Stochastic approach for daily solar radiation modeling

Mathematical modeling of solar radiation continues to be an important issue in renewable energy applications. In general, existing models are mostly empirical and data dependent. In this paper, a novel approach for solar radiation modeling is proposed and illustrated. The proposed application consists of hidden Markov processes, which are widely used in various signal processing topics including speech modeling with successful results. In the experimental work, mean of hourly measured ambient temperature values are considered as observations of the model, whereas mean of hourly solar radiation values are considered as the hidden events, which constitute the outcomes of the proposed mathematical model. Both solar radiations and temperatures are converted to quantized number of states. Finally, after a training stage that forms the transition probability values of the described states, the hidden Markov model parameters are obtained and tested. The tests are repeated for various numbers of states and observations are presented. Plausible modeling results with distinct properties in terms of accuracy are achieved.