Wind generation forecasting of short and very short duration using Neuro-Fuzzy Networks: A case study

Wind energy has become a good alternative to diversification and expansion of the Brazilian electric matrix. However, the random nature of the wind speed makes its forecast complex, directly affecting the planning and operation of generation systems. For a proper integration, it is necessary to formulate the prediction of resources available through reliable methodologies. Currently, there are several techniques and models that predict wind speed, which are divided in: historical data series analysis and numerical weather prediction models. On prediction of wind power generation, which uses wind speed data converted into electrical energy, there are physical, statistical methods and models based on Computational Intelligence. On this basis, this study presents a prediction methodology based on Neuro-Fuzzy Networks, which are prevenient from the union between the learning resources present in Artificial Neural Networks (ANNs) with the decision-making of human information processing, present in Fuzzy Logic. From the software Matlab
^®
, the wind generation forecasting of short and very short duration is done to qualify and quantify its contribution in national energy outlook.