A fault distance measurement system for overhead contact network in electrified railway stations

Abstract

The present invention relates to the field of contact network technology, and specifically to a fault distance measurement system for the contact network of an electrified railway station. First, the current data, voltage data, environmental data and historical monitoring data of the contact network monitoring point are obtained; then, the current data and the voltage data are preprocessed and the electrical data is corrected to obtain the final preprocessed electrical data; then, the final preprocessed electrical data and the preprocessed historical monitoring data are sequentially subjected to wavelet transformation and GAF ​​transformation to obtain a first GAF image group and a second GAF ​​image group; feature matching is performed on the first GAF image group and the second GAF ​​image group, and fault type matching information is obtained according to the feature matching value; finally, the final preprocessed electrical data and the first GAF image group are predicted using a fault location prediction model to obtain fault location information.

Description

Technical Field

[0001] The invention relates to the technical field of overhead contact networks, in particular to a fault distance measurement system for overhead contact networks in electrified railway stations. Background Art

[0002] The impedance method is mainly used for fault distance measurement of the overhead contact network in electrified railway stations, that is, fault distance measurement is achieved based on the current and voltage variables of the fault line.

[0003] At present, the existing technology still has shortcomings in the fault distance measurement method of the contact network in the electrified railway station. On the one hand, the existing technology does not match the detection of the contact network fault according to the instantaneous electrical characteristics. On the other hand, the existing technology usually uses the change of electrical data to locate the fault position of the contact network, but in actual application it will be affected by environmental interference, resulting in deviations in the positioning results, which will reduce the reliability and accuracy of the fault distance measurement of the contact network in the electrified railway station.

[0004] Therefore, a fault distance measurement system for overhead contact network in electrified railway stations is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a fault distance measurement system for the contact network of an electrified railway station. First, the current data, voltage data, environmental data and historical monitoring data of the contact network monitoring point are obtained; then, the current data and the voltage data are preprocessed and the electrical data is corrected to obtain the final preprocessed electrical data; then, the final preprocessed electrical data and the preprocessed historical monitoring data are successively subjected to wavelet transformation and GAF ​​transformation to obtain a first GAF image group and a second GAF ​​image group; feature matching is performed on the first GAF image group and the second GAF ​​image group, and fault type matching information is obtained according to the feature matching value; finally, the final preprocessed electrical data and the first GAF image group are predicted using a fault location prediction model to obtain fault location information.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A fault distance measurement system for an electrified railway station overhead contact network, comprising:

[0008] The system control unit is used to control the start, pause and stop of the system;

[0009] The data acquisition unit is used to obtain the current data, voltage data and environmental data of each monitoring point of the contact network; at the same time, it obtains historical monitoring data from the historical database;

[0010] The data processing and correction unit is used to preprocess the current data and the voltage data to obtain preprocessed electrical data; input the preprocessed electrical data and the environmental data into an electrical data correction weight prediction model to obtain electrical data correction weights; and weight the preprocessed electrical data using the electrical data correction weights to obtain final preprocessed electrical data;

[0011] The fault type matching unit is used to process the final preprocessed electrical data and the preprocessed historical monitoring data by combining wavelet transform and GAF ​​transform to obtain a first GAF image group and a second GAF ​​image group; perform feature matching on the first GAF image group and the second GAF ​​image group using a VGG19 network, and obtain fault type matching information according to the feature matching value;

[0012] The fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into the fault prediction model for training to obtain a fault location prediction model; use the fault location prediction model to predict the final pre-processed electrical data and the first GAF image group to obtain fault location information;

[0013] The output prompt unit is used to output the fault location information.

[0014] Furthermore, the historical monitoring data includes: historical electrical data, historical environmental data and historical fault location data; wherein the historical electrical data includes historical current data and historical voltage data; the historical current data, the historical voltage data, the current data and the voltage data are all continuous time series data.

[0015] Furthermore, the data processing and correction unit is used to preprocess the current data and the voltage data, and the specific implementation process of obtaining the preprocessed electrical data includes:

[0016] Acquiring electrical data of a contact network monitoring point; wherein the electrical data includes current data and voltage data;

[0017] Cleaning the electrical data to obtain first electrical data;

[0018] De-noising the first electrical data to obtain second electrical data;

[0019] The second electrical data is normalized to obtain preprocessed electrical data.

