Extreme Value Analysis

Extreme value analysis (EVA) was developed from extreme value theory (EVT). It allows assessment of the probability of events being more extreme than previously observed. It is used when events have a low frequency but high severity. This chapter presents an example of the variation in the LPG splitter reflux flow. There are two methods of identifying which of the values should be classed as extreme. These are period maxima approach, and generalised extreme value (GEV) distribution. There are alternatives to the GEV distribution. These include the Gumbel, Frechet and reverse Weibull distributions. The chapter shows that, like the GEV distribution, the Gumbel distribution is a special case of the Hosking distribution. Extreme value max distribution is intended for use when the extreme values are maxima. When they are minima the distribution it is modified to the extreme value min distribution.