The volatility in wind characteristics like speed and direction creates difficulty in wind farm layout optimization, designing energy storage systems, fluctuations on the grid and loss to businesses due to sub-optimal planning. Accurate forecasting of wind characteristics has the potential to resolve these issues. To capture the overall trend of wind characteristics, the wind data are split into trend and seasonality using STL decomposition method. The trend and remainder part are modelled together using least square support vector regression (LSSVR) whose hyper-parameters are tuned using a recently proposed algorithm from the literature. The accuracy of the speed and direction model is 0.9752 and 0.9792, respectively. These models are then used to forecast the future speed and direction values. Using the forecasts, frequency maps are generated for the wind farm with a layout selected from the literature. The generated power from the wind farm is noted to be 11,535.97 kW when compared to the benchmark value of 11,548.02 kW.