摘要
针对负荷预测中传统经验模态分解(empirical mode decomposition, EMD)存在模态混叠和端点效应,变分模态分解(variational modal decomposition, VMD)存在参数选择困难的问题,提出一种遗传算法(genetic algorithm, GA)、VMD、残差网络(residual network, ResNet)和长短时记忆网络(long short-term memory, LSTM)相结合而成的GA-VMD-ResNet-LSTM负荷预测方法。首先,以包络熵作为适应度函数,利用GA对VMD参数迭代寻优;其次,优化后的VMD将负荷数据分解为相对平稳的子序列并利用皮尔逊相关系数(Pearson correlation coefficient, PCC)筛选影响因素;最后,以ResNet作为特征提取单元处理各重构数据,分别通过LSTM进行时间序列预测,叠加并重构各预测结果,得到最终负荷预测值。将该方法应用于第九届电工数学建模竞赛负荷数据,有力验证了所提方法的精准性和有效性。
In order to solve the problems of modal aliasing and end effect in traditional empirical mode decomposition(EMD), and the difficulty of parameter selection in variational mode decomposition(VMD), in this paper, a load prediction method combining genetic algorithm(GA), VMD, residual network(ResNet) and long short-term memory(LSTM)network GA-VMD-ResNet-LSTM was proposed. Firstly, the envelope en-tropy was used as fitness function to optimize VMD parameters iteratively by GA. Secondly, the optimized VMD decomposed the load data into relatively stable sub-sequences and Pearson correlation coefficient(PCC) was used to screen the influencing factors. Finally, ResNet was used as the feature extraction unit to process the reconstructed data, and LSTM was used to predict the time series, and the prediction results were reconstructed superimposed to obtain the final load prediction value. The accuracy and validity of the proposed method are verified by applying the proposed method to the load data of the 9 th electrical engineering mathematical modeling contest.
作者
简定辉
李萍
黄宇航
Jian Dinghui;Li Ping;Huang Yuhang(School of Physics and Electronic-Electrical engineering,Ningxia University,Yinchuan 75002l,China)
出处
《国外电子测量技术》
北大核心
2022年第10期15-22,共8页
Foreign Electronic Measurement Technology
基金
宁夏自然科学基金(2021AAC03073)项目资助。
关键词
遗传算法
变分模态分解
残差网络
长短时记忆网络
负荷预测
genetic algorithm
variational modal decomposition
residual network
short-and long-term memory networks
load forecasting
