Comparison of energy production between observed and predicted wind speed series with artificial neural networks

Making a prediction model on possible wind energy systems is a physical modelling including several approaches: numerical methods from global to local scales, ensemble forecast, upscaling and downscaling methods; the prediction is implemented also with statistical methods and learning machines, or benchmarking techniques and uncertainty analysis. After recognizing the structure of human brain computation, a new studying area of artificial neural networks conducted. A neural network is a structure, which is modelled according to the brain performing system on a particular task to solve and neural network can be applied using either some electronic parts or designing a model thanks to using of software. During this study, Matlab ANN codes used in order to design artificial neural networks. During the first part of the calculation performance analysis of ANN was studied, by making a comparison between observed wind speed data and predicted wind speed data by the designed network. Additionally, energy calculation was studied considering both observed wind speed data and predicted data by using WAsP (Wind Atlas Analysis and Application Program). Furthermore, the best network was used for short term future prediction of wind speed. During the project, ITU Maslak Campus wind speed and direction data are used which have been observed between the years of 2001 and 2004.