Open AccessApplication of Variational Mode Decomposition and Deep Learning in Short-Term Power Load Forecasting
Author(s) -
Ping Yu,
Jie Fang,
Xu Yang,
Qiang Shi
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1883/1/012128
Subject(s) - modal , term (time) , sequence (biology) , computer science , mode (computer interface) , component (thermodynamics) , power (physics) , electric power system , series (stratigraphy) , nonlinear system , hilbert–huang transform , artificial intelligence , algorithm , deep learning , control theory (sociology) , telecommunications , paleontology , chemistry , physics , thermodynamics , control (management) , white noise , quantum mechanics , biology , polymer chemistry , genetics , operating system
Accurate load forecasting of power system operation and development is of great significance. Because of the power load time series has strong nonlinear, the traditional forecasting model does not apply. Therefore, a short-term load forecasting model based on variational model (VMD) and length (LSTM) is proposed. Firstly, the VMD decomposes the original load sequence to get the modal of different size frequency component. Then, phase space reconstruction (PSR) organizes the modal components into deep learning inputs. Then, the LSTM network is employed to predict each group of modal components. Finally, all the modal component predictive value addition to the power load to predict the future. The experimental results show that compared with the BP, LSTM and EEMD - LSTM model, the model completely weakens the non-stationary load sequence, minimize the prediction error, reached the highest prediction accuracy.