Modeling combined global, beam, and diffuse clear-sky indices with Markov-chain mixture distribution models

Abstract

This study uses the N-state Markov-chain mixture distribution model and the multiple-component N-state Markov-chain mixture distribution model to simulate global, beam, and diffuse horizontal clear-sky index. The models are data-driven such that when trained on single or multiple clear-sky index time-series, the models generate arbitrarily long synthetic clear-sky index time-series for the same components. The models were tested on solar irradiance datasets from two different climatic regions: Norrköping, Sweden, and Oahu, Hawaii, USA. The results show high probability distribution and temporal autocorrelation goodness-of-fit for all models and high cross correlation goodness-of-fit as well as accurate correlation between the component datasets for the multiple component model simulations. When combined with, e.g., the Hay and Davies model, the output from this model could, for example, be used to generate realistic time-series of incident solar irradiance on tilted planes.