A Study of Forecasting Precision Enhancement of Short-term Wind Power Generation by use of NN Method

Abstract

Recently, more and more renewable power plants are connecting into the power grids, and it becomes difficult for power system operators to grasp the system operating condition exactly. Therefore, a proper forecasting method is expected to predict the short-term power generation of the dispersed generators for use of short-term distributed control, short-term operating plan of dispersed generators and electric energy storage equipments, as well as economical employment. For this purpose, in this paper, we approach the forecasting precision enhancement method for the short-term wind power generation by use of an improved NN (Neural Network), which is only based on the real-time meteorological data. By the ways of properly selection of learning data and introduction of two types of improved NN methods called NARX and Ensemble Technique, we attempt to enhance the forecasting precision for the short-term wind power generation. The results from this study show that the Mean Absolute Percentage Errors and Maximum Errors are lower than that of the base model which is based on the generally used FNN method, and thus have verified the validity of these forecasting precision improving methods proposed in this work.