Short-Term Power Load Forecasting Method Based on Improved Sparrow Search Algorithm, Variational Mode Decomposition, and Bidirectional Long Short-Term Memory Neural Network
Ming Wen,
Bo Liu,
Hao Zhong,
Zongchao Yu,
Changqing Chen,
Xian Yang,
Xueying Dai,
Lisi Chen
Affiliations
Ming Wen
College of Electrical Engineering & New Energy, China Three Gorges University, Yichang 443002, China
Bo Liu
College of Electrical Engineering & New Energy, China Three Gorges University, Yichang 443002, China
Hao Zhong
College of Electrical Engineering & New Energy, China Three Gorges University, Yichang 443002, China
Zongchao Yu
Economic and Technical Research Institute of State Grid Hunan Electric Power Co., Ltd., Changsha 410004, China
Changqing Chen
Key Laboratory of Smart City Energy Sensing and Edge Computing of Hunan Province, Hunan City University, Yiyang 413000, China
Xian Yang
Key Laboratory of Smart City Energy Sensing and Edge Computing of Hunan Province, Hunan City University, Yiyang 413000, China
Xueying Dai
Key Laboratory of Smart City Energy Sensing and Edge Computing of Hunan Province, Hunan City University, Yiyang 413000, China
Lisi Chen
Hunan Zhongdao New Energy Co., Ltd., Yiyang 413000, China
A short-term power load forecasting method is proposed based on an improved Sparrow Search Algorithm (ISSA), Variational Mode Decomposition (VMD), and Bidirectional Long Short Term Memory (BiLSTM) neural network. First, the SSA is optimized by combining Tent chaotic mapping, reverse learning, and dynamic step adjustment strategy, and the VMD mode number and penalty factor are optimized by ISSA. Secondly, the initial load sequence is decomposed into several Intrinsic Mode Function (IMF) components using ISSA-VMD. The effective modal components are screened by Wasserstein Distance (WD) between IMF and the original signal probability density. Then, the effective modal components are reconstructed by the Improved Multi-scale Fast Sample Entropy (IMFSE) algorithm. Finally, the extracted features and IMF were input into the ISSA-BiLSTM model as input vectors for prediction.
