Short-Term Power Load Prediction Method Based on VMD and EDE-BiLSTM

Aiming at the diversity and flexibility of electric loads and their inherent nonlinearity and temporality in the context of the new era. A forecasting method is proposed combining Variational Modal Decomposition (VMD) and EDE-BiLSTM. Initially, the load data are decomposed using VMD to obtain several modal components. These components are then integrated with temperature, wind speed, precipitation, electricity price, and holiday characteristics to capture a comprehensive set of influencing factors. Subsequently, the modal components undergo training and learning through the EDE-BiLSTM model, aiming to minimize the test set’s root mean square error (RMSE). Each modal component is modelled individually, and the final prediction is derived by summing the predicted values of all components. The method’s effectiveness is validated using load datasets from Singapore and Australia. The model in this paper outperforms the comparison model with root mean square error (RMSE), mean absolute percentage error (MAPE),and R-squared (R2) for the two datasets respectively. The experimental results show that the proposed method performs better than other methods and can improve the accuracy of short-term power load forecasting.