Assessing power grid reliability using rare event simulation

Renewable energy generators such as wind turbines and solar panels supply more and more power in modern electrical grids. Although the transition to a sustainable power supply is desirable, considerable implementation of distributed and intermittent generators may strain the power grid. Since grid operators are responsible for a highly reliable power grid, they want to estimate to what extent violations of grid stability constraints occur. To assess grid reliability over a period of interest, various reliability indices exist. The main challenge of this research is to develop reliability assessment methods for a power grid given the uncertainty of power injections.

Occurrences of grid instability are typically rare in modern power grids. Using conventional Monte Carlo or Crude Monte Carlo (CMC) simulation for grid reliability estimation may therefore require a prohibitively large number of samples to achieve a sufficient level of accuracy. In this thesis we extend a CMC method with a rare event simulation technique called splitting to efficiently compute unbiased estimators for several reliability indices. We illustrate the computational gain of the splitting technique compared to CMC in several experiments. We investigate and discuss considerations when applying a splitting technique. In particular, we show that a suitable choice for the so-called importance function is crucial for achieving considerable computational gain.