Method for compensating for missing data in fault location of fully parallel AT power supply for electrified railways

Abstract

This application relates to the technical field of information compensation for fault location in electrified railways, and specifically relates to a method for compensating missing data in fault location of all-parallel AT power supply for electrified railways. The method includes: fitting the missing data based on the current balance equation and historical data, analyzing the difference between the fitted value of the missing data and the non-missing data, and calculating the dynamic fitting confidence of each substation and kiosk; constructing a fitting rationality factor based on the Bayesian network combined with historical data and real-time data, and combining the dynamic fitting confidence; dynamically adjusting the fitted value of the missing current data of the substation and kiosk according to the fitting rationality factor; performing fault location through the dynamically adjusted fitted value and the non-missing current data, improving the stability and reliability of the fault location, being able to adapt to various fault types and complex system topologies, avoiding the problem that the deviation superposition of the fitted data affects the fault location accuracy in the scenario of missing data in multiple substations and kiosks, and ensuring the safe and stable operation of electrified railways.

Description

Technical Field

[0001] This application relates to the technical field of information compensation for fault location in electrified railways, and specifically to a method for compensating for missing data in fault location of fully parallel AT power supply in electrified railways. Background Art

[0002] With the development needs of high-speed railways, the train speed has increased and the train operation density has increased. The power supply system needs to have high capacity and high reliability. The fully parallel AT power supply method is widely used because of its advantages such as large transmission power, low voltage loss, and long power supply distance. However, its complex power supply structure also increases the failure rate. During the development process, researchers have proposed various ranging principles, which basically meet the engineering needs, but are also affected by many factors. The significance of fault location technology is that by quickly and accurately judging the fault type and calibrating the fault point location, the fault can be solved in time, the power supply can be restored, and the safe and stable operation of the electrified railway can be guaranteed, which plays an important role in improving the railway transportation efficiency and safety.

[0003] However, in the scenario of missing data in multiple substations, the missing data of each substation will introduce a certain deviation, and the deviations in the scenario of multiple substations will be superimposed on each other, significantly amplifying the overall error and affecting the ranging accuracy. Summary of the Invention

[0004] In order to solve the above technical problems, this application provides a method for compensating for missing data in fault location of fully parallel AT power supply in electrified railways to solve the existing problems.

[0005] The method for compensating for missing data in fault location of fully parallel AT power supply in electrified railways of this application adopts the following technical solutions:

[0006] An embodiment of this application provides a method for compensating for missing data in fault location of fully parallel AT power supply in electrified railways, and this method includes the following steps:

[0007] Collect various current data on the same side of each substation in real time, including T-line current, F-line current, and suction current; divide all current data into the most recently collected data and historical data according to the proximity of the collection time;

[0008] Obtain the fitting values of the missing data of each substation by fitting the missing data of each substation with the non-missing data of each substation; calculate the dynamic fitting confidence of each substation based on the difference between the fitting value of the missing data and the non-missing data in the most recently collected data of each substation, and the difference between the fitting values of the missing data between each substation.

[0009] Construct and train a probabilistic graphical model based on all historical data of all substations; based on the most recently collected data of all substations, combined with the trained probabilistic graphical model, calculate the posterior probability of each current type for each substation, as well as the posterior joint probability and prior joint probability of each substation; based on the posterior joint probability and the prior joint probability, combined with the dynamic fitting confidence of each substation, calculate the fitting rationality factor of the missing data of each substation;

[0010] Based on the fitting rationality factor and the posterior probability of each current type for each substation, dynamically adjust the fitting values of the missing data of each substation to obtain the final fitting values of each missing data;

[0011] Perform fault location based on the final fitting values of the missing data of each substation and the non-missing data.

[0012] In one embodiment, the method for obtaining the fitting values of each missing data is as follows:

[0013] For each substation, if the number of missing data among the three current data at any moment of the substation is 1, solve the value of the missing data at the any moment through the current balance equation as its fitting value; if there are two or more missing data among the three current data at any moment of the substation, use the mean value of the same type of current data at all moments before the any moment as the fitting value of the corresponding missing data at the any moment.

