Short-Term Wind Power Forecasting on Multiple Scales Using VMD Decomposition, K-Means Clustering and LSTM Principal Computing

• Main Authors: Zexian Sun, Shenshen Zhao, Jingxuan Zhang
• Format: Article
• Language: English
• Published: IEEE 2019-01-01
• Series: IEEE Access
• Subjects: Wind power forecasting; variational mode decomposition; K-means clustering; long short term memory network
• Online Access: https://ieeexplore.ieee.org/document/8843987/

Wind power plays a crucial role in the secure conversion and management of the power system. Therefore, this study proposes a hybrid model for short-term wind power forecasting, which consists of the variational mode decomposition(VMD), the K-means clustering algorithm and long short term memory(LSTM) network.The combination model is conducted as follows: the VMD decomposes the raw wind power series into a certain number of sub-layers with different frequencies; K-means as a data mining approach is executed for splitting the data into an ensemble of components with similar fluctuant level of each sub-layer; LSTM is adopted as the principal forecasting engine for capturing the unsteady characteristics of each component. Eventually, the forecasting results would be generated by aggregating the predicted components.To evaluate the fitting capacity of the proposed model, seven different models including the back propagation neural network(BP) approach, the Elman neural network(ELMAN), the LSTM approach, the VMD-BP approach, the VMD-Elman approach, the VMD-LSTM approach and the VMD-Kmeans-LSTM approach are implemented on four wind power series for multiple scales. The experimental results demonstrate the best performance in favour of the proposed model.