基于数据重采样与GRU神经网络的风电功率多步提前预测  

Multi-step Advance Forecast of Wind Power Based on Data Resampling and GRU

作　　者：胡珈宁 王旭 周振雄[1] HU Jianing;WANG Xu;ZHOU Zhenxiong(Electrical and Information Engineering College of Beihua University,Jilin 132021,China)

机构地区：[1]北华大学电气与信息工程学院,吉林吉林132021

出　　处：《北华大学学报（自然科学版）》2024年第5期688-693,共6页Journal of Beihua University（Natural Science）

基　　金：吉林省科技发展计划项目(YDZJ202303CGZH001)。

摘　　要：准确预测不同时间尺度风电功率对于实现能源管理系统可靠运行至关重要。针对当前预测方法随着步数增加无法保持较高预测精度的问题,提出一种数据重采样技术与GRU神经网络相结合的风电功率多步提前预测方法;利用数据重采样技术对原始风电功率时间序列重新采样,得到新的风电功率时间序列;通过GRU神经网络对重新采样的时间序列进行单步提前预测,实现对原始风电功率时间序列的多步提前预测。利用澳大利亚某风力发电厂2022年、2023年数据进行试验,结果表明,本文方法比已有方法的平均绝对百分比误差和均方根误差至少降低了1.94%和6.13,具有更好的预测结果。Accurately predicting wind power at different times and scales is crucial for the reliable operation of energy management systems.Aiming at the problem that current prediction methods can not maintain high prediction accuracy with the increase of steps,a multi-step advance forecast of wind power method based on data resampling and GRU is proposed by combining data resampling technology with GRU neural network.A new time series of wind power is obtained by using data resampling technology to resample the original time series of wind power;Using GRU neural network to perform one-step advance prediction on the resampl ing time series,achieving multi-step advance prediction of the original wind power time series.Experiments were conducting by using data from a wind power plant in Australia in 2022 and 2023,and the results showed that the proposed method outperformed existing methods in predicting results,the mean absolute percentage error and root mean square error were reduced by at least 1.94%and 6.13.

关 键 词：风电功率预测 数据重采样 GRU神经网络 多步预测 

分 类 号：TM614[电气工程—电力系统及自动化]