[0014] In one embodiment, the process for obtaining the dynamic fitting confidence of each substation is as follows:

[0015] Calculate the fitting proportion of the missing current of each substation based on the difference between the fitting values of the missing data of each substation and those of other substations;

[0016] Denote the mean value of all types of current data in the most recently collected data of the x-th substation as the first mean value; denote the fitting value of the p-th missing data in the most recently collected data of the x-th substation as , and denote the difference between and the first mean value as ; denote the dynamic fitting confidence of the x-th substation as , The expression of is: , where is the Sigmoid function; is the number of missing data in the most recently collected data of the x-th substation;

[0017] In one embodiment, the expression of the fitting proportion of the missing current is:

[0018] , where N is the number of substations; is the number of missing data in the most recent collected data of the nth substation; is the fitted value of the mth missing data in the most recent collected data of the nth substation.

[0019] In one embodiment, the process of constructing and training the probability graphical model is as follows:

[0020] Taking various fault types as root nodes, taking various currents of each substation as intermediate nodes, and taking the fitted values of the missing data of each substation as leaf nodes, a Bayesian network model is constructed; the Bayesian network model is trained with the historical data of all substations.

[0021] In one embodiment, the process of obtaining the posterior probability of each current of each substation, as well as the posterior joint probability and prior joint probability of each substation is as follows:

[0022] Taking the most recent collected data of all substations as the input of the trained Bayesian network model, the posterior probability of each current of each substation is output respectively, and the posterior probability and prior probability of the joint of all current data of each substation are calculated through the Bayesian formula, which are respectively recorded as the posterior joint probability and prior joint probability of each substation.

[0023] In one embodiment, the expression of the fitting rationality factor of the missing data of each substation is:

[0024] , where is the fitting rationality factor of the missing data of the xth substation; is the dynamic fitting confidence of the xth substation; is the posterior joint probability of the xth substation; is the prior joint probability of the xth substation; is the exponential function with the natural constant e as the base.

[0025] In one embodiment, the process of obtaining the final fitted value of each missing data is as follows:

[0026] For each substation, if the fitting rationality factor of the substation is greater than or equal to a preset first threshold, the fitted values of the missing data of the substation are used as the final fitted values;

[0027] If the fitting rationality factor of the substation is greater than or equal to a preset second threshold and less than the preset first threshold, the posterior weight and prior weight of the substation are calculated based on the posterior joint probability, prior joint probability and the fitting rationality factor of the substation; based on the posterior weight and prior weight, combined with the posterior probability of each current of each substation and the historical data, the final fitted values of the missing data of the substation are obtained;

[0028] If the fitting rationality factor of the current kiosk is less than the preset second threshold, the missing data in the current kiosk is still regarded as missing data, and the final fitting values of the missing data of the kiosks with the fitting rationality factor greater than or equal to the preset second threshold are used as the non-missing data of the corresponding kiosks, and the fitting rationality factor of the current kiosk is recalculated until the fitting rationality factor of the current kiosk is greater than or equal to the second threshold, and the final fitting values of the missing data are obtained.

[0029] In one embodiment, the expressions for calculating the posterior weight and the prior weight of the kiosk are:

[0030]

[0031]

[0032] In the formula, and are the posterior weight and the prior weight of the x-th kiosk respectively; is the fitting rationality factor of the missing data of the x-th kiosk; is the posterior joint probability of the x-th kiosk; is the prior joint probability of the x-th kiosk.

[0033] In one embodiment, based on the posterior weight and the prior weight, combined with the posterior probability and the historical data of each current of each kiosk, obtaining the final fitting values of the missing data of each kiosk specifically includes:

[0034] Calculating the posterior expectation value of each current of each kiosk through the posterior probability of each current of each kiosk;

[0035] Denote the mean value of all data of the c-th current in the historical data of the x-th kiosk as ;

[0036] Denote the corrected fitting value of the missing data of the c-th current in the most recently collected data of the x-th kiosk as , The expression of is: In the formula, is the posterior expectation value of the c-th current of the x-th kiosk;

[0037] Taking the corrected fitting values of each current of each kiosk as the final fitting values of each missing data of each current of each kiosk.

