In order to mitigate climate change, more attention every year is being given to wind energy. However, despite minimal impact of wind turbines on the environment, there is a negative side as well. Wind speed variations are a stochastic process, and it is difficult to predict wind power accurately. Therefore, unpredictable power can disbalance the power grid; besides, huge power reserves are necessary. Wind energy can be forecasted based on statistical, physical or hybrid methods and models. However, all methods and models generate power prediction errors during different time horizons. The paper presents an analysis of wind power prediction errors determining factors based on statistical, physical and hybrid approaches. Investigation revealed that combination of statistical methods – nonlinear regression, model output statistics, the most suitable power curve and wind speed correction methods – reduced wind power prediction errors up to 1.5%. A detailed evaluation of relief variations and surface roughness increased wind power accuracy by 2%. Considering the local conditions of the western part of Lithuania, the best suitable tool for a short-term wind power prediction is a hybrid model including a detailed description of topographical conditions and the most precise statistical methods.
Marginal Bottleneck Identification in Power
System Considering Correlated Wind
Power Prediction Errors
