摘要
针对径流式小水电日发电量预测精度较低而导致的发电计划不合理和弃电问题,提出一种基于变分模态分解(Variational mode decomposition,VMD)和双向长短时记忆网络(Bi-directional long short-term memory,BiLSTM)的日发电量组合预测方法。首先分析了小水电的日发电量特性,基于此采用VMD技术将日发电量分解为低频和高频分量,以提取时间序列数据周期趋势特性和局部波动特征,采用中心频率法和皮尔逊相关系数法确定VMD的分解次数;最后,采用基于日发电量特性选择的BiLSTM预测工具对VMD分解得到的各个分量进行预测,再将各个分量的预测结果进行融合,得到最终的预测值。实验结果表明,所选模型能较好地拟合实际数据,提高预测精度,并优于其他深度学习方法。
Aiming at the problems of unreasonable power generation plan,power abandonment caused by the low accuracy of daily power generation prediction of run-of-river small hydropower,this article proposes a combined daily power generation prediction method based on variational mode decomposition(VMD)and bi-directional long short-term memory(BiLSTM).Firstly,the characteristics of daily power generation of small hydropower are analysed,based on which the daily power generation is decomposed into low-frequency and high-frequency components by VMD technique to extract the cyclic trend characteristics and local fluctuation characteristics of the time series data,and the number of decomposition times of VMD is determined by using the centre-frequency method and Pearson’s correlation coefficient method;finally,the BiLSTM prediction tool based on the selection of daily generation characteristics is used to predict each modal component of the VMD decomposition,and then the final daily generation prediction is obtained by fusing the prediction results of each component.The experimental results show that the selected model can better fit the actual data,improve the prediction accuracy,and outperform other deep learning methods.
作者
刘可真
普伟
赵庆丽
谭化平
赵贞焰
LIU Kezhen;PU Wei;ZHAO Qingli;TAN Huaping;ZHAO Zhenyan(School of Electric Power Engineering,Kunming University of Science and Technology,Kunming 650500,China;China Energy Engineering Group Yunnan Electric Power Design Institute Co.,Ltd,Kunming 650051,China)
出处
《电力科学与工程》
2023年第9期28-37,共10页
Electric Power Science and Engineering
基金
云南省教育厅科学研究基金资助项目(2022J1279)
云南电网有限责任公司科技项目(YNKJXM20180736)。
关键词
小水电
日发电量
预测
双向长短期记忆神经网络
变分模态分解
small hydropower
daily power generation
prediction
bidirectional long-and short-term memory neural networks
variational modal decomposition
作者简介
刘可真(1974-),女,教授,研究方向为新型电力系统规划分析与稳定性;通信作者:普伟(1998-),男,硕士研究生,研究方向为深度学习在电力系统中的应用;赵庆丽(1973-),女,高级工程师,长期从事电力系统规划、电力系统仿真技术研究;谭化平(1999-),男,硕士研究生,主要研究方向为新型电力系统惯量统计;赵贞焰(1998-),女,硕士研究生,研究方向为新型电力系统的频率控制。
