Long term wind power forecast using adaptive wavelet neural network

Abstract

With the growing uncertainty due to high wind power penetration, an accurate wind power forecast tool is very much essential for economic and stable operation of the electricity markets. It helps the system operators, to include wind generation into economic scheduling, unit commitment and reserve allocation problems. It also assists the wind power producers to minimize their losses through strategic bidding in the day ahead electricity markets. In this paper the problem of long-term wind power forecast is addressed, considering the numerical weather prediction (NWP) system wind speed and wind direction forecasts as inputs. An adaptive wavelet neural network is proposed for mapping the NWP's wind speed and wind direction forecasts to wind power forecasts. Wind direction inheritantly being a circular variable, for better training and function approximation, a transformed version of wind direction variables are used as inputs. Further, a closest set of patterns based on euclidean distance are chosen for training patterns and block wise training and forecast strategy is employed for carrying wind power forecast. The results show that the significant improvement over persistence method is achieved.