Comparative study of power forecasting methods for wind farms

This paper presents a comparative study of various forecasting models for wind power. With the growing wind power usage in the power system, wind power forecasting is very much needed to help the power system in unit commitment, economic scheduling and reserve allocation problems. Wind power forecasting using autoregressive integrated moving average model, surface fitting model, neural networks, extreme learning machine and online sequential extreme learning machine is carried out in this paper. The performance characteristics of different forecasting models have been compared by applying different measure of errors such as bias, mean absolute error, root mean square error and standard deviation. The effectiveness of online sequential extreme learning machine is evaluated on the given wind power data and the results demonstrate that the online sequential extreme learning machine performance characteristic is better as compared with the other forecasting models.