A nonlinear autoregressive neural network model for short-term wind forecasting

Integration of wind power into an electricity grid can be greatly optimized with accurate forecasting of wind speed and subsequently the power. These forecasts aid the power utilities operating in a competitive electricity market with planning and operational management of a wind generation unit. This paper presents a swift and less data hungry prediction method based on nonlinear autoregressive neural networks for short term wind speed prediction. An Artificial Intelligence (AI) method is chosen because AI techniques are considered to be more accurate than the conventional ones. The developed scheme is tested on two study sites and its effectiveness is demonstrated by comparison with a benchmark such as time series persistence. The impact of varying the size of required input data is also analyzed and it is concluded that using the developed method, minimal historical wind speed data is needed for one-hour ahead prediction.