[0020] Furthermore, the data processing and correction unit inputs the preprocessed electrical data and the environmental data into an electrical data correction weight prediction model to obtain an electrical data correction weight; and the preprocessed electrical data is weighted using the electrical data correction weight to obtain the final preprocessed electrical data. The specific implementation process includes:

[0021] Acquire pre-processed electrical data and environmental data; wherein the environmental data includes: vibration data, temperature data, humidity data and electric field strength data;

[0022] Constructing an electrical data correction weight prediction model, inputting the preprocessed historical monitoring data into the electrical data correction weight prediction model for training, and obtaining the historical electrical data correction weight;

[0023] Inputting the historical electrical data correction weight, the preprocessed electrical data and the environmental data into the pre-trained electrical data correction weight prediction model for prediction, optimizing the model parameters, and obtaining the electrical data correction weight;

[0024] Wherein, the electrical data correction weights include: vibration influence weight, temperature influence weight, humidity influence weight and electric field strength influence weight;

[0025] Using the electrical data correction weight to perform weighted summation on the normalized environmental data to obtain a final electrical data correction weight;

[0026] The pre-processed electrical data is weighted using the final electrical data correction weight to obtain final pre-processed electrical data.

[0027] Furthermore, the pre-processed electrical data is weighted by using the final electrical data correction weight, and the calculation formula for obtaining the final pre-processed electrical data is:

[0028] ;

[0029] in, Preprocess electrical data for final purpose; Correct weights for final electrical data; To pre-process electrical data; Weight for vibration effects; is the normalized vibration data; is the temperature impact weight; is the normalized temperature data; The weight for humidity effect; is the normalized humidity data; is the influence weight of electric field strength; is the normalized electric field strength data.

[0030] Furthermore, the fault type matching unit is used to process the final preprocessed electrical data and the preprocessed historical monitoring data in combination with wavelet transform and GAF ​​transform to obtain the first GAF image group and the second GAF ​​image group. The specific implementation process includes:

[0031] Obtain final preprocessed electrical data and preprocessed historical electrical data;

[0032] The final pre-processed electrical data and the historical electrical data are processed by wavelet transform to obtain a first wavelet transform group and a second wavelet transform group; wherein the wavelet transform group includes: a high-frequency component data group and a low-frequency component data group;

[0033] The first wavelet transform group and the second wavelet transform group are transformed by using GAF transform to obtain a first GAF image group and a second GAF ​​image group; wherein the GAF image group includes: a GAF high-frequency image group and a GAF low-frequency image group.

[0034] Furthermore, the fault type matching unit uses the VGG19 network to perform feature matching on the first GAF image group and the second GAF ​​image group, and obtains fault type matching information according to the feature matching value:

[0035] Obtain a first GAF image group, a second GAF ​​image group, and a pre-trained VGG19 network model;

[0036] Inputting the first GAF image group and the second GAF ​​image group into the pre-trained VGG19 network model to obtain a first GAF image feature group and a second GAF ​​image feature group;

[0037] Calculating the first GAF image feature group and the second GAF ​​image feature group using a feature matching formula to obtain a feature matching value;

[0038] Fault type matching information is obtained according to the feature matching value and the feature matching threshold.

[0039] Furthermore, the calculation formula of the feature matching formula is:

[0040] ;

[0041] Among them, TZMV is the feature matching value; is the current matching weight; DLMV is the current characteristic matching value; is the voltage matching weight; DYMV is the voltage feature matching value; M is the number of current matching features; is the current high frequency characteristic weight; is the i-th current high-frequency feature in the first GAF image feature group; is the i-th current high-frequency feature in the second GAF ​​image feature group; is the current low-frequency characteristic weight; is the i-th current low-frequency feature in the first GAF image feature group; is the i-th current low-frequency feature in the second GAF ​​image feature group; N is the number of voltage matching features; is the voltage high frequency feature weight; is the jth voltage high-frequency feature in the first GAF image feature group; is the jth voltage high frequency feature in the second GAF ​​image feature group; is the voltage low-frequency feature weight; is the jth voltage low-frequency feature in the first GAF image feature group; is the jth voltage low-frequency feature in the second GAF ​​image feature group.

[0042] Furthermore, the fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into a fault prediction model for training to obtain a fault location prediction model; and the specific implementation process of using the fault location prediction model to predict the final preprocessed electrical data and the first GAF image group to obtain the fault location information includes:

[0043] Acquire final preprocessed electrical data and the corresponding first GAF image group; at the same time, acquire historical monitoring data that meets the fault type matching information and the corresponding second GAF ​​image group;

[0044] Constructing a fault prediction model, taking the historical electrical data in the historical monitoring data and the second GAF ​​image group as model inputs, and the historical fault location data in the historical monitoring data as model output results, training the fault prediction model, and obtaining a fault location prediction model;

[0045] The final pre-processed electrical data and the first GAF image group are input into the fault location prediction model for prediction, and the fault location information is output.

[0046] Furthermore, the output prompt unit outputs the fault location information in the form of: voice, image and video.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] 1. The present invention proposes an electrical data correction function for correcting collected electrical data; the function first cleans, denoises and normalizes the electrical data to obtain pre-processed electrical data; then, the normalized environmental data is weighted and summed using the electrical data correction weight output by the model to obtain the final electrical data correction weight; finally, the pre-processed electrical data is weighted using the final electrical data correction weight to obtain the final pre-processed electrical data; the final electrical data correction weight can accurately reflect the impact of environmental changes on data collection, thereby effectively improving the reliability and accuracy of fault distance measurement of the contact network of the electrified railway station.

