VMD-LSTM算法在短期负荷预测中的应用  

Application of VMD-LSTM algorithm in short term load forecasting

作　　者：胡欣球 马立新[1] HU Xinqiu;MA Lixin(University of Shanghai for Science and Technology,Shanghai 200093,China)

机构地区：[1]上海理工大学光电信息与计算机工程学院,上海200093

出　　处：《电力科学与工程》2018年第6期9-13,共5页Electric Power Science and Engineering

基　　金：上海市张江国家自主创新重点项目(201310-PI-B2-008)

摘　　要：传统的负荷预测方法如回归分析法、灰色预测法、支持向量机(Support Vector Machine,SVM)等缺少对时序数据相关性、特征值的全面考虑,预测值不稳定,精度较低且波动大。为进一步提升电力系统短期负荷预测的准确率,结合变分模态分解(Variational Mode Decomposition,VMD)和长短时记忆神经网络(Long Short-Term Memory,LSTM)算法,建立了一种短期负荷预测模型(VMD-LSTM)。首先采用VMD技术将输入负荷数据分解为多个有限带宽的本征模态分量,分解结果表明了人们生产生活中不同的用电习惯,并且分离了数据中的噪声和信号,然后对每个模态分量建立LSTM神经网络进行预测,结合模型输出重构预测结果。通过实际算例,验证了该算法相比于传统的负荷预测算法适应性强,预测精度高且稳定,具有显著的理论指导意义和实用价值。Traditional load forecasting methods,such as regression analysis,grey forecasting,and support vector machine(SVM),lack comprehensive considerations of correlations and eigenvalues of time series data.The predictive values are unstable,and what’s more,they are of low precision and have large fluctuations.In order to further improve the accuracy of short-term load forecasting for power systems,this paper combines variational mode decomposition(VMD)and long-term memory(LSTM)algorithms to establish a short-term load forecasting model(VMD-LSTM).Firstly,the input load data is decomposed into multiple finite-bandwidth eigenmode components by using VMD technology.The decomposition results show that people use different electricity habits in production and life.The noise and signal in the data are separated.For each module,the LSTM neural network is used to predict the state components,and the prediction results are reconstructed with the model output.Through practical examples,it is verified that the algorithm is more adaptable than the traditional ones,and the prediction accuracy is high and stable.It has significant theoretical guidance and practical value.

关 键 词：短期负荷预测 变分模态分解 长短时记忆神经网络 VMD-LSTM 

分 类 号：TM721[电气工程—电力系统及自动化]