Multivariate load prediction method for integrated energy system based on CEEMD-LSTM

Abstract

Accurate load prediction is increasingly important for the economic dispatch of the integrated energy system. A multivariate load prediction method based on Complete Ensemble Empirical Mode Decomposition and Long Short Term Memory network (CEEMD-LSTM) is raised. Pearson correlation coefficient was used to analyze the influencing factors, and the influencing factors with high correlation degree were screened out. CEEMD was used to decompose cold, heat and electric load sequences into the mean components of Intrinsic Mode Function (IMF), and the ones with highly related to load forecasting are screened and retained. The selected IMF mean components and influencing factors were input into the LSTM model for training and learning to obtain the final CEEMD-LSTM load forecasting model. The simulation results are verified by an integrated energy system in a province. Compared with other load forecasting methods, the proposed load forecasting method has higher prediction accuracy, and considers the difference between cold, heat, electric load and the correlation of influencing factors.