[0049] 2. The present invention proposes a fault type matching function for obtaining fault type information of the contact network; the function first performs wavelet transform and GAF ​​transform on the final preprocessed electrical data and the historical electrical data in turn to obtain a first GAF image group and a second GAF ​​image group; then, feature matching calculation is performed on the first GAF image feature group and the second GAF ​​image feature group extracted by the VGG19 network, and the fault type matching information is obtained according to the calculated feature matching value; the function combines wavelet transform, GAF transform and feature matching, thereby effectively improving the reliability and accuracy of fault ranging of the contact network in the electrified railway station.

[0050] 3. The present invention proposes a fault location positioning function for obtaining the fault location information of the contact network; this function trains the fault prediction model by using historical electrical data and the second GAF ​​image group to obtain the fault location prediction model; then, finally, the fault location prediction model is used to predict the final preprocessed electrical data and the first GAF image group, and the local and global changes of the electrical data are combined to improve the accuracy of the model in predicting the fault location, thereby effectively improving the reliability and accuracy of the fault distance measurement of the contact network in the electrified railway station. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a structural schematic diagram of a fault distance measurement system for an electrified railway station overhead contact network according to the present invention;

[0052] Figure 2 It is a structural schematic diagram of the electrical data correction weight prediction model of the present invention;

[0053] Figure 3 It is a structural schematic diagram of the fault location prediction model of the present invention;

[0054] Figure 4 The present invention is a schematic flow chart of a fault distance measurement system for an electrified railway station overhead contact network. DETAILED DESCRIPTION

[0055] 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.

[0056] The overhead contact network is a key facility in the electrified railway system, and its main function is to provide power for electric locomotives. It is usually erected in a zigzag manner above the rails to provide a stable current source for the pantograph. As an important part of the electrified railway, the overhead contact network not only carries the task of power transmission, but also ensures the smooth power supply of the electric locomotive during its travel.

[0057] The number of contact networks is large and the structure is relatively complex. It is usually composed of multiple components, including contact suspension systems, support devices, positioning devices and pillars. Therefore, when a fault occurs in the contact network, it is crucial to perform fault distance measurement on the contact network in electrified railway stations in order to shorten the troubleshooting time, quickly repair the fault problem, and minimize the long-term impact on train operation.

[0058] At present, the existing technology still has shortcomings in the fault distance measurement method of the contact network in the electrified railway station. On the one hand, the existing technology does not match the detection of the contact network fault according to the instantaneous electrical characteristics. On the other hand, the existing technology usually uses the change of electrical data to locate the fault position of the contact network, but in actual application it will be affected by environmental interference, resulting in deviations in the positioning results, which will reduce the reliability and accuracy of the fault distance measurement of the contact network in the electrified railway station.

[0059] Embodiment 1

[0060] As an embodiment of the present invention, refer to Figure 1 , a fault distance measurement system for an electrified railway station contact network, comprising: a system control unit, a data acquisition unit, a data processing and correction unit, a fault type matching unit, a fault position positioning unit and an output prompt unit;

[0061] The system control unit is used to control the start, pause and stop of the system;

[0062] The data acquisition unit is used to obtain current data, voltage data and environmental data of each monitoring point of the contact network; at the same time, it obtains historical monitoring data from the historical database;

[0063] In this embodiment, the current data, voltage data and environmental data are collected by current sensors, voltage sensors and environmental sensors respectively. Table 1 is the data collection result at a monitoring point of the overhead contact network.

[0064] Table 1. Data collection results at a monitoring point on the contact network

[0065]

[0066] Furthermore, the historical monitoring data includes: historical electrical data, historical environmental data and historical fault location data; wherein the historical electrical data includes historical current data and historical voltage data; the historical current data, the historical voltage data, the current data and the voltage data are all continuous time series data.

[0067] In this embodiment, historical monitoring data is used for subsequent model training and feature matching of fault types; historical electrical data is continuous time series data to improve the model's ability to accurately identify fault modes and overall environmental impact changes, thereby reliably and accurately measuring fault distances for electrified railway station contact networks.

[0068] The data processing and correction unit is used to preprocess the current data and the voltage data to obtain preprocessed electrical data; input the preprocessed electrical data and the environmental data into the electrical data correction weight prediction model to obtain the electrical data correction weight; and weight the preprocessed electrical data using the electrical data correction weight to obtain the final preprocessed electrical data;

[0069] Furthermore, the data processing and correction unit is used to preprocess the current data and the voltage data, and the specific implementation process of obtaining the preprocessed electrical data includes:

[0070] Acquiring electrical data of a contact network monitoring point; wherein the electrical data includes current data and voltage data;

[0071] Cleaning the electrical data to obtain first electrical data;

[0072] De-noising the first electrical data to obtain second electrical data;

[0073] The second electrical data is normalized to obtain preprocessed electrical data.

