Reliability Estimation And Dynamic Risk Prediction For Overhead Contact Lines Of High-speed Railway

Up to the beginning of 2022,the total operating mileage of China high-speed railway has exceeded 40,000 km.Overhead contact lines(OCLs)are playing the prominent role in obtaining electrical energy for high-speed trains,whose operational reliability is closely associated with the safety of high-speed railway.Due to complex structure with multiple mechanical components and fixed installations without backup,this specific outdoor operation mode makes OCLs not only work under continuous vibration caused by pantograph excitation during train running,but also vulnerable to external extreme weather and environmental factors.These internal and external triggers are more likely to lead to the frequent occurrences of OCL failures and a series of serious consequences,such as affecting traffic safety,train interruption,delay,outage,etc.,even causing serious safety accidents and risks.Therefore,it is significantly indispensable to investigate the failure probabilities of OCL equipment,comprehensively consider failure mode,risk propagation,and severity of consequence,perceive its future development trend according to the current operating state,and evaluate the overall operating risk of OCLs,which can improve the ability of catenary risk prevention and guarantee the reliability of the high-speed railway transportation.In this context,on the basis of requirements for operation and maintenance of OCLs and OCL failure records,this thesis will explore the formation mechanism of OCL failures,reveal the influence laws of different causes,reduce the operational risks of OCLs,and provide theoretical basis for the design of OCLs and “predictive maintenance”.First,the related work related to reliability analysis,failure risk prediction of external weather factors,risk assessment of OCLs is provided,and further the research framework of this thesis is proposed.Based upon the reliability theory,data driven techniques,and risk theory,this thesis focuses on the reliability assessment,failure risk prediction of external weather factor,and risk assessment of OCLs.Secondly,aiming at the current maintenance mode with a corrective maintenance and preventive maintenance,Chapter 2 proposes Bayesian network(BN)model incorporating common cause failure(CCF)and imperfect maintenance to evaluate the operational reliability of OCLs.The fault tree model is established for multi-component repairable OCLs,and then the BN model is applied to evaluate the reliability of OCLs and OCL components,simultaneously considering the CCF and imperfect maintenance.Then the reliability indices of OCLs and OCL components are derived.A comparison of computational results is conducted under whether considering CCF and imperfect maintenance or not,respectively.The proposed approach can not only identify the weak point of OCLs,but also is more in line with the actual operation and maintenance scenarios.Thirdly,given that the OCLs are more vulnerable to lightning strike,the operational reliability of power supply is significantly impacted,and existed studies fail to investigate spatiotemporal difference characteristics of lightning-related trip-out of OCLs and lightning characteristic,Chapter 3 develops a data-driven lightning-related trip-out prediction approach for OCLs.After investigating the spatiotemporal difference characteristic of lightning-related tripout of OCLs and lightning movement process,four key lightning parameters are selected to reveal the stochastic nature of lightning strike,including lightning peak current,ground flashover density,maximum steepness of lightning peak current,and the distance between OCLs and lightning strike point.A new lighting intensity index is proposed and the occurrence probability of lightning strike is modelled.The spatiotemporal fragility model is proposed to capture the spatiotemporal dependencies between lightning intensity and OCL trip-out rate,combined with actual OCL trip-out records.Furthermore,the BN model is utilized to implement lightning-related trip-out prediction for OCLs.The effectiveness of the proposed approach is validated using experiments,which can capture the spatiotemporal difference characteristic of lightning-related trip-out of OCLs and work acceptably on noisy lightning data with a signalto-noise ratio of 15 d B or higher.Fourthly,considering that wind-caused floater intrusion has posed enormous threats to the safety of OCLs and there is little research,Chapter 4 proposes to wind-caused floater intrusion prediction for OCLs based on Bayesian neural network.To account for the uncertainty in wind variables,the probabilistic wind model is proposed to analyze joint probability density function of wind speed and direction.An environment sensitive parameter of floater intrusion for OCLs is designed,which can integrate the characteristic parameters of OCLs,and local environment into correlation investigation between floater intrusion risk level and wind weather condition.Furthermore,four weekly wind features are selected as the input of prediction approach to implement the wind-caused floater intrusion risk prediction,including weekly average wind speed,weekly maximum wind speed,weekly sum of the angle between OCL alignment and wind direction,and weekly sum of environment sensitive parameter.The proposed prediction approach is validated using the constructed dataset.It has the capacity of presenting great robustness in floater intrusion risk prediction with imbalanced and small dataset,and can estimate the prediction uncertainty.Fifthly,because the surface of OCL porcelain insulator is more likely to be polluted by fog-haze,then leading to flashover,Chapter 5 proposes a fog-haze-related pollution flashover prediction approach for OCLs,based on Stochastic Gradient Hamiltonian Monte Carlo inference for deep Gaussian process(SGHMC-DGP).By analyzing the significant factors related to pollution degree of OCL insulator under fog-haze,five key parameters are selected,including wind speed,PM 2.5,humidity,air quality index,and temperature.After normalization,these key features are input into the proposed approach to predict the fog-haze-related pollution flashover probability of porcelain insulators.The effectiveness of the proposed approach is demonstrated by experiments.In addition,it can not only capture the prediction uncertainty from limited fault samples,but also provide a reference for establishing the short-term forecast mechanism of haze pollution flashover.Finally,owning to multiple triggers of OCL failures,and complexity and variety of failure modes and risk propagation,Chapter 6 studies the predictive risk assessment for OCLs based on dynamic BN and the fusion of internal and external factors.Based on Chapter 2,the failure mode and effect analysis(FMEA)method is applied to analyze failure modes and consequences of OCLs.At the same time,combined with Chapter 3,4,and 5,risk propagation chain is investigated from the perspectives of three external weather conditions.In such case,the network of OCL failure risk propagation is established based on the fusion of failure rate of OCL components and external weather-related failure risks.Then the risk consequences are calculated and quantified from perspectives of economics loss,and social trust loss constrained by power outage time and train timetable.To this end,when obtaining new weather data with continuous variations,it is priority to predict the weather-related failure probability and then the risk assessment model based on dynamic BN is proposed to implement the predictive risk evaluation for OCLs.The numerical experiments demonstrate the proposed approach can fuse the internal and external failure risk and reflect the time-varying characteristic of operational risks,which is valuable to facilitate the safety operation of OCLs.In this thesis,various data-driven techniques are utilized to establish a framework of reliability estimation and risk dynamic prediction for OCLs,which provides a significant reference and a valuable insight for the intelligentization of prognostics and health management of OCLs of high-speed railway.