基于使用智能插座对电路中的电流高频次监测、获得用户用电态势数据,对电流迁移态序列、电流迁移态序列特征的一元回归模型进行了研究.基于电流稳态序列e片段、电流稳态序列片段,设计了电流迁移态序列的构造方法,提出使用一元回归模型描述每一个电流迁移态序列,实现将长度不定的电流迁移态序列到维数固定的电流迁移态序列特征空间的映射.进一步,设计了基于微环境的粒子群优化算法(Microenvironment based Particle Swarm Optimization,MPSO),实现了电流迁移态序列一元回归特征的优化.实验表明:使用所提电流迁移态序列特征进行电器状态识别,平均可以达到97.93％的准确率,且相较PSO算法与CAPSO算法,MPSO算法在使用较少的粒子数达到与这两种算法一致精度的同时,使用时间显著降低.
Research on Feature Model and Mining Method for Current Transition Sequence
In this paper,based on the high-frequency monitoring of current in the circuit via smart sockets and electrical situation data of the customer obtained,the current transition sequence and the univariate regression model for the feature of the current transition sequence are proposed.Based on the e-fragment for the current steady sequence and fragment for the current stable sequence,a mining method for current transition sequence discovery is designed.And further,a univariate regression model is proposed to describe each current transition sequence.And with the univariate regression model,each current transition sequence in variable length is mapped into the feature space for the current transition sequence with fixed dimension.Furthermore,a microenvironment-based particle swarm optimization(MPSO)algorithm is given to optimize the univariate regression features for the current transition sequences.The experimental results show that using the current transition sequence features proposed in this paper for electrical device state recognition can achieve an average accuracy of 97.93％.Compared with the PSO algorithm and CAPSO algorithm,the MPSO algorithm used in this paper achieves the same accuracy with fewer particles and significantly reduces usage time.
current transition sequenceparticle swarm optimization algorithmunivariate regression modelmicroenvironment
NETL
NSTL
万方数据
维普