[0038] This application has at least the following beneficial effects:

[0039] This application collects various current data on the same side of each substation in real time, fits the missing data based on the current balance equation and historical data, analyzes the differences between the fitted values of the missing data and the non-missing data, calculates the dynamic fitting confidence of each substation, evaluates the reliability of the fitted missing data of the substation, and provides a basis for subsequent judgment of fitting rationality; through the dynamic fitting confidence of each substation, and based on the Bayesian network combined with historical data and real-time data, evaluates the rationality of the fitted values, and excludes the cases where the fitting results deviate greatly from the historical data; dynamically adjusts the fitted values of the missing current data of the substation according to the fitting rationality factor, directly incorporates the highly reliable fitted values into the ranging calculation, performs weighted average correction on the relatively reliable fitted values, and initiates iterative calculation for the unreliable fitted values to readjust the fitted values, which helps to improve the fault ranging fault tolerance and accuracy; performs fault ranging through the dynamically adjusted fitted values and the non-missing current data, improves the stability and reliability of ranging, can adapt to various fault types and complex system topologies, avoids the problem that the fitting data deviation is superimposed in the scenario of missing data in multiple substations, affects the ranging accuracy, and ensures the safe and stable operation of the electrified railway. Brief Description of the Drawings

[0040] To more clearly illustrate the technical solutions and advantages in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0041] Figure 1 It is a flowchart of the method for compensating missing data in the fault ranging of the fully parallel AT power supply of the electrified railway provided by the present application;

[0042] Figure 2 It is a schematic diagram of the acquisition process of the final fitted values of each missing data. Detailed Embodiments

[0043] To further elaborate on the technical means and effects adopted by the present application to achieve the intended invention purpose, the following, in combination with the drawings and preferred embodiments, details the specific embodiments, structures, features, and effects of the method for compensating missing data in the fault ranging of the fully parallel AT power supply of the electrified railway proposed according to the present application. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which this application belongs.

[0045] The following specifically describes the specific solution of the method for compensating for missing fault location data in the fully parallel AT power supply of electrified railways in conjunction with the accompanying drawings.

[0046] A method for compensating for missing fault location data in the fully parallel AT power supply of electrified railways provided by an embodiment of the present application.

[0047] Specifically, the following method for compensating for missing fault location data in the fully parallel AT power supply of electrified railways is provided. Please refer to Figure 1 , and this method includes the following steps:

[0048] Step S1, collect various current data on the same side of each substation in real time, including T-line current, F-line current, and suction current; divide all current data into the most recently collected data and historical data according to the proximity of the collection time.

[0049] Assume that there are N substations in the electrified railway system, and the xth substation is denoted as .

[0050] Deploy high-precision current sensors and GPS modules in all substations. Real-time capture various current data of each substation through the current sensors. Since in the fully parallel AT power supply mode of railways, the circuit is divided into the up-line side and the down-line side, the embodiment of the present application takes the down-line side as an example for analysis. Collect the T-line current, F-line current, and suction current on the down-line side of each substation through the current sensors. Among them, the T-line current of the substation is denoted as , the F-line current is denoted as , and the suction current is denoted as .

[0051] For the data acquisition of the current sensors, preferably, in the embodiment of the present application, the data acquisition frequency is set to 1 kHz. As other embodiments of the present application, the implementer can set the data acquisition frequency according to the actual situation.

[0052] Each substation aligns its local clock through the PPS (pulses per second) signal generated by the GPS module to ensure that the timestamp error between different substations is less than or equal to . To ensure timing consistency, the PPS signal of the GPS module and the PTP protocol (Precision Time Protocol) are used to achieve hardware-level clock synchronization. Among them, the clock synchronization method is a well-known technology, and the specific process will not be elaborated here.

[0053] For the T-line current, F-line current, and pickup current data collected at each substation, implement a hierarchical storage strategy, specifically: Consider the current data collected within the most recent 1 hour from the current moment as the most recently collected data and store it in the in-memory database to support real-time queries; consider the current data collected from 1 hour ago to 1 month ago as intermediate data and store it in the time-series database; consider the current data collected more than 1 month ago as historical data and archive it to the distributed file system.

