Hybrid fuzzy clustering neural networks to wind power generation forecasting Paulo Salgado * , Paulo Afonso ** * Universidade de Trás-os-Montes e Alto Douro/ECT-Departamento de Engenharias, Vila Real, Portugal ** Universidade de Aveiro - Escola Superior de Tecnologia e Gestão de Águeda, Águeda, Portugal psal@utad.pt ; pafnaa@ua.pt Abstract— Wind power forecasting methods can be used to plan unit commitment, scheduling and dispatch by system operators and electricity traders. Because wind power is weather dependent, and therefore, is variable and intermittent over various time-scales, an accurate forecasting of wind power is recognized as a major contribution for a reliable large-scale wind power integration taking profit of economics gains. This paper explores a new approach using fuzzy clustering algorithms for obtaining one day forecast for the characteristics curves of speed wind. Moreover, a Feedforward Neural Networks (FNN) provides an estimate of the average hourly wind speed, for 24 hours horizon. I. INTRODUCTION Over the last decade there has been a rapid growth in wind generated electricity. The worldwide installed wind power capacity has increased from a total nameplate capacity of 24.3 GW to 238,351 GW in 2011, and the industry is set to grow by at least another 40 GW in 2012 [1]. The increased incidence of wind power in an energy network causes an increase of the unpredictability factor of energy production. Thus, it is difficult to predict the wind power production value, as well as its maximal or minimal values and their occurrence instants. The cause of this problem is that the wind velocity and its orientation are considered as one of the most difficult meteorological parameters to forecast. This is a result of the complex interactions between large scale forcing mechanisms such as pressure and temperature differences, the rotation of the earth, and local characteristics of the surface. The forecasting technique employed depends on the available information and the time scale in question. However one of the shortcomings in the wide use of the generation of electricity from the wind due to its intermittency regime. This factor determines the extent to which energy is produced from the wind turbine. This problem is exacerbated by the fact that wind energy cannot be stored and cannot be easily ramped up to meet load requirements [2]. To address these problems, wind power forecasting methods can be used to plan unit commitment, scheduling and dispatch by system operators, and maximize profit by electricity traders. Because wind power is weather dependent, it is variable and intermittent over various time-scales. This point makes very difficult to forecast the power which will be injected in the distribution network, which hinders the management of the networks of power centrals, in the fragile energetic balance between the production and the consumption of energy. There may also be problems in energy transportation system that connects wind farms, often placed far from the centers of consumption. A good forecast of the produced power is, therefore, very important. So, the accurate forecasting of wind power is recognized as a major contribution for reliable large-scale wind power integration. This demand of prediction accuracy motivates researchers to propose accurate short-term forecasting models of wind power. The wind power forecast should be based on the actual wind velocity forecast or on the output power of the wind turbines. Huge research is being carried out for obtaining good wind speed forecasting systems. Several mathematical models which hybridize weather forecasting models and statistical techniques have been proposed in the literature [3-4]. Also, in many cases, these systems use statistical down-scaling processes including auto- regressive models [5], artificial neural networks [6] or support vector machines [7], as a final step to improve the wind speed forecasting of the system. It may be agreed upon that wind power can be a more viable prediction parameter than wind speed for power generation purposes on the premise that predicting wind speed and converting it to power output using power curves or the following equation which relates the wind turbine’s power output to wind speed: 3 1 2 P b w P C Av   (1) where Cp is the coefficient of performance,  is the air density, Ab is the area swept by the blade and vw is the wind speed at right angles to the turbine’s blades-face. However, forecasting wind power has its limitations since it can be linked to a particular machine design or operation. Wind speed predictions is being a more logical approach for understand wind forecasting techniques for power generation. In addition, it is easier to obtain data sets for wind speed than power output with better quality. By using the power curves and equations in converting wind speed to wind turbine power output can be made through aggregate forecasting [8]. For these reasons in this paper we present a wind speed forecasting. 359 CINTI 2013 • 14th IEEE International Symposium on Computational Intelligence and Informatics • 19–21 November, 2013 • Budapest, Hungary 978-1-4799-0197-5/13/$31.00 ©2013 IEEE