摘要
为更准确预测短期风电功率,提出了一种基于误差修正的NNA-ILSTM短期风电功率预测方法。首先,采用斯皮尔曼(Spearman)等级相关系数法对风电功率影响因素分析,选出相关性较高的参量;其次,对长短期记忆网络添加注意机制与修改损失函数以解决其对有效信息筛选不足的问题,利用神经网络算法(NNA)优化改进的长短期记忆网络(ILSTM)中的神经元数量和时间步长,提高其预测精度以及泛化能力,构建NNA-ILSTM预测模型;最后,分析预测误差与风电功率、风速之间相关性,构建误差修正模型,对NNA-ILSTM模型预测结果进行修正,得到风电功率预测的最终结果。实验结果表明,所提出的模型可以显著提高风电功率预测精度。
In order to improve the accuracy of short-term wind power forecasting,the paper proposes a NNA-ILSTM short-term wind power prediction method based on error correction.Firstly,the factors influencing the wind power are analyzed by Spearman rank correlation coefficient method,and the parameters with high correlation are selected.Then an attention mechanism is added to the longterm and short-term memory network and loss function is modified to solve the problem of the insufficient screening of effective information.The NNA is used to optimize the number and time step of neurons in the ILSTM,so as to improve its prediction accuracy and generalization ability and build the NNA-ILSTM prediction model.Finally,the relationship between the prediction error and the wind power and wind speed is analyzed,the error correction model is constructed,the prediction results of the NNA-ILSTM model are corrected,and the final results of the wind power prediction are obtained.The experimental results show that the proposed model can significantly improve the accuracy of the wind power prediction.
作者
赵铁成
谢丽蓉
叶家豪
ZHAO Tiecheng;XIE Lirong;YE Jiahao(School of Electrical Engineering,Xinjiang University,Urumqi 830047,China)
出处
《智慧电力》
北大核心
2022年第1期29-36,共8页
Smart Power
基金
国家自然科学基金资助项目(62163034)。
