Verification of Cascading Events in Interconnected Stochastic Systems: Using Adaptive Parametric Importance Sampling Methods

Verification of Cascading Events in Interconnected Stochastic Systems: Using Adaptive Parametric Importance Sampling Methods

Cox, P.B.

Abate, A. (mentor)
Tkachev, I. (mentor)
Blom, H.A.P. (mentor)
Stano, P.M. (mentor)

Mechanical, Maritime and Materials Engineering

Delft Center for Systems and Control

Systems and Control Engineering

2013-08-13

Verification of rare cascading events in interconnected Markov processes are of high interest in energy grids, computer networks, and banking systems. Small defaults may lead to a global cascade of failures in the network. In this work, the rare events of contagious bankruptcies of the interconnected banking system are verified by three adaptive parametric importance sampling methods, belonging to Monte Carlo simulation based approaches. All three methods show comparable results in terms of the accuracy and the convergence speed on the medium sized benchmark example. Additionally, we claim that the adaptive parametric importance sampling methods scale linearly in time with respect to the sample size. The sensitivities of large or small size banks are clearly different; the larger, well connected banks have a significantly higher probability to trigger cascading; whereas for small, well connected banks this increase in probability is not found. The probability of triggering cascading is comparable between the larger sized, low connected banks and all small sized banks disregard its degree of connectivity. The results are in line with literature, which makes the adaptive parametric importance sampling methods suitable for verification of the network sensitivity.

Verification
Cascading Events
Stochastic Hybrid Systems
Importance Sampling

http://resolver.tudelft.nl/uuid:1ef9216e-d6f3-4687-88c5-d5fb9ce6b614

2018-08-13

Student theses

master thesis

(c) 2013 Cox, P.B.