Wind power is one of the most promising options for power generation. It ensures the diversity and security of energy supply and meets the pressing need to reduce the levels of emission of polluting gases. In the operation of electrical systems with a strong presence of wind generation, it is essential to provide at least one day in advance the future values (at least hourly) of wind speed, so that we can assess the availability of energy for the next day, a useful information in the order of the generating units and electrical control system. The purpose of this dissertation aims to develop models spe-cifically to develop models to forecast short-term wind speed, based on artificial intelligence techniques, artificial neural network model and adaptive neuro-fuzzy Systems (ANFIS) and a statistical model composed of a harmonic regression model and Box-Jenkins. For application of the methodology, the city of São João do Cariri (State of Paraíba), where a reference station of SONDA project (National Environmental Data for the energy sector) is located, was considered.To apply the methodology was consi-dered the city of the ray tracing model (State of Paraíba), which is located a station ref-erence design (National Environmental Data for the energy sector). The performance of artificial neural network model and adaptive neuro-fuzzy Systems (ANFIS) and a statis-tical model are compared mixed forecasts of 6 hours, 12 hours, 18hours and 24 hours ahead. The results show the best performance of the ANFIS model and encourage fur-ther studies on the subject.