Railway contact network reliability evaluation method based on Markov chain and dynamic Bayesian network

Abstract

The invention discloses a railway catenary reliability evaluation method based on a Markov chain and a dynamic Bayesian network, and the method achieves the systematic monitoring and fault early warning of the operation state of catenary equipment through the construction of a catenary monitoring data collection system, and the utilization of multi-source data integration, multi-node multivariable modeling and hierarchical time scale monitoring. According to the scheme, comprehensive acquisition and analysis of key parameters of a contact network system are ensured by integrating various factors such as meteorology, load and equipment state. Through combined application of the dynamic Bayesian network and the Markov chain, multi-node and multi-variable modeling is carried out, and the dynamic influence of factors such as equipment aging, load fluctuation and external environment change on the system reliability can be accurately captured. In addition, by refining the time scale, performing high-frequency sampling on short-term variables such as weather and load fluctuation, and performing low-frequency monitoring on long-term factors such as equipment aging and corrosion, the state of the contact network is finely evaluated on different time scales.

Description

Technical Field

[0001] The invention relates to the field of reliability assessment of power supply systems, in particular to a reliability assessment method for railway contact networks based on Markov chains and dynamic Bayesian networks. Background Art

[0002] In recent years, my country's high-speed rail, subway and other rail transportation have achieved rapid development. The contact network is very important for the safe running of trains and is an important guarantee for the safe and secure operation of trains. Therefore, it is very important to conduct reliability assessment of railway contact networks. There are many problems in the existing railway contact network reliability assessment methods: the reliability of the railway contact network system is affected by a variety of dynamic factors, such as weather changes, temperature changes, etc.; the contact network system changes in real time, and it is impossible to achieve real-time feedback and dynamic adjustment of the contact network system status; in the reliability assessment of the railway contact network system, the impact of components on the reliability of the contact network system is not fully considered. Summary of the invention

[0003] In order to overcome the above problems existing in the prior art, the present invention proposes a railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network.

[0004] The technical solution adopted by the present invention to solve the technical problem is: a railway contact network reliability evaluation method based on Markov chain and dynamic Bayesian network, comprising the following steps: Step 1: construct a dynamic Bayesian network model of the overhead contact network system, use the state transition characteristics of the Markov chain and the probabilistic reasoning ability of the Bayesian network to describe the state transition and fault evolution process of each subsystem at different times, conduct real-time monitoring and collection of operating data of the railway overhead contact network system, and build a real-time data collection system; Step 2, introducing the real-time monitoring data and reliability evaluation index obtained in step 1 into the overhead line reliability evaluation index system, while considering the impact of maintenance on system reliability; Step 3, classifying the reliability of the contact network system status through the historical operation data of the railway contact network and the operation status of the system; Step 4, using the Gaussian membership function to calculate the membership of the overhead line system index, by analyzing the historical monitoring and operation data of the overhead line system as sample data of different levels, and calculating the Gaussian membership function at different levels; Step 5: Determine the evaluation time interval in days, build a judgment matrix based on the importance of the indicators, and calculate the consistency ratio CR of the established matrix. If the calculated CR is less than 0.1, it is considered that the consistency of the judgment matrix is ​​within a reasonable range. If the calculated CR is greater than or equal to 0.1, the importance of the indicator is revised again. According to the judgment matrix, the weight vector is obtained, and the weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion; Step 6: Set the credibility of the system state, and combine multiple belief functions into a unified belief degree through Dempster's rule; establish a credibility distribution matrix based on the belief functions of multiple indicators, and obtain the probability distribution function of the system state through the integration of evidence theory; Step 7, divide each state according to the historical operation data of the railway overhead line system, correspond the probability distribution function of the system to obtain the state level of the system, and finally obtain the reliability evaluation result.

