embodiment of the present invention further discloses a lightning arrester health status assessment method based on a traction network line lightning arrester health status assessment system disclosed in the present invention, comprising the following steps:

[0071] Step S1, using the state data acquisition unit of the data acquisition layer to collect device state information; using the state data transmission unit of the data acquisition layer to send the operating environment of the lightning arrester and the characteristic information reflecting the working status to the data calculation layer;

[0072] Step S2, using the digital filtering module, data decomposition module and data reconstruction module of the data calculation layer to calculate and process the characteristic information of the arrester, and send the relevant data to the state assessment layer;

[0073] Step S3, using the state assessment preliminary review module, multi-factor decision-making module, model training module, state review and prediction module and data transmission module of the state assessment layer to assess the health status of the arrester and predict the future status in combination with the current influencing factors;

[0074] Step S4: Use the data storage unit and database data transmission unit of the terminal database to store the data required for training the model and evaluating the prediction results.

[0075] In this embodiment, the arrester health status assessment method uses Figure 1 A traction network line arrester health status assessment system is shown. The traction network line arrester health status assessment system includes: a data acquisition layer, a data calculation layer, a status assessment layer and a terminal database. Its functions and operation process are as follows:

[0076] Use the corresponding sensor equipment and data transmission lines in the data acquisition layer to send the characteristic information of the operating environment and reaction working status of the lightning arrester to the data calculation layer; use the digital filtering module, data decomposition module and data reconstruction module of the data calculation layer to calculate and process the characteristic information of the lightning arrester, and send relevant data to the state assessment layer; use the state assessment preliminary review module, multi-factor decision-making module, model training module, state review and prediction module and data transmission module of the state assessment layer to evaluate the health status of the lightning arrester and compare and predict the future status in combination with the current influencing factors; use the data storage unit and database data transmission unit of the terminal database to store the data required for the training model and the evaluation and prediction results.

[0077] The following is a detailed explanation of each step.

[0078] Step S1, using the state data acquisition unit of the data acquisition layer to collect device state information; using the state data transmission unit of the data acquisition layer to send the operating environment of the lightning arrester and the characteristic information reflecting the working status to the data calculation layer;

[0079] In this embodiment, the data acquisition layer uses temperature and humidity sensors, leakage current sensors and data transmission lines. The data acquisition layer includes a state data acquisition unit and a state data transmission unit; the state data acquisition unit collects device state information based on edge sensing devices such as leakage current sensors and temperature and humidity sensors, wherein the leakage current sensor is used to collect leakage current signals, and the temperature and humidity sensor is used to collect the temperature and humidity of the device operating environment; the state data transmission unit is used to upload the collected real-time state data to the data calculation layer or the state evaluation layer through the transmission line for processing.

[0080] Step S2, using the digital filtering module, data decomposition module and data reconstruction module of the data calculation layer to calculate and process the characteristic information of the arrester, and send the relevant data to the state assessment layer;

[0081] In this embodiment, if Figure 2 As shown, the data calculation layer includes a digital filtering module, a data decomposition module and a data reconstruction module, which are used to calculate and process the characteristic information of the arrester and send relevant data to the state assessment layer. The digital filtering module is run, and the convolution sliding window weighted average algorithm (SG) filtering is used to improve the smoothness of the signal and reduce the interference of noise. The data decomposition module is run, and the local mean decomposition method (LMD) is used to process the leakage current and decompose the alternating current signal into signals of different frequency bands. The data reconstruction module is run, and the FFT is used to realize the signal to reconstruct the alternating current signal into a sinusoidal signal of constant amplitude at different frequencies.

[0082] The calculation and processing of the characteristic information of the lightning arrester comprises:

[0083] Step S21, running the digital filtering module, using the convolution sliding window weighted average algorithm to filter, to improve the smoothness of the signal and reduce the interference of noise;

[0084] Step S22, running the data decomposition module, processing the leakage current using the local mean decomposition method, and decomposing the alternating current signal into signals of different frequency bands;

[0085] Step S23, running the data reconstruction module, using FFT to reconstruct the alternating current signal into a sinusoidal signal with a constant amplitude at different frequencies.