[0074] The preprocessing of electrical data in this embodiment includes data cleaning, data denoising and data normalization; data cleaning is used to eliminate erroneous data and blank data in the electrical data; data denoising is used to remove noise interference signals by using a filtering method; data normalization is used to reduce the scale of the data to improve the efficiency of model training, so that the preprocessing process can effectively improve the reliability and accuracy of fault distance measurement of the contact network of the electrified railway station.

[0075] Furthermore, the data processing and correction unit inputs the preprocessed electrical data and the environmental data into an electrical data correction weight prediction model to obtain an electrical data correction weight; and the preprocessed electrical data is weighted using the electrical data correction weight to obtain the final preprocessed electrical data. The specific implementation process includes:

[0076] Acquire pre-processed electrical data and environmental data; wherein the environmental data includes: vibration data, temperature data, humidity data and electric field strength data;

[0077] Constructing an electrical data correction weight prediction model, inputting the preprocessed historical monitoring data into the electrical data correction weight prediction model for training, and obtaining the historical electrical data correction weight;

[0078] The structure of the electrical data correction weight prediction model in this embodiment can be referred to Figure 2 , including: input layer, feature preprocessing layer, feature classification layer, feature attention layer and prediction output layer; the input layer is used to transform the historical monitoring data from the data space to the feature space to obtain the initial features; the feature preprocessing layer is used to perform 5 residual convolution operations on the initial features to obtain deep convolution features; the feature classification layer uses a fully connected layer to classify environmental features according to type to obtain multi-category features; the feature attention enhancement layer uses the self-attention mechanism to obtain the influence weights between environmental features and the influence weights of environmental features on electrical features; the prediction output layer combines the influence weights and the fully connected layer to output the correction weights of historical electrical data.

[0079] Inputting the historical electrical data correction weight, the preprocessed electrical data and the environmental data into the pre-trained electrical data correction weight prediction model for prediction, optimizing the model parameters, and obtaining the electrical data correction weight;

[0080] The electrical data correction weights include: vibration influence weight, temperature influence weight, humidity influence weight and electric field strength influence weight;

[0081] Using the electrical data correction weight to perform weighted summation on the normalized environmental data to obtain a final electrical data correction weight;

[0082] The pre-processed electrical data is weighted using the final electrical data correction weight to obtain final pre-processed electrical data.

[0083] In this embodiment, an electrical data correction function is proposed for correcting the collected electrical data; the function first cleans, denoises and normalizes the electrical data to obtain pre-processed electrical data; then, the normalized environmental data is weighted and summed using the electrical data correction weight output by the model to obtain the final electrical data correction weight; finally, the pre-processed electrical data is weighted using the final electrical data correction weight to obtain the final pre-processed electrical data; the final electrical data correction weight can accurately reflect the impact of environmental changes on data collection, thereby effectively improving the reliability and accuracy of fault distance measurement of the contact network of the electrified railway station.

[0084] Furthermore, the pre-processed electrical data is weighted by using the final electrical data correction weight, and the calculation formula for obtaining the final pre-processed electrical data is:

[0085] ;

[0086] in, Preprocess electrical data for final purpose; Correct weights for final electrical data; To pre-process electrical data; Weight for vibration effects; is the normalized vibration data; is the temperature impact weight; is the normalized temperature data; The weight for humidity effect; is the normalized humidity data; is the influence weight of electric field strength; is the normalized electric field strength data.

[0087] In this embodiment, an electrical data correction weight formula is proposed for correcting the pre-processed electrical data; the formula multiplies the pre-processed electrical data by the final electrical data correction weight, and the final electrical data correction weight is obtained by weighted summation of the normalized environmental data by the influence weight of the model output; the final pre-processed electrical data calculated by this formula can more accurately reflect the data actually collected on the contact network, thereby improving the reliability and accuracy of fault distance measurement of the contact network of the electrified railway station.

[0088] In order to conveniently illustrate the electrical data correction function proposed by the present invention, this embodiment selects three groups of test data for testing, which are recorded as test one, test two and test three; each group of data includes electrical data and environmental data, and is obtained by continuous collection on the same contact network, under the same environment and at different monitoring points; each group of electrical data is subjected to data cleaning, data denoising and data normalization processing to obtain pre-processed electrical data of each sample; wherein the pre-processed electrical data includes multiple groups of pre-processed voltage data and pre-processed current data;

[0089] Input the historical electrical data correction weight, pre-processed electrical data and environmental data into the pre-trained electrical data correction weight prediction model for prediction to obtain the electrical data correction weight; wherein the historical electrical data correction weight is obtained by inputting the pre-processed historical monitoring data into the output of the electrical data correction weight prediction model; the electrical data correction weight is used to perform weighted summation on the normalized environmental data to obtain the final electrical data correction weight; the final electrical data correction weight is used to weight the pre-processed electrical data to obtain the final pre-processed electrical data of each sample;

