基于CNN-GRU-MLR的多频组合短期电力负荷预测  

Multi-frequency Combination Short-Term Power Load Forecasting Based on CNN-GRU-MLR

作　　者：方娜[1,2] 李俊晓 陈浩[1,2] 余俊杰 FANG Na;LI Jun-xiao;CHEN Hao;YU Jun-jie(Hubei Key Laboratory for High-efficiency Utilization of Solar Energy and Operation Control of Energy Storage System,Hubei University of Technology,Wuhan Hubei430068,China;Hubei Power Grid Intelligent Control and Equipment Engineering Technology Research Center,Hubei University of Technology,Wuhan Hubei 430068,China)

机构地区：[1]湖北工业大学太阳能高效利用及储能运行控制湖北省重点实验室,湖北武汉430068 [2]湖北工业大学湖北省电网智能控制与装备工程技术研究中心,湖北武汉430068

出　　处：《计算机仿真》2023年第1期118-124,共7页Computer Simulation

基　　金：国家自然科学基金青年科学基金项目(51809097);湖北省教育厅科学技术研究计划指导性项目(B2018044);太阳能高效利用湖北省协同创新中心开放基金(HBSKFQN2016007);长江科学院开放研究基金(CKWV2018496/KY)。

摘　　要：负荷预测对于电力企业制定未来调度计划十分重要。为了进一步提高预测精度,充分挖掘负荷数据中时序特征的联系,提出一种卷积神经网络(Convolutional Neural Networks,CNN)、门控循环单元(Gate Recurrent Unit,GRU)和多元线性回归(Multiple Linear Regression,MLR)混合的多频组合电力负荷预测模型。该模型先对时间序列的负荷数据进行集合经验模态分解(Ensemble Empirical Mode Decomposition,EEMD),并将其重构为高低两种频率;同时在高频中引入影响因子较大的气象因素,使用CNN-GRU模型预测,低频部分使用多元线性回归进行预测;最后将各个模型得出的预测结果叠加,得到最终预测结果。仿真结果表明,相对于其它网络模型,提出的混合模型具有更高的预测精度,是一种有效的短期负荷预测方法。The load forecasting for the future scheduling plan ofelectric power enterprises is very important.In order to further improve the prediction accuracy and excavate the relationship of time series features in load data,this paper proposes a multi-frequency combination power load forecasting model based on convolutional neural network(CNN),gate recurrent unit(GRU)and multiple linear regression(MLR).In this model,the load data of time series were first decomposed by ensemble empirical mode decomposition(EEMD),and then reconstructed into high and low frequencies.At the same time,for the high frequency part,the GRU-CNN model was used for prediction,while for the low frequency part,MLR was used.Finally,the prediction results obtained from each model were superimposed to form the final forecast results.The simulation results show that the proposed model has higher forecast accuracy compared with other network models and it is an effective method for short-termloadforecasting.

关 键 词：集合经验模态分解 门控循环单元 多元线性回归 卷积神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]