[0005] In the above-mentioned railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network, in step 1, real-time monitoring and collection are performed through a monitoring data acquisition system, and the monitoring data acquisition system includes a data acquisition layer, a diagnostic analysis layer, a data storage layer, and an application execution layer. The data acquisition layer captures and feeds back the operating status data of the equipment in real time; the diagnostic analysis layer analyzes and diagnoses the data of the data acquisition layer to discover potential faults of the system; the data storage layer is used to store historical operating status data, operating analysis and monitoring of contact network equipment, and various intelligent algorithms related to equipment operation control based on training; the application execution layer converts data processing and management results into actual operation and maintenance measures.

[0006] In the above-mentioned railway overhead contact network reliability assessment method based on Markov chain and dynamic Bayesian network, the specific process of using Markov chain for real-time monitoring in step 1 is: using a first-order Markov chain to simulate the rapid response of short-term meteorological factors and load changes, using a two-order Markov chain to integrate long-term historical information to reflect the long-term dependence of equipment aging and failure probability, defining a unified spatial state, and merging two Markov chains; Among them, the meteorological state W t and power load status L t As a short-term state variable, it is only affected by the previous time step; the equipment aging state A t and the failure probability F t As a long-term state variable, its current state depends on the states of the previous two time steps.

[0007] In the above-mentioned railway contact network reliability evaluation method based on Markov chain and dynamic Bayesian network, the contact network reliability evaluation index system in step 2 includes a target layer, an index layer, and a sub-index layer. The target layer is the evaluation of the operation status of the contact network, and the index layer is the system real-time monitoring data. , System performance indicators , Equipment status monitoring indicators and related reliability indicators The sub-index layer consists of steady-state availability, average failure frequency, average repair time, equipment maintenance rate, and contact line relative status index.

[0008] In the above-mentioned railway contact network reliability evaluation method based on Markov chain and dynamic Bayesian network, the membership degree in step 4 is The calculation method is: ; ; ; Among them, j=1,2,3,4,5 represent the system level very good, good, average, poor and very poor respectively, n is the number of sample points, is the i-th sample point, is the center point, is the standard deviation.

[0009] In the above-mentioned railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network, the step 5 specifically includes: determining the assessment time interval in days as , assign initial weights to each indicator , ,…, The initial weight reflects the relative importance of each indicator without considering the time factor. The indicator data at different time intervals are collected and the indicator weight is adjusted according to the time interval: ; ; in, is the attenuation coefficient, is the time decay function, is the importance of indicator i relative to indicator j at time t; For each time interval , construct the judgment matrix R(t): ; Calculate the maximum eigenvalue λmax(t) of the judgment matrix for each time interval, and calculate the consistency index CI(t) and consistency ratio CR(t): ; ; Among them, RI is the random consistency index, which depends on the order of the matrix; m is the order of the judgment matrix; The judgment matrix of each time interval is checked for consistency to ensure the consistency of the evaluation. If CR(t)<0.1, the judgment matrix has satisfactory consistency; if CR(t)≥0.1, the judgment matrix needs to be readjusted; , solve for the eigenvector q, is a unit vector. After normalizing the eigenvector q, the weight vector Q(t) is: ; ; in, is the weight of criterion k at time t, is the e-th vector in the feature vector q, and r is the number of evaluation indicators; The weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion at different time points: ; Where T is the total number of time intervals.

[0010] In the above-mentioned railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network, the step 6 is specifically as follows: The identification framework of railway contact network is H={ , , , , }, set the belief degree of each state of the system The system is in state Credibility: ; in, For the system under conditions The system is in state The probability of; r is the number of evaluation indicators, k is the kth indicator; The belief functions are fused through Dempster's rule. When multiple belief functions come from different sources, they are merged into a unified belief degree: ; Among them, i is the i-th state of the system; Based on the belief functions of multiple indicators, the fused belief degrees form a credibility distribution matrix Y: ; Through the integration of evidence theory, the probability distribution of each system state is obtained, and the probability distribution function of the system state is defined for: ; in, for The degree of belief in the state; The railway overhead contact system is in a dynamic environment. The state probability of the overhead contact system changes with time. The state transfer matrix is ​​introduced. , and obtain the probability distribution of the contact network system state at each time step : .

