基于智能电表的数据分析

发布时间：2018-07-05 10:26

本文选题：智能电表 + 能量解聚合　；参考：《深圳大学》2017年硕士论文

【摘要】：基于智能电表的数据分析是在不增加额外设备的情况下,不接触的对负载设备进行监测,根据聚合总线上的电表数据解聚合出单个用电器的一种能源解聚合模型。该模型以用电器消耗的功率为特征,结合用电器的历史使用习惯,分析出用电器的当前状态,指导用户合理的用电行为,从而节约电力能源。论文分析了国内外各种已有能量解聚合模型的优缺点,结合用电设备的历史信息,提出两种基于智能电表的能量解聚合预测模型。1)基于用电器状态的因子隐马尔可夫模型。该方法引入用电器的状态来做能量的解聚合,对单状态用电设备来说,传统基于单个用电器的模型和基于用电器状态的模型效果相差不大,但在多状态用电器的解聚合研究中,基于用电器状态的模型整体上具有更高的能量分配准确率。2)基于时间划分的关联规则学习模型。该方法寻找用电器间的相互关系,结合用电器的历史使用习惯,生成经常使用设备的频繁项集,训练出具有强关联规则的模型。该模型对空闲时间段的电表数据稀释一定的倍数,降低空闲时间段的先验信息,在不影响模型解聚合准确度的条件下,降低生成模型的计算复杂度。论文将以上能量解聚合的模型在三种公开数据集上进行对比实验,采用通用的评价标准。整体上,基于用电器状态的因子隐马尔可夫模型比传统模型的能量准确分配率高,尤其是在多状态用电器的预测上。此外,提出的基于时间划分的关联规则学习模型具有最小的预测误差、最高的能量分配率和准确率,解聚合能力优于传统的能量解聚合模型,在小功率用电器、相同功率用电器和具有相同功率用电器组合的解聚合问题中同样适用。
[Abstract]:The data analysis based on intelligent ammeter is a kind of energy depolymerization model of single electric appliance, which is based on the data of ammeter on the aggregation bus, monitoring the load equipment without adding extra equipment. This model is characterized by the power consumption of electrical appliances, combined with the historical usage habits of electrical appliances, analyzes the current state of electrical appliances, instructs users to use electricity rationally, and saves power energy. This paper analyzes the advantages and disadvantages of various existing energy depolymerization models at home and abroad. Combining with the historical information of electrical equipment, two prediction models of energy depolymerization based on intelligent electric meter (1) are proposed. 1) the factor hidden Markov model based on the state of electrical appliances is proposed. This method introduces the state of electrical apparatus to depolymerize the energy. For single-state electric equipment, the effect of traditional model based on single electric appliance and the model based on state of electrical appliance is not different, but in the research of de-aggregation of multi-state electric appliance, The model based on the state of electrical appliance has higher energy allocation accuracy (.2) the learning model of association rules based on time division. Based on the historical usage habits of electrical appliances, the frequent itemsets of frequently used devices are generated, and a model with strong association rules is trained. The model dilutes a certain multiple of the ammeter data in the idle time period, reduces the prior information of the idle time period, and reduces the computational complexity of the generated model without affecting the accuracy of the model depolymerization. In this paper, the above models of energy depolymerization are compared on three kinds of open data sets, and the general evaluation criteria are adopted. On the whole, the factor hidden Markov model based on the state of electrical appliances has a higher accurate energy distribution rate than the traditional model, especially in the prediction of multi-state electrical appliances. In addition, the proposed association rule learning model based on time division has the minimum prediction error, the highest energy distribution rate and accuracy, and the ability to deaggregate is superior to the traditional energy deaggregation model. The same applies to the depolymerization problem of the same power electric appliance and the same power electric appliance combination.
【学位授予单位】：深圳大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TM933.4

【参考文献】