摘要
针对短期电力负荷数据随机性强,难以实现准确预测的问题,提出了一种基于自适应噪声完备集合经验模态分解(complete ensemble empirical mode decomposition with adaptive noise,CEEMDAN)和时间卷积网络-长短期记忆网络(temporal convolutional network-long short-term memory network,TCN-LSTM)混合模型的预测方法。所提算法先使用CEEMDAN方法将负荷数据分解为一系列相对平稳的子序列。同时为了降低后续计算规模,通过引入排列熵的方法将各子序列进行重组。然后,将各个重组序列输入到TCN-LSTM组合模型中,利用TCN模型提取特征并构建序列的特征向量,再基于LSTM模型对其进行训练及预测。最后把全部预测值进行相加得到完整的预测负荷值。通过使用欧洲某地真实负荷数据进行验证。结果表明:所提算法与其他常见的预测算法相比具有更高的预测精度,可为负荷预测等研究工作提供相关参考。
Based on the complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN)and temporal convolutional network-long short-term memory network(TCN-LSTM)hybrid model,a forecasting method was proposed for the problem that the short-term power load data are so stochastic that it is difficult to achieve accurate forecasting.Firstly,the proposed algorithm decomposed the load data into a number of relatively smooth subseries using the CEEMDAN algorithm.Secondly,in order to reduce the computational scale of the subsequent model,the series were recombined by using the permutation entropy algorithm.Then,the individual recombinant series were input into the TCN-LSTM hybrid model,and the TCN model was used to extract features and construct feature vectors of the series,which were then trained and predicted based on the LSTM model.Finally,all the obtained predicted values were summed to obtain the complete predictive value of power load values.By using the real load data from a European site for validation,the results justify the higher prediction accuracy compared with other common prediction algorithms,which provides a relevant reference for research work such as load prediction.
作者
赵星宇
吴泉军
朱威
ZHAO Xing-yu;WU Quan-jun;ZHU Wei(Smart Energy Mathematics Research Center of College of Mathematics and Physics,Shanghai University of Electric Power,Shanghai 201306,China)
出处
《科学技术与工程》
北大核心
2023年第4期1557-1564,共8页
Science Technology and Engineering
基金
国家自然科学基金(61903244)
上海辰仕科技发展有限公司一般企事业单位资助项目(H2019-269)。