[0086] The data calculation layer includes: a digital filtering module, a data decomposition module and a data reconstruction module; the data calculation layer calculates and processes the characteristic information of the arrester and sends relevant data to the state evaluation layer;

[0087] In this embodiment, the digital filtering module of the data calculation layer is run to remove anomalies and reduce noise on the device status data uploaded by the data acquisition layer, and then the normal data after noise reduction is stored and transmitted to the data decomposition module; the data decomposition module of the data calculation layer is run to perform modal decomposition on the data transmitted by the digital filtering module using the local mean decomposition method, decompose the mixed signal of variable frequency and variable amplitude into a signal of a fixed frequency band, eliminate the interference components with extremely low content, store the components with higher content and transmit the signal components of each frequency band to the data reconstruction module; the data reconstruction module of the data calculation layer is run to calculate and process the characteristic information of the lightning arrester, and send the relevant data to the status evaluation layer.

[0088] In a preferred embodiment, the specific steps of the convolution sliding window weighted average algorithm filtering in step S21 include:

[0089] Step S211, determine the window width and the polynomial order; the window width is a positive odd integer, which represents the number of neighboring points used for fitting; the size of the window width determines the smoothing effect and the computational complexity; the polynomial order is a positive integer, which represents the order of the polynomial used for fitting, wherein the polynomial order is smaller than the window width;

[0090] Step S212, construct a data matrix; assuming the window width is 2M+1 and the polynomial order is N; for a point X[-M, -M+1, ... 0, 1 ... M-1, M] in a signal window, construct a matrix with 2M+1 rows and N+1 width:

[0091]

[0092] In the formula, y is the output value to be solved, x is the measurement point, and a is the parameter to be solved;

[0093] Step S213, determine the fitting parameters A' and B' by the least squares method, express the matrix equation in S212 as Y = XA + B, then the predicted value Y' = XA' = X(X T X) -1 X T Y;

[0094] Step S214, move the pane backwards, and repeat steps S212 and S213 until all points are reached.

[0095] In a preferred embodiment, the local mean decomposition method in step S22 specifically includes:

[0096] Step S221: Let the original signal be x(t), find each local extreme point, and calculate n i and n i+1 The average value of all the average values ​​m i At the corresponding extreme point time tni and t ni+1 The linear extension is performed between the two values, and the extended straight line is smoothed by the sliding average method to obtain the local mean function m 11 (t).

[0097] Step S222: Calculate the local amplitude a i , All local amplitudes ai at the corresponding extreme point t ni and t ni+1 A straight line is extended between the two values, and the extended straight line is smoothed by the sliding average method to obtain the local mean function a 11 (t);

[0098] Step S223: Separate the local mean function m from the original signal 11 (t)

[0099] h 11 (t) = x(t) - m 11 (t)

[0100] Step S224, use a 11 (t) for h 11 (t) is modulated to obtain:

[0101]

[0102] At this point we need to judge s 11 (t) Is it a pure frequency modulation function (the amplitude of a pure frequency modulation function is always 1, and -1 11 (t)<1; if a 12 (t) is s 11 (t) function is the envelope estimation function, then a 12 (t) = 1), if it is not a pure frequency modulation function, return to step (1) and 11 (t) Repeat the above iterative process until a pure FM signal s is obtained. 1n (t), then:

[0103] h 11 (t) = x(t) - m 11 (t)

[0104] h 12 (t) = s 11 (t)-m 12 (t) . .

[0107] h ln t=s 1n-1 tm ln t​

[0108] Step S225: multiply all local envelope functions generated in the entire iterative process to obtain an envelope signal a1(t):

[0109]

[0110] Step S226: Determine that the first PF component of the original signal is the product of the envelope signal a1(t) and the pure frequency modulation signal sin(t):

[0111] PF1(t)=a1(t)s ln (t)

[0112] Step S227: Separate the PF1(t) component from the original signal x(t), and use the obtained new signal u1(t) as a new original signal. Repeat steps S221 to S226 for k cycles until u1(t) is equal to or greater than u1(t). k (t) is a monotonic function.