[0090] In order to facilitate subsequent display, a group of preprocessed voltage data and preprocessed current data in the preprocessed electrical data are randomly selected as preprocessed electrical reference data for processing, and the final electrical data reference correction weight and the final preprocessed electrical reference data are obtained; wherein, the left side of the final electrical data reference correction weight is the final voltage data reference correction weight, and the right side is the final current data reference correction weight; the left side of the preprocessed electrical reference data is the preprocessed voltage reference data, and the right side is the preprocessed current reference data; the left side of the final preprocessed electrical reference data is the final preprocessed voltage reference data, and the right side is the final preprocessed current reference data; the electrical data correction test results are shown in Table 2:

[0091] Table 2. Electrical data correction test results

[0092]

[0093] The fault type matching unit is used to process the final preprocessed electrical data and the preprocessed historical monitoring data in combination with wavelet transform and GAF ​​transform to obtain a first GAF image group and a second GAF ​​image group; perform feature matching on the first GAF image group and the second GAF ​​image group using a VGG19 network, and obtain fault type matching information according to the feature matching value;

[0094] Furthermore, the fault type matching unit is used to process the final preprocessed electrical data and the preprocessed historical monitoring data in combination with wavelet transform and GAF ​​transform to obtain the first GAF image group and the second GAF ​​image group. The specific implementation process includes:

[0095] Obtain final preprocessed electrical data and preprocessed historical electrical data;

[0096] The final pre-processed electrical data and the historical electrical data are processed by wavelet transform to obtain a first wavelet transform group and a second wavelet transform group; wherein the wavelet transform group includes: a high-frequency component data group and a low-frequency component data group;

[0097] The first wavelet transform group and the second wavelet transform group are transformed by using GAF transform to obtain a first GAF image group and a second GAF ​​image group; wherein the GAF image group includes: a GAF high-frequency image group and a GAF low-frequency image group.

[0098] In this embodiment, the wavelet transform can capture the instantaneous change characteristics in the electrical data through multi-level decomposition, and the GAF transform can transform the regular information in the time series electrical data into spatial information in the image, which helps the model to analyze its fault mode, thereby improving the reliability and accuracy of fault distance measurement of the contact network in the electrified railway station.

[0099] Furthermore, the fault type matching unit uses the VGG19 network to perform feature matching on the first GAF image group and the second GAF ​​image group, and obtains fault type matching information according to the feature matching value:

[0100] Obtain a first GAF image group, a second GAF ​​image group, and a pre-trained VGG19 network model;

[0101] Inputting the first GAF image group and the second GAF ​​image group into the pre-trained VGG19 network model to obtain a first GAF image feature group and a second GAF ​​image feature group;

[0102] Calculating the first GAF image feature group and the second GAF ​​image feature group using a feature matching formula to obtain a feature matching value;

[0103] Fault type matching information is obtained according to the feature matching value and the feature matching threshold.

[0104] In this embodiment, a fault type matching function is proposed to obtain fault type information of the contact network; the function first performs wavelet transform and GAF ​​transform on the final preprocessed electrical data and the historical electrical data in turn to obtain a first GAF image group and a second GAF ​​image group; then, feature matching calculation is performed on the first GAF image feature group and the second GAF ​​image feature group extracted by the VGG19 network, and the fault type matching information is obtained according to the calculated feature matching value; the function combines wavelet transform, GAF transform and feature matching, thereby effectively improving the reliability and accuracy of fault distance measurement of the contact network of the electrified railway station.

[0105] Furthermore, the calculation formula of the feature matching formula is:

[0106] ;

[0107] Among them, TZMV is the feature matching value; is the current matching weight; DLMV is the current characteristic matching value; is the voltage matching weight; DYMV is the voltage feature matching value; M is the number of current matching features; is the current high frequency characteristic weight; is the i-th current high-frequency feature in the first GAF image feature group; is the i-th current high-frequency feature in the second GAF ​​image feature group; is the current low-frequency characteristic weight; is the i-th current low-frequency feature in the first GAF image feature group; is the i-th current low-frequency feature in the second GAF ​​image feature group; N is the number of voltage matching features; is the voltage high frequency feature weight; is the jth voltage high-frequency feature in the first GAF image feature group; is the jth voltage high frequency feature in the second GAF ​​image feature group; is the voltage low-frequency feature weight; is the jth voltage low-frequency feature in the first GAF image feature group; is the jth voltage low-frequency feature in the second GAF ​​image feature group.

[0108] In this embodiment, the current matching weight and the voltage matching weight are both set to 0.5; the current high-frequency feature weight and the voltage high-frequency feature weight are set to 0.6; the current low-frequency feature weight and the voltage low-frequency feature weight are set to 0.4; of course, the setting of the weight coefficient is not unique and can be adjusted according to actual conditions.

