A novel method for forecasting surface wind speed using wind-direction based on hierarchical markov model

Abstract

This article presents a new method for detecting heterogeneities in wind data set to predict wind speed based on the well-known Hidden Markov Model (HMM). In the proposed method, the HMM categorizes the wind time series into some groups in which each group represents a wind regime. Each regime uses an internal first-order Markov Chain (MC) for forecasting, and the combination of all regimes outputs generates the final wind speed forecast. The model proposed in this study is called “Hierarchical Markov Model”. The first layer detects and separates wind regimes as heterogenic groups of wind data by the use of wind direction data, based on HMM, and the second layer forecasts the wind speed using MC. The proposed model is implemented and tested using real data. Its effectiveness in terms of temporal stationary index is compared with that of a first-order MC-based method. The results showed that more than 70% improvement can be achieved in wind speed prediction by the proposed method. Moreover, it gives a probability distribution function of wind speed prediction, which is sharper than the one obtained with the first-order MC; means that more precise prediction.