Short-Term Power Load Forecasting Based on a Combination of VMD and ELM

Accurate short-term power load forecasting is becoming more and more important for the stable operation and improved economic benefits of electric power systems. However, when affected by various factors the power load shows non-linear and non-stationary characteristics. In order to forecast power load precisely, we propose an extreme learning machine (ELM) combined with variational mode decomposition (VMD), as a new hybrid time series forecasting model. In the first stage, since decomposed modes and hidden layer nodes have great influence on prediction accuracy, a three-dimensional relationship has been established to determine them in advance. In the second stage, using VMD, the time series of power load is decomposed into predetermined modes that are then used to construct training parts and forecast outputs. Then every individual mode is taken as an input data to the ELM. Eventually, in the third stage, the final forecasted power load data is obtained by aggregating the forecasting results of all the modes. To testify the forecasting performance of the proposed model, a five-minute power load data in Hebei of China is used for simulation, and comprehensive evaluation criteria is proposed for quantitative error evaluation. Simulation results demonstrate that the proposed model performs better than some previous methods.