[0054] Step S2: Fit the missing data at each substation through the non-missing data at each substation to obtain the fitted values of each missing data; Calculate the dynamic fitting confidence level of each substation based on the difference between the fitted values of the missing data and the non-missing data in the most recently collected data at each substation, as well as the difference in the fitted values of the missing data between each substation.

[0055] In the scenario of missing data at multiple substations, the missing data at each substation will introduce a certain deviation, and the deviations in the multi-substation scenario will be superimposed on each other, significantly amplifying the overall error and affecting the ranging accuracy.

[0056] First, for each piece of missing data at each substation, fit each piece of missing data through the non-missing data at each substation by existing methods, specifically:

[0057] Taking substation as an example, through the principle of current balance, a current balance equation can be established: , where is the T-line current of substation , is the F-line current of substation , is the pickup current of substation . Obtain the fitted values of each missing current data through the non-missing data and the current balance equation: If the number of missing data among the three current data collected at substation at the same moment is 1, that is, only one of the three current data is missing, then solve the missing data at this moment through the current balance equation. For example, assume that is missing at this moment, while and are known, then the fitted value of the missing pickup current value can be obtained through ; If more than two of the three current data collected at substation at the same moment are unknown, then the mean or standard deviation of historical data can be used to fit the missing data. For example, if and If it is missing, the average value of the pick-up current at all times before this time is used as the fitting value of the pick-up current at this time, and the average value of the T-line current at all times before this time is used as the fitting value of the T-line current at this time. Thus, the fitting value of each missing data can be obtained through the above method. Among them, the fitting value of the p-th missing data of the is denoted as .

[0058] Then, to analyze the reliability of the fitting values of the missing data of each , taking as an example, the most recent collected data of this is obtained from the in-memory database, and the average value of all kinds of current data in the most recent collected data of this is denoted as the first average value; for each missing data of this , the difference between the fitting value of this missing data and the first current average value is denoted as the first difference of this missing data;

[0059] Furthermore, calculate the dynamic fitting confidence of each , and the expression is:

[0060]

[0061]

[0062] In the formula, is the dynamic fitting confidence of the x-th ; is the Sigmoid function; is the number of missing data in the most recent collected data of the x-th ; is the first difference of the p-th missing data in the most recent collected data of the x-th ; is the proportion of the missing current fitting of the x-th ;

[0063] is the fitting value of the p-th missing data in the most recent collected data of the x-th ; N is the number of ; is the number of missing data in the most recent collected data of the n-th ; is the -th missing data in the most recent collected data of the n-th .

[0064] quantifies the degree of deviation of the fitting value from the normal working condition. The larger its absolute value, the less credible the fitting value; represents the proportion of the total fitting value of the missing current of the x-th in the total fitting value of the missing current of all . The numerator represents the actual contribution of this in the fault current. The larger its value, the higher its current intensity or the more missing data. The denominator represents the scale of the overall missing current of all , reflecting the severity and distribution range of the current fault, The larger the value, the higher the proportion of the missing current of the substation kiosk, which may be located at key topological positions, such as near the substation or high-fault areas.

[0065] It is used to quantify the fitting reliability of the missing data points of the x-th substation kiosk. The closer its value is to 1, the more reliable the fitting result of the missing data of the substation kiosk. On the contrary, when its value is close to 0, it indicates that the fitting value of the missing data of the substation kiosk deviates from the normal operating condition or the contribution ratio of the substation kiosk is low, and further correction is required.

[0066] Step S3, construct a probabilistic graphical model based on the historical data of all substation kiosks and train it; based on the recently collected data of all substation kiosks, combined with the trained probabilistic graphical model, calculate the posterior probability of each current of each substation kiosk, as well as the posterior joint probability and prior joint probability of each substation kiosk; based on the posterior joint probability and the prior joint probability, combined with the dynamic fitting confidence of each substation kiosk, calculate the fitting rationality factor of the missing data of each substation kiosk.

