Short-Term Electricity Load Forecasting Based on Ensemble Empirical Mode Decomposition and Long Short-Term Memory Neural Network

Electricity load forecasting forms a basis for planning and economic operation of power systems, so accurate load forecasting is conducive to improving the safety and stability thereof, however, owing to load data being non-linear and fluctuating, electricity load forecasting is difficult. Therefore, a novel hybrid model based on ensemble empirical mode decomposition (EEMD) and long short-term memory neural network (LSTM), namely the EEMD-LSTM model was proposed to forecast short-term electricity load. Firstly, by using EEMD, the original load series was broken down into a residual error component (Re) and a sequence of intrinsic mode functions (IMFs) with varying frequencies. So converting the load series into a series that is comparatively stationary. High similarity components were aggregated using the sample entropy (SE) approach in order to simplify the model. Then, by using an LSTM method suitable for processing time series problem, appropriate forecasting models were established for each group of components and the sum of the forecast values of each sub-model was the initial forecast value. The final forecast value of load was the sum of initial forecast value and forecasting errors. Finally, electricity loads of a city in Liaoning Province, China in spring, summer, autumn, and winter were predicted. The model suggested in this study outperforms the other ten forecasting techniques in terms of performance and has less statistical errors, which raises the accuracy of electricity load forecasting, according to the results.