摘要
充电桩运行的稳定性直接影响整体充电网络的运行效率,预测充电桩的故障可以为相关的运营管理提供有力的数据支撑,为此提出一种基于深度学习的公共充电桩故障预测模型。给出受限玻尔兹曼机模型,通过Gibbs采样法求得隐层单元的具体数值,归一化处理后将其加入逐层预训练中,得到深度受限玻尔兹曼机模型,以运维处理及时率、充电桩硬件质量、维修及时率、充电区停电时长计算运维影响指数,结合天气因素计算环境影响指数,二者融合实现综合预测模型的构建。仿真实验结果表明,所提模型的预测能力强且准确率高,具有一定的实际应用价值。
The stability of the charging point directly affects the operation efficiency of the whole charging network.The prediction of the charging point fault can provide strong data support for the relevant operation management.Therefore,a public charging point fault prediction model based on deep learning is proposed.The restricted Boltzmann machine model is given,and the specific value of hidden layer unit is obtained by Gibbs sampling method.After normalization,it is added into layer by layer pre training to obtain the deeply restricted Boltzmann machine model.The operation and maintenance impact index is calculated by operation and maintenance processing timeliness rate,charging pile hardware quality,maintenance timeliness rate and charging area blackout duration,and the environmental impact index is calculated by combining with weather factors.The fusion of the two realizes the construction of comprehensive prediction model.Simulation results show that the proposed model has strong prediction ability and high accuracy,and has a certain practical application value.
作者
吴丹
王俊
许燕
WU Dan;WANG Jun;XU Yan(State Grid Shanghai Municipal Electric Power Company,Shanghai 200122,China;Shanghai Electric Industrial Co.,Ltd.,Shanghai 200025,China)
出处
《电子设计工程》
2022年第5期127-130,135,共5页
Electronic Design Engineering
关键词
深度学习
Gibbs采样法
归一化处理
故障预测
deep learning
Gibbs sampling method
normalization processing
fault prediction
