风电功率超短期多步预测误差评估及分析
Evaluation and Analysis of Wind Power Ultra-Short-Term Multi-Step Forecast Error
李凤名 1车润棋 2杨茂2
作者信息
- 1. 中国石化胜利油田新能源开发中心,山东 东营 257001
- 2. 现代电力系统仿真控制与绿色电能新技术教育部重点实验室(东北电力大学),吉林 吉林 132012
- 折叠
摘要
超短期风电功率预测误差分析有助于改进预测精度,进而降低风电不确定性对电力系统带来的不利影响.以LSTM模型为例,对超短期多步预测的误差特性进行分析.首先,对超短期风电功率预测误差进行静态特性分析,研究了预测误差随预测步长的动态变化特性;然后,提出了数值天气预报在超短期风电功率预测中的误差占比定量评估模型;最后,提出了一种综合考虑形状和时间损失的神经网络损失函数,降低由于输入信息不足引起的时滞和幅值误差.结合吉林省20 个风电场的实测数据,对风电功率多步预测误差特性进行全面分析,为风电功率多步预测模型的评估、修正提供了参考.
Abstract
The ultra-short-term wind power forecast error analysis contributes to improving predictive accuracy and consequently reducing the adverse effects of wind power uncertainty on the power system.Using the Long Short-Term Memory(LSTM)model as an example,this paper analyzes the error characteristics of ultra-short-term multi-step forecasting.Firstly,the static and qualitative analysis of the ultra-short term wind power prediction error is carried out,and the dynamic characteristics of the prediction error change with the prediction step are studied.Then,a quantitative evaluation model of the error ratio of numerical weather prediction in ultra-short term wind power forecast is proposed.Finally,this paper presents a neural network loss function which takes shape and time loss into account to reduce delay and amplitude error caused by insufficient input informa-tion.Through comprehensive analysis of the multi-step wind power forecasting error characteristics based on the measured data from 20 wind farms in Jilin Province,this research provides a reference for the evaluation and cor-rection of multi-step wind power forecasting models.
关键词
超短期风电功率多步预测/误差分析/评估指标/误差分布特性/可预测分析/损失函数Key words
ultra-short-term wind power multi-step forecast/error analysis/evaluation index/error distribu-tion characteristics/predictable analysis/loss function引用本文复制引用
基金项目
国家重点研发计划项目(2022YFB2403000)
中国石油化工集团有限公司课题(JR22081)
出版年
2024
NETL
NSTL
维普
万方数据