[0011] In the above-mentioned railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network, the step 7 is specifically: according to the state distribution of the historical operation data of the railway contact network system, the boundary points of each state are calculated to obtain the dynamic boundary value b 1， b 2， b 3， b4, maps different status levels to corresponding evaluation results according to the probability distribution of the system: Very good: the probability distribution value is higher than b1, the reliability of the system is very good, and the probability of failure is very low; Better: The probability distribution value is between [b2, b1), the system status is relatively healthy, but there are some minor faults; Normal: The probability distribution value is between [b3, b2), the system has certain faults and may need attention and maintenance; Poor: The probability distribution value is between [b4, b3), the system has a greater risk or multiple failures, and the reliability is poor; Very bad: The probability distribution value is lower than b4, the failure risk of the system is very high, and the reliability is seriously reduced.

[0012] The beneficial effect of the present invention is that the present invention realizes systematic monitoring of the operating status of the contact network equipment and fault warning by constructing a contact network monitoring data acquisition system, utilizing multi-source data integration, multi-node multi-variable modeling and hierarchical time scale monitoring. By integrating multiple factors such as meteorology, load, and equipment status, comprehensive collection and analysis of key parameters of the contact network system are ensured. Through the joint application of dynamic Bayesian networks and Markov chains, multi-node and multi-variable modeling is performed, which can accurately capture the dynamic impact of factors such as equipment aging, load fluctuations, and external environmental changes on system reliability. In addition, by refining the time scale, high-frequency sampling is performed on short-term variables such as meteorology and load fluctuations, and low-frequency monitoring is performed on long-term factors such as equipment aging and corrosion, so as to perform a detailed assessment of the contact network status at different time scales.

[0013] Through the monitoring and acquisition system, the system integrates multiple data sources such as meteorology, load and equipment status, and can fully grasp the operating status of the contact network, thereby effectively improving the accuracy of fault identification and early warning. Secondly, based on the multi-node and multi-variable modeling method of dynamic Bayesian network and Markov chain, the interaction between equipment and environmental factors is analyzed to improve the accuracy of the reliability assessment of the contact network system. In addition, the present invention adopts hierarchical time scale monitoring, and realizes short-term fluctuation response and long-term trend analysis through high-frequency sampling of short-term variables and low-frequency sampling of long-term variables, and considers the impact of maintenance rate on the contact network system in the reliability assessment system. Finally, the dynamic hierarchical analysis method is used to dynamically adjust the weights of each indicator in the model according to the system status reflected by the real-time monitoring data, so as to more accurately reflect the reliability of the contact network system. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a schematic diagram of the process of the present invention; Figure 2 It is a structural framework diagram of the monitoring data acquisition system of the present invention; Figure 3 It is a schematic diagram of the dynamic network structure of the contact network of the present invention; Figure 4 It is a schematic diagram of the contact network reliability evaluation index system of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0016] The present invention discloses a railway contact network reliability evaluation method based on Markov chain and dynamic Bayesian network. Figure 1 As shown, the specific steps include: Step 1: construct a dynamic Bayesian network model of the overhead contact network system, use the state transition characteristics of the Markov chain and the probabilistic reasoning ability of the Bayesian network to describe the state transition and fault evolution process of each subsystem at different times, conduct real-time monitoring and collection of operating data of the railway overhead contact network system, and build a real-time data collection system; Step 2, introducing the real-time monitoring data and reliability evaluation index obtained in step 1 into the overhead line reliability evaluation index system, while considering the impact of maintenance on system reliability; Step 3, classifying the reliability of the contact network system status through the historical operation data of the railway contact network and the operation status of the system; Step 4, using the Gaussian membership function to calculate the membership of the overhead line system index, by analyzing the historical monitoring and operation data of the overhead line system as sample data of different levels, and calculating the Gaussian membership function at different levels; Step 5: Determine the evaluation time interval in days, build a judgment matrix based on the importance of the indicators, and calculate the consistency ratio CR of the established matrix. If the calculated CR is less than 0.1, it is considered that the consistency of the judgment matrix is ​​within a reasonable range. If the calculated CR is greater than or equal to 0.1, the importance of the indicator is revised again. According to the judgment matrix, the weight vector is obtained, and the weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion; Step 6: Set the credibility of the system state, and combine multiple belief functions into a unified belief degree through Dempster's rule; establish a credibility distribution matrix based on the belief functions of multiple indicators, and obtain the probability distribution function of the system state through the integration of evidence theory; Step 7, divide each state according to the historical operation data of the railway overhead line system, correspond the probability distribution function of the system to obtain the state level of the system, and finally obtain the reliability evaluation result.

