Research on Multi-Step Short-Term Power Load Forecasting Based on Optimized VMD Signal Processing and PSO-SVR

Accurate load forecasting is beneficial for the rational scheduling of power generation energy, meeting people's daily electricity needs, and facilitating the steady development of the Chinese economy. It is difficult to predict short-term load demand using historical load data, mainly due to t...

Full description

Saved in:
Bibliographic Details
Main Authors: Tan, Xuebiao, Wang, Jing, Ding, Dan, Kang, Lun, Fu, Wentian, Wang, Haiwen, Ran, Yunzheng, Zhong, Jianwei
Format: Conference Proceeding
Language:English
Subjects:
Accuracy
Computational modeling
Correlation
Load forecasting
Load modeling
Modal optimization reorganization
Particle swarm optimization
Particle swarm optimization algorithm
Prediction algorithms
Predictive models
Short-term load forecasting
Signal processing algorithms
Support vector machines
Variational modal decomposition
Vectors
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Accurate load forecasting is beneficial for the rational scheduling of power generation energy, meeting people's daily electricity needs, and facilitating the steady development of the Chinese economy. It is difficult to predict short-term load demand using historical load data, mainly due to the difficulties caused by its nonlinear characteristics. Therefore, this article proposes a VMD-PSO-SVR short-term forecasting model for the power load of an industrial park. This model uses Variational Mode Decomposition (VMD) to split the original historical load data into several nonlinear subsequences. Then, selective reconstruction is performed based on the degree of correlation between the subsequences and the original sequence to form several new subsequences. Next, the easily predictable subsequences are input into a support vector regression prediction model optimized by particle swarm optimization (PSO) for prediction. Finally, the predicted results of each component are reconstructed to obtain the final prediction result. A case study in an industrial park concludes that the average absolute percentage error (MAPE%) and goodness of fit test (R2%) evaluation indicators of the VMD-PSO-SVR prediction model are 0.6415% and 99.7476%, respectively. Compared with other models, it can more accurately predict future short-term load demand.
ISSN:2770-663X
DOI:10.1109/ICISCAE62304.2024.10761644