基于CBAM-LSTM的风电集群功率短期预测方法
Short-term Power Prediction Method of Wind Power Cluster Based on CBAM-LSTM
张哲 1王勃2
作者信息
- 1. 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
- 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;新能源与储能运行控制国家重点实验室(中国电力科学研究院有限公司),北京 100192
- 折叠
摘要
风电功率的精准预测对我国实现"碳达峰"、"碳中和"的目标具有重要意义.传统的风电功率预测方法往往忽视了时间序列数据中的长期依赖关系和空间相关性,导致预测结果不准确.为了解决这个问题,文中提出了了卷积块注意力机制(Convolutional Block Attention Module,CBAM)和长短时记忆网络(Long Short-Term Memory,LSTM)相结合的模型.首先,使用CBAM对风电功率时间序列数据特征和数值天气预报中蕴含的空间特性进行提取,该模块能够自适应地学习时间和空间上的重要特征;然后,将提取的特征输入到LSTM层结构中进行功率预测.为了验证所提方法的有效性,使用中国吉林省某风电场的数据集进行验证,实验结果表明,与其他功率预测方法相比,文中所提方法平均绝对误差(Mean Absolute Error,MAE)平均降低2.67%;决定系数(R-Square,R2)平均提高 23%;均方根误差(Root Mean Square Error,RMSE)平均降低 2.69%.
Abstract
The accurate prediction of wind power is of great significance for China to achieve the goal of'carbon peak'and 'carbon neutrality'.Traditional wind power prediction methods often ignore the long-term dependence and spatial correlation in time series data,resulting in inaccurate prediction results.In order to solve this problem,this paper proposes a model combining Convolutional Block Attention Module(CBAM)and Long Short-Term Memory(LSTM).Firstly,CBAM is used to extract the characteristics of wind power time series data and the spatial characteristics contained in numerical weather prediction.This module can adaptively learn important features in time and space.Then,the extracted features are input into the LSTM layer structure for power prediction.In order to verify the effectiveness of the proposed method,a data set of a wind farm in Jilin Province,China is used for verification.The experimental results show that compared with other power prediction methods used in this paper,the mean absolute error(MAE)of the proposed method is reduced by an average of 2.67%.The coefficient of determination(R-Square,R2)increased by an average of 23%.The root mean square error(RMSE)decreased by 2.69%on average.
关键词
风电功率/卷积块注意力机制/长短时记忆神经网络/短期风电集群功率预测Key words
wind power/convolution block attention mechanism/long short-term memory neural network/Short-term wind farm cluster power prediction引用本文复制引用
基金项目
国家重点研发计划(2022YFB2403000)
吉林省发改委创新能力建设项目(2023C033-5)
出版年
2024
NETL
NSTL
万方数据