注意力机制下的EMD-GRU短期电力负荷预测

为进一步提高短期电力负荷预测精度,构建一种基于注意力机制的经验模态分解(EMD)和门控循环单元(GRU)混合模型,对时间序列的短期负荷进行预测.首先,对负荷序列进行EMD,将数据重构成多个分量;再通过GRU提取各分量中时序数据的潜藏特征;经注意力机制突出关键特征后,分别对各分量进行预测;最后,将各分量的预测结果叠加,得到最终预测值.仿真结果表明:相对于BP网络模型、支持向量机(SVR)模型、GRU网络模型和EMD-GRU模型,基于EMD-GRU-Attention的混合预测模型能取得更高的预测精度,有效地提高短期电力负荷预测精度.

Short-Term Power Load Forecasting Under EMD-GRU Attention Mechanism

In order to further improve the accuracy of short-term power load forecasting, a mixed model of empirical mode decomposition(EMD)and gated recurrent unit(GRU)based on attention mechanism is constructed to forecast short-term load of time series. Firstly, EMD is performed on the load sequence, reconstitute data into multiple components. Then, GRU is used to extract the underlying characteristics of time series data in each component. After the key features are highlighted by combining the attention mechanism, each component is predicted respectively. Finally, the forecast results of each component are superimposed to obtain the final predictive value. The simulation results show that the EMD-GRU-Attention mixed prediction model has higher forecast accuracy compared with BP network mode, support vector machine(SVR)mode, GRU network mode and EMD-GRU mode, and it can effectively improve the accuracy of short-term power load forecasting.