An N-state Markov-chain mixture distribution model of the clear-sky index

Abstract

This paper presents an N-state Markov-chain mixture distribution approach to model the clear-sky index. The model is based on dividing the clear-sky index data into bins of magnitude and determining probabilities for transitions between bins. These transition probabilities are then used to define a Markov-chain, which in turn is connected to a mixture distribution of uniform distributions. When trained on measured data, this model is used to generate synthetic data as output. The model is an N-state generalization of a previously published two-state Markov-chain mixture distribution model applied to the clear-sky index. The model is tested on clear-sky index data sets for two different climatic regions: Norrköping, Sweden, and Oahu, Hawaii, USA. The model is also compared with the two-state model and a copula model for generating synthetic clear-sky index time-series as well as other existing clear-sky index generators in the literature. Results show that the N-state model is generally on par with, or superior to, state-of-the-art synthetic clear-sky index generators in terms of Kolmogorov–Smirnov test statistic, autocorrelation and computational speed.