[0017] In a specific embodiment, in step 1, real-time monitoring and collection are performed through a monitoring data collection system, and the structure framework of the monitoring data collection system is as follows: Figure 2 As shown, it specifically includes a data acquisition layer, a diagnostic analysis layer, a data storage layer, and an application execution layer. The data acquisition layer captures and feeds back the operating status data of the equipment in real time; the diagnostic analysis layer analyzes and diagnoses the data of the data acquisition layer to discover potential faults of the system; the data storage layer is used to store historical operating status data, operating analysis and monitoring of contact network equipment, and various intelligent algorithms related to equipment operation control based on training; the application execution layer converts data processing and management results into actual operation and maintenance measures.

[0018] The contact network system is composed of multiple subsystems connected in series, such as contact suspension, positioning device, pillars and foundation, support device, additional device and compensation device. There are complex dependencies and mutual influences between the subsystems. When any subsystem fails, the entire system may not operate normally. Therefore, when constructing the dynamic Bayesian network (DBN) model of the contact network system, the state transition characteristics of the Markov chain and the probabilistic reasoning ability of the Bayesian network are used to describe the state transition and fault evolution process of each subsystem at different times. By analyzing the structural characteristics of the contact network system, the short-term factors and long-term factors are modeled separately, and combined with the state transfer matrix between time slices, a unified contact network dynamic network structure is constructed, such as Figure 3As shown, D represents contact network fault, M1 represents contact suspension fault, M2 represents positioning device fault, M3 represents pillar and foundation fault, M4 represents supporting device fault, M5 represents additional device fault, M6 represents compensation device fault, X1 represents suspension string fault, X2 represents load-bearing cable fault, X3 represents contact wire fault, X4 represents wire clamp fault, X5 represents positioning rod fault, X6 represents combined positioner fault, X7 represents positioning hook fault, X8 represents windproof pull wire fault, X9 represents pillar fault, X10 represents foundation fault, X11 represents insulator fault, X12 represents wrist arm fault, X13 represents base fault, X14 represents bushing double ear fault, X15 represents additional conductor fault, X16 represents additional conductor pipe suspension device fault, X17 represents compensation rope fault, X18 represents tension compensation fault, G1 represents positioning wire clamp fault, G2 represents positioning pin fault, and G3 represents insulator fault.

[0019] In order to accurately evaluate the reliability of the contact network, it is necessary to combine short-term and long-term monitoring information and select the appropriate Markov chain order to capture the dependencies at different time scales. By monitoring and recording multivariate influencing factors at different scales, the contact network system can be effectively evaluated for reliability. First, meteorological factors and contact network load change information are monitored frequently in the short term. Combined with the historical data, fault information and equipment status information of the long-term operating contact network system, the contact network system can respond in the short term through the processing of multivariate at different time scales, reduce the risk of emergencies, better adapt to the changing characteristics of different factors in the contact network, and improve the accuracy of contact network reliability assessment.

