Usually, a typical wind turbine system cannot generate electricity power consistently, because the power outputs are heavily depended on wind speed. However, terrain, temperature, humidity and other factors can also affect wind speed. Therefore, wind power forecasting is a complex, multi-dimensional, and highly non-linear system. Neural network is able to learn the relationship between system inputs and outputs without mathematical conversion, and perform complex non-linear mapping, data classification, knowledge processing, and so forth. In addition, neural network also has the ability of parallel processing to reduce computing time, so it is suitable for wind power forecasting. The purpose of this paper is to use neural network technology to design a wind power forecasting system. Moreover, the efficiency analysis of the proposed wind power forecasting system in Kinmen farm is described. Finally, we use MATLAB to implement the proposed wind power forecasting system in Kinmen farm, which is capable of forecast within 48-hours ahead.