Wind Farm Power Prediction Based on Wind Speed and Power Curve Models

Abstract

Accurate prediction of wind farm power is essential for increasing wind penetration in the electricity grid. It also aids the power system operators in planning unit commitment, economic scheduling, and dispatch. In this paper, combined wind farm power prediction models have been built based on wind speed prediction models and power curve models. The wind speed prediction models have been built using nonlinear autoregressive models with and without external variables. The wind turbine power curve has been modeled using parametric and nonparametric models. Parametric models are built using four- and five-parameter logistic expression, whose parameters are solved using particle swarm optimization (PSO) and differential evolution (DE). Nonparametric models were built based on data mining algorithms. Multistep prediction model for wind power forecasting has been developed for very short-term forecasting of wind power. Real-time data obtained from Sotavento Galicia Plc. has been used for testing the proposed model.