[0020] The specific process of using Markov chain for real-time monitoring in step 1 is as follows: using a first-order Markov chain to simulate the rapid response of short-term meteorological factors and load changes, using a two-order Markov chain to integrate long-term historical information to reflect the long-term dependence of equipment aging and failure probability, defining a unified spatial state, and merging the two Markov chains; Among them, the meteorological state W t and power load status L t As a short-term state variable, it is only affected by the previous time step; the equipment aging state A t and the failure probability F t As a long-term state variable, its current state depends on the states of the previous two time steps. This effectively captures the long-range dependencies across time steps. The equipment maintenance status 𝐵𝑡 is directly related to the equipment maintenance status and has an important impact on system reliability.

[0021] Assumed weather conditions There are m possible states, so the transition matrix of the meteorological state is for Each element in the matrix Indicates weather conditions from Transfer to Probability: .

[0022] Assuming power load status There are n possible states. The transfer of load state depends not only on the current load state, but also on the meteorological state. Assumed weather conditions , load state transfer matrix For one The matrix of Indicates the weather conditions Under the load state from Transfer to Probability: ; The rows represent the load status at the current moment, the columns represent the load status at the next moment, and the matrix is ​​for specific meteorological conditions. The sum of the elements in each row should be 1.

[0023] The state transition matrix of a short-term Markov chain: ; Among them, short-term factors =( , ).

[0024] Assuming the device is aged There are p possible states, and the transition of the equipment aging state depends on the history of the load state. The transition matrix of the equipment aging state is For one The matrix, element Indicates that the load state Under the influence of Transfer to Probability: .

[0025] Equipment failure status There are four possible states: normal operation, minor fault, serious fault and under maintenance. The transition of equipment fault state depends not only on the current equipment fault state, but also on the equipment aging state, load state and meteorological state. Equipment fault state transition matrix is a 4x4 matrix with elements Indicates that the device is in state Transfer to Probability: ; Among them, long-term factors .

[0026] The state transition matrix of a long-term Markov chain: ; Among them, long-term factors .

[0027] Define a unified spatial state , merging two Markov chains, the space includes all possible state combinations of short-term and long-term factors , the merged state transfer matrix for: ; A comprehensive framework is provided through the combined Markov chain to monitor the changes of multivariables of the railway overhead line system at different time scales and their impact on the system status.

[0028] In a specific embodiment, the contact network reliability evaluation index system in step 2 is as follows: Figure 4 As shown, it specifically includes a target layer, an indicator layer, and a sub-indicator layer. The target layer is the evaluation of the operation status of the overhead contact network, and the indicator layer is the system real-time monitoring data. , System performance indicators , Equipment status monitoring indicators and related reliability indicators The sub-index layer consists of steady-state availability, average failure frequency, average repair time, equipment maintenance rate, and contact line relative status index.

[0029] Taking maintenance factors into consideration, the identification framework of railway contact network status is H={ , , , , }, the steady-state availability of the overhead line system refers to the state in which the system is in normal operation for a long time. The probability of , the steady-state probability distribution satisfies: , if the steady-state distribution is =[ , , , ], ; The steady-state availability A of the overhead line system is: ; in, is the probability of normal operation of the system in steady state, is the state transfer matrix; Average fault frequency of overhead line system is the frequency of transition from normal state to fault state: ; Among them, j is the different status levels of the system, t is the current running time, T0 is the system running time cycle, H t is the system state at time t, H j is the state corresponding to system level j, and P is the conditional probability; The mean repair time (MTTR) of the overhead line system is the average rate at which the fault state is transferred to the normal state: ; Among them, j is the different status levels of the system.

[0030] In a specific embodiment, in order to more intuitively reflect the actual operating status of the system, the reliability assessment of the contact network system status is divided into five levels based on historical experience:

[0031] (1) Very good ): In this state, the monitoring data of the contact network system or equipment indicators are in the best condition, and the possibility of contact network system failure is extremely low.

