A Novel Short-term Residential Load Forecasting Model Combining Machine Learning Method with Empirical Mode Decomposition

Accurate load forecasting is vital for the stability and efficiency of power system’s operation. This paper proposes a model combining the empirical mode decomposition (EMD) and the neural network regression methods to forecast an individual load customer. The EMD approach is introduced to break down the load profile into multiple intrinsic mode functions (IMFs) and residuals. These components are then forecasted with the long short-term memory (LSTM) network respectively and reconstructed to form a final forecasting value. The results in case study reveals that the LSTM based on the EMD model provides a high prediction accuracy.