基于序列到序列和注意力机制的超短期风速预测

ULTRA-SHORT-TERM WIND SPEED PREDICTION BASED ON SEQUENCE-TO-SEQUENCE AND ATTENTION MECHANISM

风速具有较大的波动性,给风速预测带来较大难度。为解决以上问题,该文提出基于序列到序列和注意力机制的深度学习模型来进行超短期风速预测,首先采用1维卷积神经网络和门控循环单元对风速序列数据做编码处理,得到语义向量,然后使用注意力机制和门控循环单元对语义向量做动态解码,最后输出预测值,通过反向传播算法和梯度下降算法训练模型参数。在实际采集的风速数据上对模型的预测精度和性能做评估,实验结果表明:相较于其他模型,该模型提高了超短期风速预测精度和鲁棒性,具有较好的泛化能力,验证了所提模型的有效性。

Wind is intermittent in nature,which makes it difficult to predict wind speed.In order to solve this problems,a deep learning model based on sequence-to-sequence and attention mechanism is proposed to predict ultra-short-term wind speed.Firstly,1-D convolution neural network and Gate Recurrent Unit are used to encode the wind speed sequence data and obtain semantic vectors.Secondly,attention mechanism and Gate Recurrent Unit are used to predict ultra-short-term wind speed.The semantics vectors are decoded dynamically,and the predicted values are output.The model parameters are trained by back propagation algorithm and gradient descent algorithm.The experimental results show that the forecasting model improves the accuracy and robustness of ultra-short-term wind speed prediction.Compared with other models,the proposed model has better generalization ability.