We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
基于 PSO-SVR 模型的短期天然气负荷预测.
- Authors
杨奕; 刘金源; 陈天民; 魏王颖; 王寿喜
- Abstract
In view of the difficulty in selecting influencing factors of natural gas load, redundancy factors will affect the accuracy and speed of load prediction, feature selection relief algorithm was used to screen influencing factors of load value and remove redundant influencing factors, so as to provide certain data basis for accurate natural gas load prediction. When performing load forecasting, a single support vector regression (SVR) was difficult to select the optimal selection of important parameters. In order to solve this problem, a composite model of particle swarm optimization (PSO) based on feature selection was proposed to optimize important parameters of SVR. Firstly, the proposed algorithm used feature selection to select influencing factors, which provided the main data support for load prediction. Then, the initial values of the three key parameters of SVR were set, and the optimal key parameter values were obtained through iteration. Finally, the influence factors and load values were input into the PSO-SVR model for training and prediction. Load data of Yuji pipeline were used to forecast and compare. The results show that the prediction accuracy of the proposed algorithm is higher than that of other single models, which can provide reference for the research of natural gas load prediction and provide basis for natural gas companies to purchase gas.
- Publication
Science Technology & Engineering, 2023, Vol 23, Issue 35, p15210
- ISSN
1671-1815
- Publication type
Article