A Deterministic Bases Piecewise Wind Power Forecasting Models

Abstract

Continue emphasis in mitigating the environmental impacts of fossil generated electrical energy has fuelled interest in sustainable and renewable energy; as a result of this interest, renewable energy penetration into power utilities energy mix has increased significantly. Two major issues delaying further increase of renewable energy are supply intermittency and availability. Prediction of renewable energy availability can never be over emphasized. In this paper we propose a simple nonlinear least square piecewise model to predict output power of a small Canadian wind farm. The proposed model decomposes the wind speed sweeping the wind turbine into three major speed groups, slow, moderate and fast speed. The dynamics of the wind speed in each group defines the model and the prediction error performance. We showed that the piecewise model outperformed the manufacturer’s power curve that is traditionally uses by wind farms. We present typical predictions for Fall, Winter, Spring and Summer and compared results from our proposed model to the manufacturer’s power curve. The piecewise model as well as the manufacturer’s power curve performances are both related to the skill of the wind speed estimator, accurate wind speed estimates will result to excellent forecast for both models.