[0067] When processing the missing data of substation kiosks in the electrified railway system, not only the reliability of data fitting should be concerned, but also the fitting problem of current missing data relative to historical data should be noted. However, due to the diversity of fault types and the complexity of system operating conditions, a single data fitting method often fails to comprehensively reflect the actual scenario. Different fault types may lead to significant differences in the current distribution pattern, and the historical data may contain current characteristics under various operating conditions, which makes the fitting result need to consider both the rationality of current observed data and the consistency of historical data. In addition, changes in system topology, equipment aging, and external environmental factors may also affect the distribution of current data, further increasing the difficulty of fitting. Therefore, when evaluating the fitting effect of missing data, multiple factors need to be comprehensively considered to ensure that the fitting result can adapt to the current fault conditions and maintain a certain coherence with historical laws.

[0068] (1) Construct a Bayesian network model according to the current data of the substation kiosk:

[0069] Take the three fault types of TR, FR, and TF as the top-level inputs of the Bayesian network, that is, take the three fault types as the root nodes, take the T-line current, F-line current, and pickup current of each substation kiosk as the intermediate nodes, and take the fitting values of the missing data of each substation kiosk as the leaf nodes to construct a Bayesian network model;

[0070] Extract features based on the historical data of all substation kiosks and calculate the conditional probability table, and then use these conditional probability tables to train the Bayesian network model; the calculation of the conditional probability table and the training of the Bayesian network model are well-known technologies, and the specific process will not be elaborated here;

[0071] Take the most recently collected data of all kiosks as the input of the trained Bayesian network model, and respectively output the posterior probabilities of the T-line current, F-line current, and pickup current of each kiosk. Calculate the posterior expected value of each current of the kiosk based on the posterior probability of each current of the kiosk. At the same time, calculate the posterior probability and prior probability of the joint of all current data of each kiosk through Bayes' formula, and record them as the posterior joint probability and prior joint probability of the kiosk respectively. The calculation of the posterior probability, prior probability, and posterior expected value is a well-known technology, and the specific process will not be elaborated here.

[0072] (2)Based on the above analysis, take the xth kiosk as an example to calculate the fitting rationality factor of the missing data of each kiosk. The expression is:

[0073]

[0074] In the formula, is the fitting rationality factor of the missing data of the xth kiosk; is the dynamic fitting confidence of the xth kiosk; is the posterior joint probability of the xth kiosk; is the prior joint probability of the xth kiosk; is the exponential function with the natural constant e as the base.

[0075] Evaluates the reliability of the fitting of the missing data in the kiosk The larger the value, the more credible the fitted missing value; is the posterior probability calculated through the Bayesian network model based on the observed data, that is, the data in the in-memory database, indicating the rationality of the fitted value of the missing current data in the kiosk under the current network topology and fault conditions. The larger the value, the better the fitted value matches the current observed data; is the prior probability obtained based on historical data, that is, the data archived in the distributed file system, indicating the matching degree of the fitted value of the missing current data in the kiosk with the historical pattern. The larger the value, the more the fitted value conforms to the historical law; is used to quantify the improvement or decrease degree of the rationality of the fitted value under the current observed data relative to the historical pattern. When the difference between its value and 1 is smaller, it indicates that the fitted value is more similar to the historical pattern. At this time, it shows that the fitted value can more accurately conform to the past data law, has good generalization ability, and the fitting result is more stable and reliable.

[0076] is used to quantify the kiosk The overall reliability of the fitted values of the missing current data. When the value of B is closer to 1, the rationality of the fitted values under the current observed data is highly consistent with the historical pattern, indicating that the fitting results not only conform to the current observations but also to the historical laws, and the fitting results of the missing data are more reasonable.

[0077] Step S4: Based on the fitting rationality factor and the posterior probability of each type of current in each substation, dynamically adjust the fitted values of the missing data in each substation to obtain the final fitted values of each missing data.

