融合日类型注意力的电力负荷短期预测方法研究OA

To fully exploit the daily and weekly periodicity characteristics of power load,a short-term load forecasting method integrating a day-type attention mechanism is proposed.Firstly,the applicability of a multiple linear regression model was verified using linear regression prediction results of single-day power load,and the historical data were grouped accordingly.Secondly,an attention mechanism was intro-duced to highlight the contributions of different historical days to the target day,and a neural network-based forecasting model with integrated day-type attention was constructed.Finally,the proposed model was compared with multiple linear regression,and the Bidirectional Long Short-term Memory(BiLSTM)networks were applied to case studies.The results indicate that the proposed model outperforms the com-parative models in terms of Mean Absolute Percentage Error(MAPE),Normalized Root Mean Square Er-ror(NRMSE),and root mean square percentage error(RMSPE),demonstrating its effectiveness in predic-tion accuracy and stability.