Wind Power Prediction Based on VMD-Neural Network

Abstract

Large-scale wind power integration brings severe challenges to the safe and stable operation of power system. Wind power prediction is an important way to deal with the uncertainty of wind power and promote the absorption of wind power. Because the traditional forecasting method has a single factor and limited space to improve its accuracy, it is urgent to study new theories and methods. To improve the accuracy of wind power forecasting, a new wind power forecasting method based on variational mode decomposition and neural network is proposed. This paper studies the analysis and processing methods of abnormal data of wind farms to provide high-quality basic data for forecasting. An improved K-Means clustering algorithm is proposed to cluster similar days and continuous periods of wind speed, which divides the whole year into several continuous periods with stronger regularity of wind speed, and then classifies wind power in the same period of time. Data are decomposed by VMD, and wind speed subsequences with different frequencies but stronger regularity are obtained. Finally, each subsequence is modeled and predicted based on radial basis function neural network, and wind power prediction power is obtained by superposition and wind speed-power curve conversion.