The use of Markov chains in forecasting wind speed: Matlab source code and applied case study Ionuţ Alexandru Petre 1 , Mihai Rebenciuc 2 , Ștefan Cristian Ciucu 3 1 PhD Candidate, Economic Informatics Doctoral School, The Bucharest University of Economics Studies (Piața Romană 6, București 010374), email: ionut_petre33@yahoo.com 2 PhD Lecturer, Dept. of Applied Mathematics, The Faculty of Applied Sciences, University Politehnica of Bucharest (Splaiul Independenței 313, București 060042), Romania, email: m_rebenciuc@yahoo.com 3 PhD Candidate, Cybernetics and Statistics Doctoral School, The Bucharest University of Economics Studies (Piața Romană 6, București 010374), Corresponding author, email: stefanciucu@yahoo.com Abstract The ability to predict the wind speed has an important role for renewable energy industry which relies on wind speed forecasts in order to calculate the power a wind farm can produce in an area. There are several well-known methods to predict wind speed, but in this paper we focus on short-term wind forecasting using Markov chains. Often gaps can be found in the time series of the wind speed measurements and repeating the measurements is usually not a valid option. In this study it is shown that using Markov chains these gaps from the time series can be filled (they can be generated in an efficient way), but only when the missing data is for a short period of time. Also, the developed Matlab programms that are used in the case study, are included in the paper beeing presented and commented by the authors. In the case study data from a wind farm in Italy is used. The available data are as average wind speed at an interval of 10 minutes in the time period 11/23/2005 - 4/27/2006. Keywords: Markov chain, wind speed, Matlab, Chapman-Kolmogorov, forecast 1. Introduction and literature review The market of wind energy is under development in the last years and many wind turbines installations are to be constructed in the following period, beeing considered a driver of the economy. This rapid development of the wind energy area is also due to scientifical reseach studies. Among the challenges in the domain are the understanding of the wind speed and applying the gained knowledge in the industry. Knowing the wind behavior in certain wind farms is extremely important for today's power systems, especially in the programming and operating means, wind power becoming an important part of future energy sources. Futher there will be presented some of the existing models that were studied in modeling of wind speed and a case study on one of these models will be done. Wind, solar, and biomass are three emerging renewable sources of energy, renewable energy replaceing conventional fuels. One of the benefits of this type of energy is the reduction of the production of CO 2 in the atmosphere. It can be said that global warming is also among the reasons for searching of alternative sources of energy production due to the fact that conventional sources are rich in CO 2 production. This type of renewable energy can be achieved by installing wind turbines in the areas where the wind is suitable. For this reason, forecasting and analysis of wind speed can lead to the decision whether a wind turbine for a home (micro generation) is suitable or not, or whether to invest or not into so called wind farms which can fill the need of energy. The problem of forecasting wind speed is quite known and has been dealt a lot in the literature. Among the models used in forecasting of wind speed are: Weibull and Rayleigh distribution (Aksoy 2005; Odo 2012; Ahmad 2009; van Donk 2005; Philippopoulos 2009; Ahmeda 2012), the AR (1) and AR (2) models (Aksoy 2005), the ARMA models (Philippopoulos 2009), Markov chains (Brokish 2009, Song 2011, Chen 2009, Aksoy 2005), wavelet transformation (Aksoy 2005), the Mycielski algorithm (Hocaoğlu 2009; Fidan 2012) and Weibull distribution (Akiner 2008; Ruigang 2011; Zuwei 2008; Isaic-Maniu 1983). In this paper we propose a short-term simulation of wind speed using Markov chains, namely transition matrices. In the case study it will be noticed that this technique is useful to generate data which is very close to the actual values, which makes us think that Markov chains can be used to fill gaps in the data series, when these