[0109] In this embodiment, a feature matching formula is proposed for feature matching of electrical data; the formula obtains a feature matching value by calculating the absolute error between the electrical features extracted by the VGG19 network model and the historical electrical features; the feature matching value fully reflects the overall fault laws and trend differences between the two, which helps to improve the reliability and accuracy of fault distance measurement of the contact network in electrified railway stations.

[0110] In order to conveniently illustrate the electrical feature matching function proposed in the present invention, this embodiment selects three groups of electrical monitoring data for testing, which are recorded as test sample one, test sample two and test sample three; each group of data is obtained by continuous collection on the same contact network, under the same environment and at different monitoring points; each group of electrical data is preprocessed, wavelet transformed and GAF ​​transformed to obtain the first GAF image group of each sample; then, the first GAF image group after the historical electrical data processing and the first GAF image group of each sample are input into the VGG19 model to obtain the first GAF image feature group and the second GAF ​​image feature group, and the feature matching threshold is set to 0.95; then, combined with the calculation process of the feature matching formula, the current feature matching value, voltage feature matching value and feature matching result of each sample are obtained, and the electrical feature matching test results are shown in Table 3:

[0111] Table 3. Electrical characteristics matching test results

[0112]

[0113] The fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into the fault prediction model for training to obtain a fault location prediction model; use the fault location prediction model to predict the final pre-processed electrical data and the first GAF image group to obtain fault location information;

[0114] Furthermore, the fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into a fault prediction model for training to obtain a fault location prediction model; and the specific implementation process of using the fault location prediction model to predict the final preprocessed electrical data and the first GAF image group to obtain the fault location information includes:

[0115] Acquire final preprocessed electrical data and the corresponding first GAF image group; at the same time, acquire historical monitoring data that meets the fault type matching information and the corresponding second GAF ​​image group;

[0116] Constructing a fault prediction model, taking the historical electrical data in the historical monitoring data and the second GAF ​​image group as model inputs, and the historical fault location data in the historical monitoring data as model output results, training the fault prediction model, and obtaining a fault location prediction model;

[0117] The final pre-processed electrical data and the first GAF image group are input into the fault location prediction model for prediction, and the fault location information is output.

[0118] The structure of the fault location prediction model in this embodiment is as follows: Figure 3 As shown, it includes: 2 input layers, 2 feature extraction layers, 2 feature fusion layers, 3 ConvLSTM layers, 3 SwimTransformer layers and 1 fully connected output layer; wherein the input layer converts the final preprocessed electrical data and the first GAF image group into the feature space for further processing; the feature extraction layer is used to extract deep features to enhance the feature expression capability; the ConvLSTM layer and the SwimTransformer layer are respectively in different branches, each for obtaining time series features and global features; the feature fusion layer is used for fusing features; the fully connected output layer is used to output fault location information.

[0119] In this embodiment, a fault location positioning function is proposed for obtaining the fault location information of the contact network; this function trains the fault prediction model by using historical electrical data and the second GAF ​​image group to obtain the fault location prediction model; then, finally, the fault location prediction model is used to predict the final preprocessed electrical data and the first GAF image group, and the local and global changes of the electrical data are combined to improve the accuracy of the model in predicting the fault location, thereby effectively improving the reliability and accuracy of the fault distance measurement of the contact network in the electrified railway station.

[0120] The output prompt unit is used to output the fault location information.

[0121] Furthermore, the output prompt unit outputs the fault location information in the form of: voice, image and video.

[0122] In this embodiment, the output prompt unit provides timely feedback to the staff and informs them of the fault location information in the form of voice, images and videos; these forms are relatively intuitive and easy to understand, which can help to quickly handle contact network fault problems, ensure the stability of the railway power system, and further improve the reliability and accuracy of fault distance measurement of the contact network in electrified railway stations.

[0123] This embodiment proposes a fault distance measurement system for the contact network of an electrified railway station. First, the current data, voltage data, environmental data and historical monitoring data of the contact network monitoring point are obtained; then, the current data and the voltage data are preprocessed and the electrical data is corrected to obtain the final preprocessed electrical data; then, the final preprocessed electrical data and the preprocessed historical monitoring data are successively subjected to wavelet transformation and GAF ​​transformation to obtain a first GAF image group and a second GAF ​​image group; feature matching is performed on the first GAF image group and the second GAF ​​image group, and fault type matching information is obtained according to the feature matching value; finally, the final preprocessed electrical data and the first GAF image group are predicted using a fault location prediction model to obtain fault location information; this system combines fault type matching with fault location positioning, and can effectively improve the reliability and accuracy of fault distance measurement for the contact network of an electrified railway station.

