Ultra-short-term Power Load Forecasting Based on Multi-head Attention Mechanism and BiLSTM

Ultra-short-term power load forecasting helps the power system to achieve reasonable dispatch and optimized operation and ensure the balance between power supply and demand by accurately predicting the power load in the next tens of minutes to hours. This is of great significance for improving the efficiency of power grid operation, reducing costs and reducing energy waste. However, when faced with ultra-short-term power loads with strong nonlinearity and fast changes, the accuracy of traditional forecasting methods is relatively low. To this end, this paper proposes an ultra-short-term power load forecasting method based on a multi-head attention mechanism and a BiLSTM neural network model. The accuracy and robustness of the model are verified using a public data set of power loads in a certain region. By the performance comparison with the traditional LSTM and BiLSTM models, the superiority of the network model proposed in this paper in ultra-short-term power load forecasting is proved.