基于注意力机制的VMD-CNN-LSTM短期风电功率预测

VMD-CNN-LSTM Short-Term Wind Power Prediction Based on Attention Mechanism

摘要: 针对传统物理机理驱动预测风电功率的方法存在预测结果精确度欠佳、泛化能力弱的问题，提出一种基于注意力机制的VMD-CNN-LSTM短期风电功率预测方案。首先采用变分模态分解算法将风电功率序列分解并进行重构。然后利用注意力机制对每个特征分配不同权重。最后通过CNN-LSTM组合网络对每个分量进行训练和预测并重构后输出预测结果。实验结果表明，基于注意力机制的VMD-CNN-LSTM模型在风电功率预测方面具有更高的预测精确度和泛化性。

Abstract: Aiming at the problems of poor prediction accuracy and weak generalization ability of the traditional physical mechanism-driven wind power prediction method, a VMD-CNN-LSTM short-term wind power prediction scheme based on attention mechanism is proposed. First, the variational mode decomposition algorithm is used to decompose and reconstruct the wind power sequence. Then use the attention mechanism to assign different weights to each feature. Finally, the CNN-LSTM combination network is used to train and predict each component and output the prediction result after reconstruction. The experimental results show that the VMD-CNN-LSTM model based on the attention mechanism has higher prediction accuracy and generalization in wind power prediction.