STOCHASTIC GENERATION OF ARTIFICIAL WEATHER DATA FOR SUBTROPICAL CLIMATES USING HIGHER-ORDER MULTIVARIATE MARKOV CHAIN MODEL

Liquid desiccant air conditioning systems provide an efficient and less energyintensive alternative to conventional vapour compression systems due to their ability to use low-grade energy provided by a hybrid photovoltaic and thermal solar power module. Air conditioning systems are major energy consumers in buildings especially in extreme climatic conditions and are therefore primary targets in so far as energy efficiency is concerned. Building energy performance has traditionally been simulated using typical meteorological year (TMY) and test reference year (TRY) weather tools. In both cases, the value allocation is pegged on the least nonconformity from the longrange data of the past 29 years. The extreme low and high points are successively disregarded which means that the actual prevailing hourly mean settings are not precisely represented. The multivariate Markov chain provides flexibility for use in circumstances where dynamic sequential and categorical weather data for a given region is required. This study presents a simplified higher order multivariate Markov chain analysis founded on a combination of a mixture-transition and a stochastic technique to project the solar radiation, air humidity, ambient temperatures as well as wind speeds and their interrelationships in sub-tropical climates, typically the coastal regions of South Africa. The generic simulation of weather parameters is produced from 20 years of actual weather conditions using a stochastic technique. The series of weather parameters developed are then implemented in the simulation of solar powered air dehumidification and regeneration processes. The outcomes indicate that the model is devoid of constraints and more accurate in the estimation of variable parameters implying that a properly designed solar-powered liquid desiccant air conditioning system is capable of supplying the majority of the latent cooling load.