[0113] u1(t)=x(t)-PF1(t)

[0114] u2(t)=u1(t)-PF2(t) .

[0116] u k (t) = u k-1 (t)-PF k (t)

[0117] In this embodiment, the originally received current signal is decomposed into multiple PF components to obtain i1 to i k There are k components in total, corresponding to the k components of PF.

[0118] Step S3, using the state assessment preliminary review module, multi-factor decision-making module, model training module, state review and prediction module and data transmission module of the state assessment layer to assess the health status of the arrester and predict the future status in combination with the current influencing factors;

[0119] The state assessment layer includes: a state assessment preliminary review module, a multi-factor decision module, a model training module, a state review and prediction module, and a data transmission module; the state assessment layer implements the assessment of the health state of the arrester and compares and predicts the future state in combination with the current influencing factors;

[0120] In this embodiment, if Figure 3The state evaluation layer shown includes: a state evaluation preliminary review module, a multi-factor decision module, a model training module, and a state review and prediction module; the evaluation preliminary review module performs state evaluation, and the multi-factor decision module includes temperature, humidity, operating time, external pollution index, and other factors that are artificially determined to have an impact. Each external condition is paired with the preliminary review result of its corresponding period and a multi-factor decision tree model is established. After that, the state review and prediction module is used to perform preliminary training and calculation on the established multi-factor decision tree model, and by establishing an XGBoost model, the multi-factor decision tree model is trained quickly and accurately to judge the current state of the equipment, predict the future state of the equipment, diagnose possible equipment failures and give early warnings. Finally, the state review and prediction module is used to use the multi-factor decision model trained by the model training module to make a secondary judgment on the current preliminary review results under the influence of the current factors, further improve the evaluation accuracy, and compare with the trained model to complete the prediction of the future state.

[0121] like Figure 4 The status evaluation preliminary review module shown in the figure is running to conduct status evaluation, which is used to evaluate the reconstructed signals of each frequency band, weight the current effect by the content size of different frequencies, and then compare the ratio with the current size under normal operation with the judgment threshold, so as to conduct a preliminary review of the health status of the lightning arrester, and locate the preliminary review results at three levels: normal, warning and fault. The preliminary review results are passed to the multi-factor decision model as training and the first status evaluation data.

[0122] The step S3 specifically includes:

[0123] Step S31, the operation status evaluation module performs status evaluation, which is used to evaluate the signals of each frequency band reconstructed in S23, weight the current effect by the content of different frequencies, and then compare the ratio with the current size under normal operation with the judgment threshold, so as to conduct a preliminary evaluation of the health status of the lightning arrester, and locate the preliminary evaluation results into three levels: normal, warning and fault;

[0124] Step S32, running a multi-factor decision-making module, including temperature, humidity, operating time, external pollution index and other factors that are artificially determined to have an impact, pairing each external condition with the preliminary review result of its corresponding time period and establishing a multi-factor decision tree model;

[0125] Step S33, the operation status review and prediction module is used to perform preliminary training and calculation on the established multi-factor decision tree model, and to establish an XGBoost model to achieve fast and high-precision parallel training of the multi-factor decision tree, judge the current status of the equipment, predict the future status of the equipment, diagnose possible equipment failures and provide early warning;

[0126] Step S34, the operating status review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review results under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

[0127] The data transmission module includes a data receiving unit and an evaluation data transmission unit, which are used to upload and send relevant data through wireless communication transmission methods such as LoRa or 5G;

[0128] The data receiving unit is used to receive the device status information of the data acquisition layer, the prediction results of the data reconstruction module and the status review and prediction module;

[0129] The data transmission module is used to transmit the reconstruction signal of the data reconstruction module to the state evaluation layer and the terminal database or to send the updated state evaluation information and prediction results of the state review and prediction module to the model training module;

[0130] In a preferred embodiment, in the step S31, the characteristic of the current increase when the arrester ages is utilized, and the operating current is compared with the operating current under normal working conditions and the weight is calculated to analyze the health status of the arrester; a health coefficient is set as H, and a quantitative health determination is finally obtained by calculating and analyzing the health coefficient; the calculation method for the quantitative health determination is:

[0131]

[0132] Where, 1≤n≤k and it is an integer, H is the overall health assessment coefficient;

[0133] According to the weighting of each frequency component after the reconstructed signal, the decision condition is obtained:

[0134]

[0135] Where i m It means taking the current amplitude at each frequency and setting the health critical standard to x, where x is set based on the empirical parameters obtained from actual engineering or based on actual conditions:

[0136] When H<x, it means the arrester is in healthy working condition;

[0137] Set an early warning when x<H<1.2x;

[0138] When 1.2x<H<1.5x, it should be considered as serious degradation and in a fault state.

[0139] In a preferred embodiment, the multiple factors in step S32 include temperature, humidity, operating time, external pollution index and other factors that are artificially determined to have an impact. Each external condition is paired with the preliminary review result of its corresponding time period and a multi-factor decision tree model is established.

[0140] In a preferred embodiment, the multi-factor decision tree model established in the step S33 is preliminarily trained and calculated, and by establishing an XGBoost model, the multi-factor decision tree training is completed quickly and accurately, the current state of the equipment is judged, the future state of the equipment is predicted, and possible equipment failures are diagnosed and warned in advance.

[0141] Step S4: Use the data storage unit and database data transmission unit of the terminal database to store the data required for training the model and evaluating the prediction results.

[0142] In this embodiment, the data storage unit in the terminal database is used to store data such as real-time original status data of the device, historical data and historical models, and newly trained and updated models.

Claims

1. A traction network line arrester health status assessment system, characterized in that: include: Data collection layer, data calculation layer, status assessment layer and terminal database; The data acquisition layer includes: a state data acquisition unit and a state data transmission unit; the data acquisition layer sends characteristic information about the operating environment and the working status of the arrester to the data calculation layer; The data calculation layer includes: a digital filtering module, a data decomposition module and a data reconstruction module; the data calculation layer calculates and processes the characteristic information of the arrester and sends relevant data to the state evaluation layer; The state assessment layer includes: a state assessment preliminary review module, a multi-factor decision module, a model training module, a state review and prediction module, and a data transmission module; the state assessment layer implements the assessment of the health state of the arrester and compares and predicts the future state in combination with the current influencing factors; The terminal database includes: a data storage unit and a database data transmission unit; the terminal database implements the storage of data required for the training model and the evaluation prediction results.

2. The traction network line arrester health status assessment system according to claim 1, characterized in that: The state data collection unit collects device state information based on the edge sensing device; The edge sensing device includes: a leakage current sensor and a temperature and humidity sensor; the leakage current sensor is used to collect leakage current signals; the temperature and humidity sensor is used to measure the temperature and humidity of the equipment operating environment; The state data transmission unit is used to upload the collected real-time state data to the data calculation layer and the state evaluation layer through the data transmission line for processing.

3. The traction network line arrester health status assessment system according to claim 1, characterized in that: The digital filtering module uses an SG filter to remove abnormalities and reduce noise from the device status data uploaded by the data acquisition layer, and then stores the normal data after noise reduction and transmits it to the data decomposition module; The data decomposition module uses the local mean decomposition method to perform modal decomposition on the data transmitted by the digital filter module, decomposes the mixed signal with variable frequency and amplitude into signals of fixed frequency bands, removes the interference components with extremely low content, stores the components with higher content and transmits the signal components of each frequency band to the data reconstruction module; The data reconstruction module uses fast Fourier decomposition to reconstruct the variable amplitude signal into a sinusoidal signal with constant amplitude and frequency, thereby reducing the difficulty of leakage current calculation, and stores the reconstructed signal and uploads it to the status evaluation layer through the data transmission module to perform arrester status evaluation.

