Hybrid model based on VMD decomposition, clustering analysis, long short memory network, ensemble learning and error complementation for short-term wind speed forecasting assisted by Flink platform

Carving out the stochastic wind speed is still a challenge due to its intrinsic nature. With the wide spread of Internet of Things technology, the amount of data have presented explosive growth, thereby enhancing the difficulty of capturing its inherent characteristics. Therefore, the improvement of training efficiency requires keeping abreast of the forecasting accuracy and stability, which still has the large promotion space. These drawbacks motivate the propose of the hybrid model based on the variational mode decomposition(VMD), clustering analysis, LSTM network, stacking ensemble learning and error complementation for wind speed forecasting in which all the components are performed on Flink platform to ensure the forecasting efficiency. More specifically, the VMD module is employed to disintegrate the wind speed series into a primary trend and several fluctuate sub-series; Next, kmeans clustering and LSTM networks are conducted to deduce the latent characteristics of the primary trend and the stacking ensemble learning consisting of two stages is applied to infer the fluctuate abstractions of the other sub-series; Furthermore, the error complement is incorporated for assessing the error sequence created by the preliminary results. Finally, the experimental results have demonstrated that the proposed model exceeds the contrastive models on forecasting accuracy and efficiency.