基于CNN+LSTM混合模型的电力消耗预测研究  

Research on Power Consumption Prediction Based on CNN+LSTM Model

作　　者：刘义卿 陈新房[1] LIU Yi-qing;CHEN Xin-fang(Institute of Disaster Prevention,Langfang 065201,Hebei)

机构地区：[1]防灾科技学院,河北廊坊065201

出　　处：《电脑与电信》2023年第7期65-69,共5页Computer & Telecommunication

基　　金：2022防灾科技学院大学生创新创业项目,项目编号:S202211775045。

摘　　要：预测电力消耗是一项重要的任务,它在保障电力系统安全运行、均衡能源分配等方面至关重要,精准的负荷预测能有效减少用电事故的发生,并提高系统的生产效率。电力预测是基于一个地区的历史电力数据来预测该地区未来一段时间的有功和无功电力负荷。研究利用神经网络组合模型的优势通过CNN优化和过滤多维输入参数,提取特征向量;并将提取的特征向量作为LSTM的输入进行电力预测。实验表明CNN+LSTM混合模型的泛化能力更强,准确率更高。Predicting power consumption is an important task,as it provides intelligence for public utilities to ensure the safe opera-tion of the power system,balance energy distribution,and other aspects.Accurate load forecasting can effectively reduce the occur-rence of electricity accidents and help them improve the production rate and efficiency of the system.Power forecasting is based on historical power data of a region to predict the active and reactive power loads for a period of time in the future.This paper utilizes the advantages of neural network combination models to optimize and filter multidimensional input parameters through CNN,and extract feature vectors.Then the extracted feature vectors are used as input to LSTM for power load prediction.Experiments show that the CNN+LSTM model has more generalization ability and higher accuracy.

关 键 词：卷积神经网络 混合模型 电力预测 

分 类 号：TM715[电气工程—电力系统及自动化]