A statistical analysis of probabilistic functions of natural inflows
Abstract
For medium- and long-term planning of hydropower plant operations, the target observation period of several months to a year is usually divided into smaller time intervals, one week or one month long. The paper analyzes the results of several years of natural inflow measurements at one measurement site, grouped into monthly and weekly intervals. It was examined whether the theoretical probability functions of continuous distributions, which are commonly used in hydrology for inflow modeling, well approximate the empirical data of natural inflows that we had available. The compatibility of the series of average monthly and series of average weekly empirical data with the theoretical probability distribution functions was tested using the χ2 (chi-square) test. We ranked the distributions in each observation period according to the magnitude of the χ2 test. The aim of the paper is to find a type of distribution that fits well with the measured data in a large number of observed basic time periods, and is also relatively simple to use - preferably a two-parameter distribution and that the distribution parameters are determined by a simple analytical procedure. Based on this type of distribution that best describes the frequency of flow occurrence over months or weeks later, we want to build a universal inflow model in some basic time period.
