This paper addresses the wind power forecasting problem from the standpoint of performance and solution requirements. These requirements stem from the need to trade wind power on real-time energy markets, the need for reliable and accurate forecasts, the ability to learn from new data as it becomes available, the ability of the solution to generalise and scale to a large number of wind farms, and the need to compute the solutions in real time. An analogs-based solution methodology is implemented and demonstrated to show potential in fulfilling the solution requirements. Analogs-based modelling represent a non-parametric modelling solution which exploit similarities in historical data with the current forecasting window. The methodology is applied to both local wind speed forecast correction and, subsequently, to wind farm power forecasting. Results for two wind farms in Europe are presented, confirming analogs as a viable and attractive solution methodology.