Application of decision tree methods for wind speed estimation

Abstract

An area's wind speed forecasting is very important to investigate whether the area is available for wind power. The wind speed estimation has been carried out by means of other machine learning methods, mostly artificial neural networks. Because, in such methods, it is aimed to estimate the wind speed with the highest accuracy along with the decimal. However, if a wind farm is to be installed, the wind speed, which is a variable in the range of 0-20 m / s, can easily be estimated with round values. If the wind speed values ​​obtained with round values ​​are forecasted with a high accuracy rate, the wind speed that is necessary for the establishment of a wind power plant in a region is obtained by a shorter and easier way. In this study, the decision tree method was used in order to reach wind speed information with an easier method and with a very short training period. Decision tree methods were examined in three different structures and three different decision tree models were designed. Additionally the estimation results of all three methods were very high, the most accurate estimation was obtained by the “Coarse Decision Tree” method which is much simpler and faster than the others.