[0078] When dealing with the challenge of missing data in multiple substations of the fully parallel AT power supply system for electrified railways, traditional methods are difficult to accurately locate faults due to the non-linear superposition of current vector relationships and the ambiguity of contribution ratios. Especially in multi-missing scenarios, the error accumulation effect significantly reduces the ranging accuracy. Therefore, a fitting rationality factor is introduced as the core evaluation index. By comprehensively considering the degree of fit between the current fitting data and the historical pattern, a high-tolerance and high-precision solution for fault ranging in complex scenarios is provided. Specifically:

[0079] (1) Taking the x-th substation as an example, use the fitting rationality factor of the missing data of the x-th substation as the weight factor to dynamically adjust the fitted values of the missing data of the x-th substation: Set the first threshold and the second threshold . Preferably, in the embodiments of the present application, the first threshold is set to 0.8, and the second threshold is set to 0.5; As other embodiments of the present application, the implementer can set the first threshold and the second threshold according to the actual situation.

[0080] (2) When , it indicates that the reliability of the fitted values of the missing data in the substation is high. Directly use the fitted values of the missing data in the substation as the final fitted values and include them in the ranging calculation;

[0081] (3) When , it indicates that there are certain deviations in the fitted values of the missing data in the substation . Then, combine the posterior joint probability and the prior joint probability generated by the Bayesian network, and correct the current value through weighted average. The specific correction method is:

[0082] First, according to the posterior joint probability, the prior joint probability and the fitting rationality factor of the substation , calculate the posterior weight and the prior weight of the substation . The expression is:

[0083]

[0084]

[0085] This weight dynamically adjusts the contribution ratio of the posterior probability to the prior probability through a rationality factor. When is relatively high, it emphasizes the local features of the current observed data ; when is relatively low, it enhances the global regularity constraint of historical data . Thus, while retaining the advantage of real-time performance, it suppresses the fitting deviation caused by data missing or noise, and improves the stability and physical consistency of the compensation result.

[0086] Furthermore, the mean value of all data of each current in the historical data of the x-th substation is used as the prior current estimation value of the missing data of each current in this substation. The fitting values of the missing data of each current in the substation are weighted and averaged for correction. The specific correction formula is:

[0087] , where is the corrected fitting value of the missing data of the c-th current in the most recently collected data of the x-th substation, is the posterior expected value of the c-th current in the x-th substation, is the prior current estimation value of the missing data of the c-th current in the x-th substation.

[0088] The corrected fitting values of each current in each substation are used as the final fitting values of each missing data of each current in each substation, and are incorporated into the ranging calculation.

[0089] Through weighted average correction, the posterior estimation with high confidence is fused with the historical prior estimation, which not only avoids the error accumulation of a single data source, but also balances the real-time data and long-term regularity, enabling the corrected current value to adapt to the current fault scenario and meet the system topology constraint, and finally reducing the ranging error in multiple missing scenarios.

[0090] (4) When , start iterative calculation, delete the fitting values of the substations with the fitting rationality factor less than the second threshold, that is, still regard the missing data in the substations as missing data, and use the final fitting values of each missing data of the substations with the fitting rationality factor greater than as the known current data of the corresponding substations. Repeat the calculations in steps S2 - S4 and perform iteration until the fitting rationality factors of all substations are greater than the second threshold, and the final fitting values of the missing data are obtained.

[0091] Through the collaborative regulation of the threshold and the fitting rationality factor, this compensation method achieves high-precision fault location in multiple missing scenarios. When the fitting value of the missing data is highly reliable, the fitting value is directly adopted to ensure real-time performance. When the fitting value of the missing data is relatively reliable, the Bayesian posterior probability and historical prior data are fused through weighted averaging to dynamically adjust the posterior weight and prior weight. This not only utilizes the local characteristics of real-time data to adapt to the current fault type but also uses historical rules to restrain and suppress noise interference to ensure physical consistency. When the missing fitting value is unreliable, the low-confidence fitting values are iteratively deleted and recalculated to avoid error accumulation. Ultimately, the ranging accuracy is significantly improved, and the system fault tolerance is enhanced. It supports simultaneous missing data in multiple substations and kiosks, taking into account both real-time performance and stability, providing a reliable guarantee for the safe operation of electrified railways.

[0092] Step S5: Perform fault location based on the final fitting values of the missing data of each substation and kiosk and the non-missing data.

