A large scale introduction of wind energy in power sector causes a number of challenges for electricity market and wind farm operators who will have to deal with the variability and uncertainty in the wind power generation in their scheduling and trading decisions. Numerical wind power forecasting has been identified as an important tool to address the increasing variability and uncertainty and to more efficiently operate power systems with large wind power penetration. It has been observed that even when the wind magnitude and direction recorded at a wind mast are the same, the corresponding energy production can vary significantly. In this work we try to introduce improvements on three different areas. Firstly, we develop a more accurate wind forecast system for a complex terrain. Secondly, we improve the wind to power conversion curve and lastly we develop a strategy to quantify uncertainties associated with the prediction. The concept has been put to practical use in the form of Norway’s first multiscale wind forecast system.