基于特征优选和麻雀搜索优化门控循环单元短期风电功率预测  

Short-term wind power forecasting based on feature optimization and sparrow search algorithm-gated recurrent unit

作　　者：胡道波 陈芳芳[1] 张倩倩 文博 罗银榕 HU Daobo;CHEN Fangfang;ZHANG Qianqian;WEN Bo;LUO Yinrong(School of Electrical and Information Technology,Yunnan Minzu University,Kunming 650504,China)

机构地区：[1]云南民族大学电气信息工程学院,云南昆明650504

出　　处：《应用科技》2023年第6期63-68,共6页Applied Science and Technology

摘　　要：针对短期风电出力预测目前存在的难点与问题,提出一种基于特征优选和麻雀搜索优化门控循环单元(gated recurrent unit,GRU)神经网络预测模型,实现对风电功率的短期预测。首先,分别使用Kendall秩相关系数、灰色关联度和互信息对原始特征数据进行特征优选,选择有效特征作为输入特征集;其次,使用麻雀搜索算法(sparrow search algorithm,SSA)对GRU神经网络超参数进行优化,获取最优超参数;最后,结合西北某风电场实测数据验证了该方法的有效性。实验结果表明,本文所提出的预测模型与文中其他传统预测模型相比,均方根误差和平均绝对误差平均下降了25.3%和31.3%,拟合优度系数平均提高了10.2%,表现出更好的预测性能。Aiming at the current difficulties and problems in short-term wind power output prediction,a gated recurrent unit(GRU)neural network prediction model based on feature optimization and sparrow search is proposed to achieve short-term wind power prediction.Firstly,Kendall rank correlation coefficient,grey correlation degree and mutual information are used to optimize the original feature data,and the effective feature is selected as the input feature set.Secondly,sparrow search algorithm(SSA)was used to optimize the GRU neural network hyperparameters and obtain the optimal hyperparameters.Finally,the effectiveness of the proposed method was verified by the measured data of a wind farm in northwest China.The experimental results show that compared with other traditional prediction models,the root-mean-square error and average absolute error decrease by 25.3%and 31.3%on average,and the goodness of fit coefficient increases by 10.2%on average,showing better prediction performance.

关 键 词：风力出电 短期预测 特征优选 麻雀搜索算法 优化门控循环单元神经网络 互信息 灰色关联度 Kendall秩相关系数 

分 类 号：TM614[电气工程—电力系统及自动化]