[0124] Embodiment 2

[0125] An electrified railway needs to perform fault distance measurement on the contact network of the station yard, and applies a fault distance measurement system for the contact network of the electrified railway station yard provided by the present invention. Figure 4 The process of the system in practical application is described as follows:

[0126] First, the data acquisition unit in the system uses sensors to obtain electrical data from each monitoring point of the contact network. In addition to obtaining electrical data, it also includes environmental data and historical monitoring data; environmental data includes: vibration data, temperature data, humidity data and electric field strength data; historical monitoring data is directly obtained through the railway background system to ensure the comprehensive coverage of monitoring data. These data provide a reliable data basis for subsequent data correction, data matching and data prediction.

[0127] Then, after acquiring the sensor data, the system will input the electrical data into the data processing and correction unit for preprocessing and correction; the data processing and correction unit will first perform preprocessing operations on the electrical data, and after completing data cleaning, denoising and normalization operations, the unit will input the preprocessed electrical data into the electrical data correction weight prediction model to obtain the electrical data correction weight; the electrical data correction weight will be used to perform electrical data correction on the preprocessed electrical data to obtain the final preprocessed electrical data.

[0128] Then, the system will call the fault type matching unit to match the final pre-processed electrical data; the unit first uses wavelet transform and GAF ​​transform to process the final pre-processed electrical data in turn to obtain the first GAF image group; the unit uses the pre-trained VGG19 network to perform feature matching on the first GAF image group and the second GAF ​​image group obtained by processing the current historical monitoring data to obtain the current feature matching value; if the current feature matching value does not exceed the pre-set feature matching threshold, the next batch of feature matching is performed; otherwise, the historical monitoring data and the second GAF ​​image group that meet the feature matching conditions are fed back to the fault location unit as training sets, and the fault type information is fed back to the output prompt unit in combination with the historical monitoring data.

[0129] Finally, the system will call the fault location unit to predict the contact network fault location; the unit first receives the training set fed back by the fault type matching unit in advance, and uses the training set to train the fault location prediction model; after that, the unit will input the final preprocessed electrical data and its first GAF image group into the pre-trained fault location prediction model for prediction, obtain the fault location information, and feed it back to the output prompt unit.

[0130] The system will receive information data fed back by the fault type matching unit and the fault location positioning unit by calling the output prompt unit, and provide timely feedback to the staff and inform them of the fault location information in the form of voice, image and video, which will help to quickly deal with contact network faults, ensure the stability of the railway power system, and further improve the reliability and accuracy of fault ranging of the contact network in electrified railway stations.

[0131] The system analyzes electrical data and environmental data through the prediction model, and realizes accurate data correction of pre-processed electrical data. At the same time, combined with wavelet transform, GAF transform and VGG feature matching, it realizes efficient matching of fault types of contact network; finally, the system uses the fault location prediction model trained by electrical data and GAF ​​image data to obtain accurate location information; its efficient data processing and accurate prediction capabilities provide a powerful support tool for fault distance measurement of contact network in electrified railway stations. The application of the system improves the reliability and accuracy of fault distance measurement of contact network in electrified railway stations.

[0132] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A fault distance measurement system for an electrified railway station overhead contact network, characterized in that: include: The system control unit is used to control the start, pause and stop of the system; The data acquisition unit is used to obtain the current data, voltage data and environmental data of each monitoring point of the contact network; at the same time, it obtains historical monitoring data from the historical database; The data processing and correction unit is used to preprocess the current data and the voltage data to obtain preprocessed electrical data; Inputting the preprocessed electrical data and the environmental data into an electrical data correction weight prediction model to obtain an electrical data correction weight; weighting the preprocessed electrical data using the electrical data correction weight to obtain final preprocessed electrical data; The fault type matching unit is used to process the final pre-processed electrical data and the pre-processed historical monitoring data in combination with wavelet transform and GAF ​​transform to obtain a first GAF image group and a second GAF ​​image group; Using the VGG19 network to perform feature matching on the first GAF image group and the second GAF ​​image group, and obtaining fault type matching information according to the feature matching values; The fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into the fault prediction model for training to obtain a fault location prediction model; use the fault location prediction model to predict the final pre-processed electrical data and the first GAF image group to obtain fault location information; The output prompt unit is used to output the fault location information.

2. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The historical monitoring data includes: historical electrical data, historical environmental data and historical fault location data; wherein the historical electrical data includes historical current data and historical voltage data; the historical current data, the historical voltage data, the current data and the voltage data are all continuous time series data.

3. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The data processing and correction unit is used to preprocess the current data and the voltage data to obtain the preprocessed electrical data, and the specific implementation process includes: Acquiring electrical data of a contact network monitoring point; wherein the electrical data includes current data and voltage data; Cleaning the electrical data to obtain first electrical data; De-noising the first electrical data to obtain second electrical data; The second electrical data is normalized to obtain preprocessed electrical data.

4. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The data processing and correction unit inputs the pre-processed electrical data and the environmental data into an electrical data correction weight prediction model to obtain an electrical data correction weight; The specific implementation process of weighting the pre-processed electrical data by using the electrical data correction weight to obtain the final pre-processed electrical data includes: Acquire pre-processed electrical data and environmental data; wherein the environmental data includes: vibration data, temperature data, humidity data and electric field strength data; Constructing an electrical data correction weight prediction model, inputting the preprocessed historical monitoring data into the electrical data correction weight prediction model for training, and obtaining the historical electrical data correction weight; Inputting the historical electrical data correction weight, the preprocessed electrical data and the environmental data into the pre-trained electrical data correction weight prediction model for prediction, optimizing the model parameters, and obtaining the electrical data correction weight; Wherein, the electrical data correction weights include: vibration influence weight, temperature influence weight, humidity influence weight and electric field strength influence weight; Using the electrical data correction weight to perform weighted summation on the normalized environmental data to obtain a final electrical data correction weight; The pre-processed electrical data is weighted using the final electrical data correction weight to obtain final pre-processed electrical data.

5. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 4, characterized in that: The final electrical data correction weight is used to weight the pre-processed electrical data, and the calculation formula for obtaining the final pre-processed electrical data is: ； in, Preprocess electrical data for final purpose; Correct weights for final electrical data; To pre-process electrical data; Weight for vibration effects; is the normalized vibration data; is the temperature impact weight; is the normalized temperature data; The weight for humidity effect; is the normalized humidity data; is the influence weight of electric field strength; is the normalized electric field strength data.

6. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The specific implementation process of the fault type matching unit for processing the final preprocessed electrical data and the preprocessed historical monitoring data in combination with wavelet transform and GAF ​​transform to obtain the first GAF image group and the second GAF ​​image group includes: Obtain final preprocessed electrical data and preprocessed historical electrical data; The final pre-processed electrical data and the historical electrical data are processed by wavelet transform to obtain a first wavelet transform group and a second wavelet transform group; wherein the wavelet transform group includes: a high-frequency component data group and a low-frequency component data group; The first wavelet transform group and the second wavelet transform group are transformed by using GAF transform to obtain a first GAF image group and a second GAF ​​image group; wherein the GAF image group includes: a GAF high-frequency image group and a GAF low-frequency image group.

7. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The fault type matching unit uses the VGG19 network to perform feature matching on the first GAF image group and the second GAF ​​image group, and obtains fault type matching information according to the feature matching value: Obtain a first GAF image group, a second GAF ​​image group, and a pre-trained VGG19 network model; Inputting the first GAF image group and the second GAF ​​image group into the pre-trained VGG19 network model to obtain a first GAF image feature group and a second GAF ​​image feature group; Calculating the first GAF image feature group and the second GAF ​​image feature group using a feature matching formula to obtain a feature matching value; Fault type matching information is obtained according to the feature matching value and the feature matching threshold.

8. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 7, characterized in that: The calculation formula of the feature matching formula is: ； Among them, TZMV is the feature matching value; is the current matching weight; DLMV is the current characteristic matching value; is the voltage matching weight; DYMV is the voltage feature matching value; M is the number of current matching features; is the current high frequency characteristic weight; is the i-th current high-frequency feature in the first GAF image feature group; is the i-th current high-frequency feature in the second GAF ​​image feature group; is the current low-frequency characteristic weight; is the i-th current low-frequency feature in the first GAF image feature group; is the i-th current low-frequency feature in the second GAF ​​image feature group; N is the number of voltage matching features; is the voltage high frequency feature weight; is the jth voltage high-frequency feature in the first GAF image feature group; is the jth voltage high frequency feature in the second GAF ​​image feature group; is the voltage low-frequency feature weight; is the jth voltage low-frequency feature in the first GAF image feature group; is the jth voltage low-frequency feature in the second GAF ​​image feature group.

9. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The fault location positioning unit is used to input the historical monitoring data and the second GAF ​​image group that meet the fault type matching information into the fault prediction model for training to obtain the fault location prediction model; and the specific implementation process of using the fault location prediction model to predict the final pre-processed electrical data and the first GAF image group to obtain the fault location information includes: Acquire final preprocessed electrical data and the corresponding first GAF image group; at the same time, acquire historical monitoring data that meets the fault type matching information and the corresponding second GAF ​​image group; Constructing a fault prediction model, taking the historical electrical data in the historical monitoring data and the second GAF ​​image group as model inputs, and the historical fault location data in the historical monitoring data as model output results, training the fault prediction model, and obtaining a fault location prediction model; The final preprocessed electrical data and the first GAF image group are input into the fault location prediction model for prediction, and the fault location information is output.

10. The fault distance measurement system for the overhead contact network of an electrified railway station according to claim 1, characterized in that: The output prompt unit outputs the fault location information in the form of: voice, image and video.

Images