基于重组二次分解及LSTNet-Atten的短期负荷预测
刘洪伟,王磊,刘阳,张鹏超,乔石

Short term load forecasting based on recombination quadratic decomposition and LSTNet-Atten
Hongwei LIU,Lei WANG,Yang LIU,Pengchao ZHANG,Shi QIAO
表 4 不同特征筛选方法的结果对比
Tab.4 Comparison of result of different feature screening methods
特征筛选方法${\hat R^{\text{2}}}$MAPE/%MAE/MWRMSE/MW
D-S优化组合-
异常值处理
0.99061.2097373.7098433.3744
D-S优化组合0.97361.2775405.8748513.1733
皮尔逊相关系数0.97451.3759423.2626504.6831
斯皮尔曼相关系数0.97041.4085427.3745542.7894
最大信息系数0.97491.3591420.7723502.0692
未处理特征0.93442.1423693.5612800.8953