4. The traction network line arrester health status assessment system according to claim 1, characterized in that: The state assessment preliminary review module uses a high- and low-frequency current amplitude weighted comparison method, including: weighting the current effect by the content size of different frequencies, and then calculating the ratio with the current size under normal operation and comparing it with the judgment threshold, so as to conduct a preliminary review of the health status of the lightning arrester and obtain the preliminary review result level; the initial result level includes: normal level, warning level and fault level; The multi-factor decision module pairs the external conditions with the preliminary review result levels of the corresponding time periods of the external conditions and establishes a multi-factor decision tree model; the external conditions include: temperature, humidity, running time, external pollution index and pre-set factor information that may cause an impact; The model training module performs preliminary training and calculation on the established multi-factor decision tree model. By establishing an XGBoost model, the multi-factor decision tree model training is completed quickly and accurately, the current state of the equipment is judged, the future state of the equipment is predicted, and possible equipment failures are diagnosed and warned in advance. The state review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review result under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

5. The traction network line arrester health status assessment system according to claim 4, characterized in that: The data transmission module includes: a data receiving unit and an evaluation data transmission unit, which are used to upload and send relevant data through a wireless communication transmission method; the wireless communication transmission method includes: LoRa or 5G; The data receiving unit is used to receive the device status information of the data acquisition layer, the prediction results of the data reconstruction module and the status review and prediction module; The evaluation data transmission unit is used to transmit the reconstruction signal of the data reconstruction module to the state evaluation layer and the terminal database; and to send the updated state evaluation information and prediction results of the state review and prediction module to the model training module.

6. The traction network line arrester health status assessment system according to claim 1, characterized in that: The data storage unit is used to store real-time original status data of the device, historical data and historical models, and newly trained and updated model data.

7. A lightning arrester health status assessment method of a traction network line lightning arrester health status assessment system according to any one of claims 1 to 6, characterized in that: The steps include: Step S1, using the state data acquisition unit of the data acquisition layer to collect device state information; using the state data transmission unit of the data acquisition layer to send the operating environment of the lightning arrester and the characteristic information reflecting the working status to the data calculation layer; Step S2, using the digital filtering module, data decomposition module and data reconstruction module of the data calculation layer to calculate and process the characteristic information of the arrester, and send the relevant data to the state assessment layer; Step S3, using the state assessment preliminary review module, multi-factor decision-making module, model training module, state review and prediction module and data transmission module of the state assessment layer to assess the health status of the arrester and predict the future status in combination with the current influencing factors; Step S4: Use the data storage unit and database data transmission unit of the terminal database to store the data required for training the model and evaluating the prediction results.

8. The method for assessing the health status of a lightning arrester according to claim 7, characterized in that: The calculation and processing of the characteristic information amount of the lightning arrester in step S2 includes: Step S21, running the digital filtering module, using the convolution sliding window weighted average algorithm to filter, to improve the smoothness of the signal and reduce the interference of noise; Step S22, running the data decomposition module, processing the leakage current using the local mean decomposition method, and decomposing the alternating current signal into signals of different frequency bands; Step S23, running the data reconstruction module, using FFT to reconstruct the alternating current signal into a sinusoidal signal with a constant amplitude at different frequencies.

9. The method for assessing the health status of a lightning arrester according to claim 7, characterized in that: The step S3 comprises: Step S31, the operation status evaluation module is used to evaluate the status, which is used to evaluate the signals of each frequency band reconstructed in S23, weight the current effect by the content of different frequencies, and then compare the ratio with the current size under normal operation with the judgment threshold, so as to conduct a preliminary evaluation of the health status of the lightning arrester, and locate the preliminary evaluation results into three levels: normal, warning and fault; Step S32, running a multi-factor decision-making module, including temperature, humidity, operating time, external pollution index and other factors that are artificially determined to have an impact, pairing each external condition with the preliminary review result of its corresponding time period and establishing a multi-factor decision tree model; Step S33, the operation status review and prediction module is used to perform preliminary training and calculation on the established multi-factor decision tree model, and to establish an XGBoost model to achieve fast and high-precision parallel training of the multi-factor decision tree, judge the current status of the equipment, predict the future status of the equipment, diagnose possible equipment failures and provide early warning; Step S34, the operating status review and prediction module uses the multi-factor decision-making model trained by the model training module to make a secondary judgment on the initial current review results under the influence of current factors, further improve the evaluation accuracy, and compare it with the trained model to complete the prediction of future status.

Images