A novel hybrid (Mycielski-Markov) model for hourly solar radiation forecasting

Abstract

Short term solar radiation forecasting is significant for the operation of big installations and the electrical grid. There are many short term forecasting techniques in the literature. In this study, short term predictions of solar radiation are reviewed. An alternative approach and model is proposed. The approach assumes that solar radiation data repeats itself in the history. According to this initial assumption, a novel Mycielski based model is proposed. This model considers the recorded hourly solar radiation data as an array and starting from the last record value, it tries to find most similar sub-array pattern in the history. This sub-array pattern corresponds to the longest matching solar radiation data array in the history. The data observed after this longest array in history is considered as the prediction. In case several sub-arrays are obtained, the model decides the selection according to the probabilistic relations of the sub-patterns last values to the next value. To model the probabilistic relations of the data, a Markov chain model is adopted and used. By this way historical search model is strengthened. Thanks to the proposed model, accurate predictions are obtained. To show the effectiveness, the results are compared with the models of others in the literature as a final task.