Wind speed forecasting for power system operational planning

Wind power is a necessary addition to traditional power market. Wind power prediction therefore is necessary because of the intermittence nature of wind. A lot of studies have been performed to accurately predict wind power and local wind speed. In this paper, an artificial neural network based predictor is described to predict wind speed, which can be mathematically modeled as a highly nonlinear random process. In this new algorithm, the short-term patterns in wind speed data are grasped by artificial neural networks and the long-term pattern are classified as increasing, decreasing and almost stable. The whole process is also divided into two parts: artificial neural networks predict short-term value and the results are modified according to the long term pattern. The experimental results are compared with those of linear regression approaches. It shows the prediction-modification process improves short-term as well as long-term predictions.