A data-driven integrated framework for predictive probabilistic risk analytics of overhead contact lines based on dynamic Bayesian network

Abstract

Due to completely working under open-air conditions without backup equipment, the overhead contact lines (OCLs) are suffering from external extreme weather conditions, except for the long-term dynamic vibrations of catenary-pantograph system. These risk factors are prone to cause failures of OCL components and power outages, which may further result in transportation interruptions, enormous economic losses, serious social impacts, and even catastrophic safety accidents. To comprehensively investigate the associated risks in OCLs, a data-driven integrated predictive probabilistic risk analytics framework based on dynamic Bayesian network is proposed to identify the significant risk factors and analyse the time-dependant failure patterns in dynamic risk propagation network of OCLs. After exploiting the weather-driven analytics for failure probability prediction, an integrated failure probability modelling for OCL components is developed, simultaneously incorporating internal, and weather-driven hazard factors of OCLs. A predictive risk metric is suggested based on the newly established risk propagation network of OCLs, which can account for weather hazards, system failures, financial costs, and social trust losses, in response to external weather conditions over time. Numerical studies conducted on the actual OCLs demonstrate that the proposed framework can dynamically model and evaluate risks of failure patterns that expose to OCLs.