A Bayesian Estimation Approach to the Post-Earthquake Recovery Trajectories of Electric Power Systems in Japan

Abstract

Post-disaster recovery modelling of engineering systems has become an important facet of disaster risk management. The post-disaster recovery trajectory of a civil infrastructure system can be quantified using (a) their initial post-disaster functionality level Qo (i.e., the ratio of the number of serviced customers/end users post-disaster to that pre-disaster); (b) rapidity h (i.e., the rate of functionality restoration); and (c) recovery time (Rt) (i.e., the total time to restore full functionality to the entire community). This study uses a Bayesian estimation approach to develop probabilistic models for characterising the relationships between seismic intensity, exposed population (PEX), Qo, h, and Rt of electric power networks (EPNs) using post-earthquake recovery data of large earthquakes occurring in Japan. Firstly, a data collection exercise was carried out to aggregate publicly available data on the aforementioned parameters for different seismic events in Japan. Based on the quality of available information, 18 strong motion events between 2003 and 2022 were selected. Next, a set of probabilistic models to estimate Qo, h, and Rt were developed using Bayesian parameter estimation to capture uncertainties. The data analysis suggests that the initial post-disaster functionality level depends on the seismic intensity and exposed population. The post-disaster recovery time is found to be dependent on the initial post-disaster functionality level, event magnitude, and year of occurrence. The rapidity of recovery is found to be dependent on the initial post-disaster functionality level. Apart from being an efficient stand-alone tool, the proposed data-driven models can be a useful benchmarking tool for simulation-based models.