中国电机工程学报2024，Vol.44Issue(14)：5451-5462,中插2,13.DOI:10.13334/j.0258-8013.pcsee.230590
基于贝叶斯深度学习的新能源多场站临界短路比区间预测方法
Multiple Renewable Energy Station Critical Short Circuit Ratio Interval Prediction Method Based on Bayesian Deep Learning
摘要
Abstract
After large-scale renewable energy is connected to the grid,the problem of voltage security and stability is prominent,so it is urgent to find a method with accuracy and practicability to evaluate the voltage support strength of the system.Therefore,an intelligent enhancement method of multiple renewable energy station short circuit ratio(MRSCR)based on Bayesian deep learning is proposed in this paper.First,focusing on the lack of accurate critical short circuit ratio(CSCR)of MRSCR,the construction process of CSCR sample set is proposed,and the batch simulation program of samples is developed accordingly.Then,the multi-gate mixture-of-experts is used to synchronously predict the CSCR of each new energy access point,and Bayesian deep learning is combined to improve the prediction accuracy and quantify the prediction uncertainty.Finally,considering the disadvantages of point estimation,an inequality method based on dynamic threshold values is proposed to provide reliable and clear interval estimation,which can provide multiple attributes of predicted values for different decision requirements.The test results on the CEPRI-FS-102 bus system show that the proposed method can effectively improve the evaluation accuracy and speed of voltage support strength,and the prediction information can provide important guidance for the decision-making process.关键词
电压支撑强度/贝叶斯深度学习/多任务学习/短路比Key words
system strength/bayesian deep learning/multi-tasking learning/short circuit ratio分类
信息技术与安全科学引用本文复制引用
李保罗,徐式蕴,李宗翰,孙华东,于琳..基于贝叶斯深度学习的新能源多场站临界短路比区间预测方法[J].中国电机工程学报,2024,44(14):5451-5462,中插2,13.基金项目
国家重点研发计划项目(2021YFB2400800).National Key R&D Program of China(2021YFB2400800). （2021YFB2400800）