[0032] (2) Better ): In this state, it means that the monitoring data of the contact network system or equipment indicators are in normal working state, and the overall operating state of the contact network system in this operating state is good.

[0033] (3) General ): In this operating state, there is a possibility of failure in the contact network system, and the performance of the contact network system will be reduced to a certain extent, but it can still meet the basic operation of the contact network system.

[0034] (4) Poor ): The contact network system or equipment in this operating state is at risk of failure or has experienced a minor failure, requiring maintenance personnel to perform maintenance.

[0035] (5) Very poor ): In this operating state, the overhead contact network system is already in a fault and cannot operate normally. The system needs to be inspected and maintained immediately. The performance of the overhead contact network system has been significantly reduced. Ensure that the system resumes normal operation as soon as possible.

[0036] When calculating the membership, it is necessary to select a suitable membership calculation method. In this embodiment, the Gaussian membership function is used to calculate the membership of the overhead line system index. By analyzing the historical monitoring operation data of the overhead line system as sample data of different levels, the mean and standard deviation of the sample data of different levels are calculated as the center point C and standard deviation of the Gaussian membership function. : ; ; Among them, j=1,2,3,4,5 represent the system level very good, good, average, poor and very poor respectively, n is the number of sample points, is the ith sample point.

[0037] The calculated center point and standard deviation are applied to different status levels to calculate the Gaussian membership function at different levels.

[0038] The membership of the system operation status is: ; in, is the i-th sample point, is the center point, is the standard deviation.

[0039] The railway contact network system is complex, so the hierarchical analysis method is usually used to construct a multi-level analysis structure model to determine the weight of each indicator. Taking into account the dynamic characteristics of the railway contact network, this patent constructs a contact network monitoring data acquisition system. According to the system status reflected by the real-time monitoring data, the weight of each indicator in the model is dynamically adjusted. On the basis of the traditional hierarchical analysis method, the time dimension is added so that the indicator weight can be adjusted with time and environmental changes, which more accurately reflects the reliability of the contact network system.

[0040] In a specific embodiment, step 5 specifically includes: determining the evaluation time interval in days as , assign initial weights to each indicator , ,…, The initial weight reflects the relative importance of each indicator without considering the time factor. The indicator data at different time intervals are collected and the indicator weight is adjusted according to the time interval: ; ; in, is the attenuation coefficient, is the time decay function, is the importance of indicator i relative to indicator j at time t.

[0041] For each time interval , construct the judgment matrix R(t): ; Calculate the maximum eigenvalue λmax(t) of the judgment matrix for each time interval, and calculate the consistency index CI(t) and consistency ratio CR(t): ; ; Among them, RI is the random consistency index, which depends on the order of the matrix; m is the order of the judgment matrix; The judgment matrix of each time interval is checked for consistency to ensure the consistency of the evaluation. If CR(t)<0.1, the judgment matrix has satisfactory consistency; if CR(t)≥0.1, the judgment matrix needs to be readjusted; , solve for the eigenvector q, is a unit vector. After normalizing the eigenvector q, the weight vector Q(t) is: ; ; in, is the weight of criterion k at time t, is the e-th vector in the eigenvector q.

[0042] The weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion at different time points: ; Where T is the total number of time intervals.

[0043] In a specific embodiment, step 6 is specifically: The identification framework of railway contact network is H={ , , , , }, set the belief degree of each state of the system The system is in state Credibility: ; in, For the system under conditions The system is in state The probability of ; r is the number of evaluation indicators, and k is the kth indicator.

[0044] The belief functions are fused through Dempster's rule. When multiple belief functions come from different sources, they are merged into a unified belief degree: ; Among them, i is the i-th state of the system.

[0045] Based on the belief functions of multiple indicators, the fused belief degrees form a credibility distribution matrix Y: .

[0046] Through the integration of evidence theory, the probability distribution of each system state is obtained, and the probability distribution function of the system state is defined for: ; in, for The degree of belief in the state.

