To discuss the possibility of applying a numerical Weather Research and Forecasting (WRF) model to wind power forecasting, the summer and winter hourly wind speed forecasting performance of the WRF model used in Rudong, Jiangsu Province, is analyzed in this paper. It is determined that the WRF model can forecast winter wind speed in Rudong 24 h in advance very well with a correlation coefficient as high as 0.61 compared to observations and is statistically significant at the 0.01 level. Similarly, the forecasts with leading times of 48 h and 72 h are also sufficiently accurate, according to the correlation coefficients. In summer, however, forecasting is less accurate, although the correlation coefficient of the 24 h leading forecast is 0.59, which is still statistically significant at the 0.01 level. The forecasting of wind speed in Nantong was also examined with observed data and was determined to be more accurate than that in Rudong in both winter and in summer. For a larger domain, the performance of wind forecasting is also analyzed. Results show that the performance of the WRF model is better for ocean than for land and is better in winter than in summer. In general, the forecasted wind speed is greater than the actual wind speed and the performance of the WRF model foresting differs significantly according to location and season. Essentially, short-term wind forecasting is valuable and is worthy of further investigation.