[0093] The final fitting values of each missing data in the most recently collected data are obtained through the above method, thereby filling in the missing values of the current data of each substation and kiosk. Further, based on the current data of all substations and kiosks, combined with the principle of the cross-link line current ratio, fault location is performed. This principle obtains the cross-link line current ratio formula for fault location by analyzing the power equations satisfied by various fault lines in the fully parallel AT power supply traction network. When a short-circuit fault occurs, by measuring the cross-link line currents of the up and down lines, the position of the fault point is calculated using the current ratio formula. This method can accurately determine the fault type and calibrate the position of the fault point, which is crucial for ensuring the safe and reliable operation of electrified railways. The principle of the cross-link line current ratio is applicable to the case where the cross-link line is not split in the up and down fully parallel AT power supply mode, and can accurately calibrate TR, FR, TF faults, as well as open-circuit faults and fault directions. Among them, fault location by the cross-link line current ratio method is a well-known technology, and the specific process will not be elaborated here.

[0094] The schematic diagram of the acquisition process of the final fitting values of each missing data is as Figure 2 shown.

[0095] In summary, in the embodiments of the present application, various current data on the same side of each substation kiosk are collected in real time. The missing data are fitted based on the current balance equation and historical data. The differences between the fitted values of the missing data and the non-missing data are analyzed, the dynamic fitting confidence of each substation kiosk is calculated, and the reliability of the fitted missing data of the substation kiosk is evaluated, providing a basis for subsequent judgment of fitting rationality. Through the dynamic fitting confidence of each substation kiosk, and based on the Bayesian network combined with historical data and real-time data, the rationality of the fitted values is evaluated, and the situations where the fitting results deviate greatly from the historical data are excluded. According to the fitting rationality factor, the fitted values of the missing current data of the substation kiosk are dynamically adjusted. The fitted values with high reliability are directly incorporated into the ranging calculation, the relatively reliable fitted values are corrected by weighted average, and the unreliable fitted values are started for iterative calculation to readjust the fitted values, which helps to improve the fault ranging fault tolerance and accuracy. Fault ranging is performed through the dynamically adjusted fitted values and the non-missing current data, improving the stability and reliability of ranging, being able to adapt to various fault types and complex system topologies, avoiding the problem that the fitting data deviation is superimposed in the scenario of missing data in multiple substation kiosks and affecting the ranging accuracy, and ensuring the safe and stable operation of electrified railways.

[0096] It should be noted that the above sequence of the embodiments of the present application is only for description and does not represent the superiority or inferiority of the embodiments. And the above specific embodiments of the present application have been described. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0097] The embodiments in the present application are all described in a progressive manner. The same or similar parts between the embodiments can be referred to each other, and the key points of each embodiment are the differences from other embodiments.

[0098] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Modifying the technical solutions recorded in the foregoing embodiments, or equivalently replacing some of the technical features, does not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A method for compensating for missing data of fault distance measurement in full parallel AT power supply of electrified railway, characterized in that: The method comprises the following steps: Collect various current data on the same side of each booth in real time, including T-line current, F-line current and suction current; divide all current data into the most recently collected data and historical data according to the time of collection; Fitting the missing data of each kiosk with the non-missing data of each kiosk to obtain the fitting value of each missing data; calculating the dynamic fitting confidence of each kiosk based on the difference between the fitting value of missing data and the non-missing data in the most recently collected data of each kiosk, and the difference between the fitting values ​​of missing data between the kiosk; Based on the historical data of all the pavilions, a probability graphical model is constructed and trained; based on the most recently collected data of all the pavilions, combined with the trained probability graphical model, the posterior probability of each current of each pavilion, as well as the posterior joint probability and the prior joint probability of each pavilion are calculated; based on the posterior joint probability and the prior joint probability, combined with the dynamic fitting confidence of each pavilion, the fitting rationality factor of the missing data of each pavilion is calculated; Based on the fitting rationality factor and the posterior probability of each current in each pavilion, dynamically adjust the fitting value of the missing data of each pavilion to obtain the final fitting value of each missing data; Fault location is performed based on the final fitted values ​​of the missing data of each kiosk and the non-missing data; The expression of the fitting rationality factor of the missing data of each pavilion is: , where is the fitting rationality factor for the xth missing data; is the dynamic fitting confidence of the x-th pavilion; is the posterior joint probability of the xth pavilion; is the prior joint probability of the xth pavilion; is an exponential function with the natural constant e as base; The process of obtaining the final fitting value of each missing data is as follows: For each pavilion, if the fitting rationality factor of the pavilion is greater than or equal to a preset first threshold, the fitting value of each missing data of the pavilion is used as the final fitting value; If the fitting rationality factor of the pavilion is greater than or equal to the preset second threshold and less than the preset first threshold, the posterior weight and the prior weight of the pavilion are calculated based on the posterior joint probability, the prior joint probability and the fitting rationality factor of the pavilion; based on the posterior weight and the prior weight, combined with the posterior probability and historical data of each current of each pavilion, the final fitting value of each missing data of the pavilion is obtained; If the fitting rationality factor of the current station is less than the preset second threshold, the missing data in the current station is still regarded as missing data, and the final fitting value of each missing data of the station whose fitting rationality factor is greater than or equal to the preset second threshold is used as the non-missing data of the corresponding station, and the fitting rationality factor of the current station is recalculated until the fitting rationality factor of the current station is greater than or equal to the second threshold, so as to obtain the final fitting value of the missing data.

2. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 1, characterized in that: The method for obtaining the fitting value of each missing data is: For each pavilion, if the number of missing data in the three current data at any time of the pavilion is 1, the value of the missing data at any time is solved by the current balance equation as its fitting value; If two or more of the three current data at any moment are missing, the mean of the same current data at all moments before the said moment is used as the fitting value of the missing data corresponding to the said moment.

3. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 1, characterized in that: The acquisition process of the dynamic fitting confidence of each pavilion is as follows: Calculate the missing current fitting ratio of each pavilion based on the difference between the missing data fitting values ​​of each pavilion and other pavilions; The mean of all current data in the most recent data collected at the x-th station is recorded as the first mean; the fitted value of the p-th missing data in the most recent data collected at the x-th station is recorded as ,Will The difference between the first mean and ; The dynamic fitting confidence of the xth pavilion is recorded as , The expression is: , where is the Sigmoid function; is the number of missing data in the most recently collected data of the x-th station; is the missing current fitting ratio of the x-th pavilion.

4. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 3, characterized in that: The expression of the missing current fitting ratio is: , where N is the number of pavilions; is the number of missing data in the most recently collected data of the nth station; The most recent data collected by the nth station The fitted values ​​for the missing data.

5. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 1, characterized in that: The process of constructing and training the probabilistic graphical model is as follows: A Bayesian network model is constructed by taking various fault types as root nodes, various currents of each station as intermediate nodes, and the fitting values ​​of missing data of each station as leaf nodes; the Bayesian network model is trained through the historical data of all stations.

6. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 1, characterized in that: The process of obtaining the posterior probability of each current of each booth, and the posterior joint probability and the prior joint probability of each booth is as follows: The most recently collected data of all the pavilions are used as the input of the trained Bayesian network model, and the posterior probability of each current in each pavilion is output respectively. The posterior probability and prior probability of the joint of all kinds of current data of each pavilion are calculated by the Bayesian formula, which are recorded as the posterior joint probability and prior joint probability of each pavilion respectively.

7. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 1, characterized in that: The expressions of the posterior weight and the prior weight of the calculation are: In the formula, , are the posterior weight and prior weight of the x-th pavilion respectively; is the fitting rationality factor for the xth missing data; is the posterior joint probability of the xth pavilion; is the prior joint probability of the xth pavilion.

8. The method for compensating for missing data of fault ranging in full parallel AT power supply of electrified railway according to claim 7, characterized in that: Based on the posterior weights and the prior weights, combined with the posterior probability and historical data of each current of each pavilion, the final fitting value of each missing data of the pavilion is obtained, specifically: Calculate the posterior expected value of each current in each pavilion through the posterior probability of each current in each pavilion; The mean value of all data of the cth current in the xth historical data of the pavilion is recorded as ; The corrected fitting value of the missing data of the cth current in the most recently collected data of the xth pavilion is recorded as , The expression is: , where is the posterior expected value of the cth current at the xth pavilion; The corrected fitting value of each current in each booth is used as the final fitting value of each missing data of each current in each booth.

Images