Development of a Wind to Power Model for Wind Farm Power Production Forecasting

Abstract

The objective of this study is to develop a Wind to Power forecasting methodology where wind farm power curves are used to convert the predicted wind speeds to predicted power productions of wind farms. A methodology is proposed to develop a wind farm power curve by performing several experiments using historical wind farm power productions and wind measurements. A simple method, the Bins Method, is compared to an advanced method based on Artificial Neural Networks (ANN). It is shown that the advanced method does not have significant advantages in terms of accuracy. However, due to its flexibility on the choice of input parameters, the ANN method is best suited for studies such as the optimal selection of input parameters. For this reason, these two methods are used alternatively in this study according to the applications. The influence of different meteorological parameters on the wind farm power curve has also been investigated with the ANN method. The wind farm power curves developed in this work have been tested for the prediction of the wind power productions using forecasted wind speed data obtained from a Numerical Weather Prediction model (GEM-LAM) as input variables to the Wind to Power model. The predicted power productions compare well with those recorded at the wind farm. The proposed method is applied to a second operating wind farm to demonstrate its validity.

On the Improvement of Wind Power Predictions Based on Terrain Characteristics and Measurements of the Annual Energy Production

“…For example, Rehman et al [6] [7] presented wind resource assessments, along with the design and the economic assessment of a 20 MW wind power plant, using 2 MW turbines, located in the Eastern region of Saudi Arabia. Specific to wind farms, Touani et al [8] developed of a Wind to Power Model for wind farm power production forecasting, where wind farm power curves are used to convert the predicted wind speeds to predicted power productions of wind farms. In their work, they have however not considered the site characteristics of the wind farms.…”

Wind power forecasting approach using neuro-fuzzy system combined with wavelet packet decomposition, data preprocessing, and forecast combination framework

“…The physical models try to estimate first the speed of the wind using numerical weather predictions (NWPs) and some information about the site (terrains, topography, roughness, obstacles) and then to estimate the generated power at the farm (Touani et al, 2013). In comparison with the statistical modeling, the physical models need an important computation time, which is not appropriate for providing very short-term predictions (Jiang et al, 2013).…”