基于EWT-TOPSIS融合和波动性感知注意力的短期电力负荷预测

打开文本图片集

中图分类号：TP391 文献标志码：A 文章编号：1001-3695（2026）04-010-1046-08

doi：10.19734/j. issn.1001-3695.2025.08.0295

Short-term electricity load forecasting using EWT-TOPSIS fusion and volatility-aware attention

Chen Siyi，HuShuanglin，Yuan Bo （SchoolofAutomationandElectronic Information，Xiangtan University，XiangtanHunan41iOo，China）

Abstract：Acuratepowerload forecasting serves asacritical technologyforensuring thesafeand stableoperationof power gridsandoptimizingenergydispatchAddressingthenonliearity，non-stationarity，andcomplextemporalvaiationchacteristics of power load sequences，this study proposedaprediction model integrating empirical wavelet transform（EWT），technique fororder preference bysimilaritytoidealsolution（TOPSIS）decision-making，andvolatility-awareatention mechanism.The modelfirstly employed EWTto dynamicalldecompose load sequences and generate five intrinsic mode functions.It thendesigedulalelatilityalyrttracttiomainsatisticalatur.ubseutlyouctedilit awareattentionmechanismtodynamicallymodulatesequenceweights.Finally，itintroducedatwo-stageTOPSIS featurefusion layertoachieveoptimalintegrationofmulti-sourceheterogeneousinformation.Experimentalresultsdemonstratethatonthe Panama dataset，the modelachievesa mean absolute error（MAE）of39.87MW，root mean squared eror（RMSE）of 57.25 MW，and mean absolute percentage error （MAPE）of 3.21% . Compared to baseline models such as Crossformer and PatchTST，MAE reduces by 10.73%～67.09% and RMSE reduces by 5.81%～61.25% .On the Australian dataset，the modelobtainsMAEof191.11MW，RMSEof 244.96MW ，and MAPEof2. 06% ，with MAE reduction of 2.31% ～ 54. 57% .Ablation experiments validatethe effectivenessof thefrequency-time-statistical featurecollaborativemodelingand volatility-aware adaptive mechanism.

Key Words：short-term load forecasting;empirical wavelet transform；TOPSIS feature fusion；volatility-awareatention；deep learning； time series forecasting

0引言

电力负荷预测是电力系统规划、运行和控制的重要基础，准确的负荷预测对于保证电网安全稳定运行、优化电力资源配置和提高经济效益具有重要意义。（剩余18524字）