[0047] The railway overhead contact system is in a dynamic environment. The state probability of the overhead contact system changes with time. The state transfer matrix is ​​introduced. , and obtain the probability distribution of the contact network system state at each time step : .

[0048] In a specific embodiment, step 7 is specifically as follows: according to the state distribution of the historical operation data of the railway contact network system, the boundary points of each state are calculated to obtain the dynamic boundary value b 1， b 2， b 3， b4, maps different status levels to corresponding evaluation results according to the probability distribution of the system: Very good: the probability distribution value is higher than b1, the reliability of the system is very good, and the probability of failure is very low; Better: The probability distribution value is between [b2, b1), the system status is relatively healthy, but there are some minor faults; Normal: The probability distribution value is between [b3, b2), the system has certain faults and may need attention and maintenance; Poor: The probability distribution value is between [b4, b3), the system has a greater risk or multiple failures, and the reliability is poor; Very bad: The probability distribution value is lower than b4, the failure risk of the system is very high, and the reliability is seriously reduced.

[0049] The above embodiments are only exemplary embodiments of the present invention and are not intended to limit the present invention. Those skilled in the art may make various modifications or equivalent substitutions to the present invention within the essence and protection scope of the present invention, and such modifications or equivalent substitutions shall also be deemed to fall within the protection scope of the present invention.

Claims

1. A railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network, characterized in that: The steps include: Step 1: construct a dynamic Bayesian network model of the overhead contact network system, use the state transition characteristics of the Markov chain and the probabilistic reasoning ability of the Bayesian network to describe the state transition and fault evolution process of each subsystem at different times, conduct real-time monitoring and collection of operating data of the railway overhead contact network system, and build a real-time data collection system; Step 2, introducing the real-time monitoring data and reliability evaluation index obtained in step 1 into the overhead line reliability evaluation index system, while considering the impact of maintenance on system reliability; Step 3, classifying the reliability of the contact network system status through the historical operation data of the railway contact network and the operation status of the system; Step 4, using the Gaussian membership function to calculate the membership of the overhead line system index, by analyzing the historical monitoring and operation data of the overhead line system as sample data of different levels, and calculating the Gaussian membership function at different levels; Step 5: Determine the evaluation time interval in days, build a judgment matrix based on the importance of the indicators, and calculate the consistency ratio CR of the established matrix. If the calculated CR is less than 0.1, it is considered that the consistency of the judgment matrix is ​​within a reasonable range. If the calculated CR is greater than or equal to 0.1, the importance of the indicator is revised again. According to the judgment matrix, the weight vector is obtained, and the weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion; Step 6: Set the credibility of the system state, and combine multiple belief functions into a unified belief degree through Dempster's rule; establish a credibility distribution matrix based on the belief functions of multiple indicators, and obtain the probability distribution function of the system state through the integration of evidence theory; Step 7, divide each state according to the historical operation data of the railway overhead line system, correspond the probability distribution function of the system to obtain the state level of the system, and finally obtain the reliability evaluation result.

2. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: In the step 1, real-time monitoring and collection are performed through a monitoring data acquisition system, and the monitoring data acquisition system includes a data acquisition layer, a diagnostic analysis layer, a data storage layer, and an application execution layer. The data acquisition layer captures and feeds back the operating status data of the equipment in real time; the diagnostic analysis layer analyzes and diagnoses the data of the data acquisition layer to discover potential faults of the system; the data storage layer is used to store historical operating status data, operating analysis and monitoring of contact network equipment, and various intelligent algorithms related to equipment operation control based on training; the application execution layer converts data processing and management results into actual operation and maintenance measures.

3. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The specific process of using Markov chain for real-time monitoring in step 1 is: using a first-order Markov chain to simulate the rapid response of short-term meteorological factors and load changes, using a two-order Markov chain to integrate long-term historical information to reflect the long-term dependence of equipment aging and failure probability, defining a unified spatial state, and merging two Markov chains; Among them, the meteorological state W t and power load status L t As a short-term state variable, it is only affected by the previous time step; the equipment aging state A t and the failure probability F t As a long-term state variable, its current state depends on the states of the previous two time steps.

4. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The contact network reliability evaluation index system in step 2 includes a target layer, an index layer, and a sub-index layer. The target layer is the evaluation of the contact network operation status. The index layer is the system real-time monitoring data. , System performance indicators , Equipment status monitoring indicators and related reliability indicators The sub-index layer consists of steady-state availability, average failure frequency, average repair time, equipment maintenance rate, and contact line relative status index.

5. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The membership degree in step 4 The calculation method is: ; ; ; Among them, j=1,2,3,4,5 represent the system level very good, good, average, poor and very poor respectively, n is the number of sample points, is the i-th sample point, is the center point, is the standard deviation.

6. The railway overhead contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The step 5 specifically includes: determining the evaluation time interval in days as , assign initial weights to each indicator , ,…, The initial weight reflects the relative importance of each indicator without considering the time factor. The indicator data at different time intervals are collected and the indicator weight is adjusted according to the time interval: ; ; in, is the attenuation coefficient, is the time decay function, is the importance of indicator i relative to indicator j at time t; For each time interval , construct the judgment matrix R(t): ； Calculate the maximum eigenvalue λmax(t) of the judgment matrix for each time interval, and calculate the consistency index CI(t) and consistency ratio CR(t): ； ； Among them, RI is the random consistency index, which depends on the order of the matrix; m is the order of the judgment matrix; The judgment matrix of each time interval is checked for consistency to ensure the consistency of the evaluation. If CR(t)<0.1, the judgment matrix has satisfactory consistency; if CR(t)≥0.1, the judgment matrix needs to be readjusted; , solve for the eigenvector q, is a unit vector. After normalizing the eigenvector q, the weight vector Q(t) is: ； ； in, is the weight of criterion k at time t, is the e-th vector in the feature vector q, and r is the number of evaluation indicators; The weights of all time intervals are combined to obtain the dynamic weight of each indicator criterion at different time points: ； Where T is the total number of time intervals.

7. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The step 6 is specifically as follows: The identification framework of railway contact network is H={ , , , , }, set the belief degree of each state of the system The system is in state Credibility: ； in, For the system under conditions The system is in state The probability of; r is the number of evaluation indicators, k is the kth indicator; The belief functions are fused through Dempster's rule. When multiple belief functions come from different sources, they are merged into a unified belief degree: ； Among them, i is the i-th state of the system; Based on the belief functions of multiple indicators, the fused belief degrees form a credibility distribution matrix Y: ； Through the integration of evidence theory, the probability distribution of each system state is obtained, and the probability distribution function of the system state is defined for: ； in, for The degree of belief in the state; The railway overhead contact system is in a dynamic environment. The state probability of the overhead contact system changes with time. The state transfer matrix is ​​introduced. , and obtain the probability distribution of the contact network system state at each time step : 。 8. The railway contact network reliability assessment method based on Markov chain and dynamic Bayesian network according to claim 1 is characterized in that: The step 7 is specifically as follows: according to the state distribution of the historical operation data of the railway contact network system, the boundary points of each state are calculated to obtain the dynamic boundary value b 1， b 2， b 3， b4, maps different status levels to corresponding evaluation results according to the probability distribution of the system: Very good: the probability distribution value is higher than b1, the reliability of the system is very good, and the probability of failure is very low; Better: The probability distribution value is between [b2, b1), the system status is relatively healthy, but there are some minor faults; Normal: The probability distribution value is between [b3, b2), the system has certain faults and may need attention and maintenance; Poor: The probability distribution value is between [b4, b3), the system has a greater risk or multiple failures, and the reliability is poor; Very bad: The probability distribution value is lower than b4, the failure risk of the system is very high, and the reliability is seriously reduced.

Images