Application of ANN for Short Term Forecasting of Wind Power Density


Akıncı T. Ç. , Noğay H. S. , Guseinoviene E., Dikun J., Şeker Ş. S.

Renewable Energy and Innovative Technologies, Smolyan, Bulgaria, 10 June 2016 - 11 August 2017, vol.1, pp.157-163

  • Publication Type: Conference Paper / Full Text
  • Volume: 1
  • City: Smolyan
  • Country: Bulgaria
  • Page Numbers: pp.157-163

Abstract

Short-term wind power density forecasting of a region is an important aspect in order to affect all the decisions to be taken with regard to the power stations to be constituted and developed within the scope of the development plan to be arranged for the mentioned region. Forecasting wind power density by any means whatsoever is important, different from wind speed, with regard to the assessment of the size, type, wing shape, and wing size of the wind turbine which is essential for use in a power station to be constituted in the region. In this study, wind power density obtained from a wind turbine which is suitable for the regional conditions was forecasted in connection with the future constitution of a wind station in Siverek District of Sanliurfa, a Southeastern Anatolian Province of Turkey. Respectively, multi-layered Artificial Neural Networks (ANN) model was used. ANN model in use at the study was developed in order to forecast wind power density directly. ANN model was trained by making use of back propagation learning algorithm. So as to have ANN model trained, a data set was organized by making use of time series of wind speed data. ANN model was trained by means of 80% (7028) of 8784 wind speed data having been used in the study. In accordance with the approach (R) value, training thereof was realized by 99.83%. Having ANN model tested by 10% of the data set, a success rate of 99.79% was attained.