Adaptability Comparison of Three Mainstream Short-term Wind Power Prediction Methods

The intermittence and variability of wind power are the main challenges for perfect predicts. Besides, the meteorological and topological complexity makes it even harder to apply any prediction methods in specific case. In this paper, there are three mainstream wind power prediction methods to be discussed on their performance on different time and spatial scale, which are BP neural network optimized by genetic algorithm (GA-BP), radial basis function neural network (RBF) and support vector machines (SVM). The research helps users select the most suitable algorithm towards different terrains and climates, so the predicted accuracy is improved. Taking three wind farms in China with different terrains and climates as examples, the accuracy, computational efficiency and adaptability of three models are compared. Results show that RBF and SVM generally have better predicted accuracy than GA-BP model. Nevertheless, three models show advantages in different seasons and terrains. What's more, predicted model built with month as time interval can increase the accuracy of short-term wind power prediction.