融合编码解码和线性回归的短期风速预测

风电场风速的准确预测对电力系统的稳定运行至关重要。提出了一种融合编码解码和线性回归的短期风速预测模型。为了减小前一时刻的风速数值与预测值的强相关性,采用多分类建模方法构建编码解码结构的多步风速预测模型。编码解码模型的风速预测输出值与线性回归风速预测值进行加权,获得最终的风速预测结果。在近五年内的风速数据上进行实验的验证与测试。结果表明,相比较已有的编码解码模型、线性模型、神经网络模型等四种风速预测方法,提出的模型在短期风速预测方面,具有更优的预测性能。

A short-term wind speed prediction model is proposed which combines an encode-decoder model and a linear regression model.In order to reduce the strong correlation between the previous wind speed and the predicted wind speed,an encoder-decoder method of the multi-class prediction is applied to construct a multi-step wind speed prediction model in this paper.To obtain the final predicted wind speed,the wind speed prediction outputs of the encoder-decoder model and the linear regression model are weighted.Extensive experiments are conducted on five-year wind speed data.