基于FTI与CNN-BiLSTM网络的短期用电量预测  

Short-Term Power Consumption Prediction Based on Financial Technical Indicators and CNN-BiLSTM Network

作　　者：雒亚锋 赵庆生[1] 梁定康 王旭平[1] LUO Ya-feng;ZHAO Qing-sheng;LIANG Ding-kang;WANG Xu-ping(College of Electrical and Power Engineering,Taiyuan University of Technology,Taiyuan Shanxi 030024,China)

机构地区：[1]太原理工大学电气与动力工程学院,山西太原030024

出　　处：《计算机仿真》2025年第2期72-77,共6页Computer Simulation

基　　金：国家自然科学青年基金项目资助(51907138);国网山西省电力有限公司科技项目资助(5205E0220003)。

摘　　要：为了更好地利用智能电表中的用户数据,提高短期用电量的预测精度,提出一种基于金融技术指标与CNN-BiLSTM组合模型的短期用电量预测方法。首先,根据用电量数据计算顺势指标(Commodity Channel Index,CCI)、趋向指标(Directional Movement Index,DMI)、布林线(BOLLinger band,BOLL)等金融技术指标,用以挖掘用电量数据的趋势特征,并将其添加到神经网络模型的输人数据中;其次,利用卷积神经网络(Convolutional Neural Networks,CNN)对输人数据进行特征提取,将输人到双向长短时记忆网络(Bidirectional Long and Short-Term Memory,BiLSTM)中的数据进行优化;最后,利用双向长短时记忆神经网络提取输人数据的内在特征并对用电量数据进行预测。对美国公共数据集OPENEI中某商业用户进行算例分析,通过与其它单一模型及混合模型对比,验证了所提模型具有更高的预测精度。In order to make better use of the user data in smart meter and improve the prediction accuracy of short-term power consumption,a short-term power consumption prediction method based on the combination of financial technical indicators and CNN BiLSTM model is proposed.Firstly,financial technology indicators such as the Commodity Channel Index(CCI),Directional Movement Index(DMI),and Bollinger band(BOLL)are calculated based on electricity consumption data to explore the trend characteristics of electricity consumption data and add them to the input data of the neural network model;Secondly,convolutional neural networks(CNN)are used to extract features from the input data,and the data inputted into Bidirectional Long and Short Term Memory(BiLSTM)is optimized;Finally,the Bidirectional Long and Short Term Memory neural network is used to extract the internal characteristics of the input data and predict the power consumption data.An example of a commercial user in the American public data set OPENEI is analyzed.By comparing with other single models and mixed models,the proposed model has higher prediction accuracy.

关 键 词：智能电表 用电量预测 金融技术指标 卷积神经网络 双向长短时记忆神经网络 

分 类 号：TM715[电气工程—电力系统及自动化]