基于主成分分析的用电模式稳定性分析

发布时间：2018-03-01 03:11

本文关键词： 用电模式稳定性主成分分析相似性判定负荷预测　出处：《电力系统自动化》2017年19期 　论文类型：期刊论文

【摘要】：用电模式稳定性分析是实施用户用电量预测的前提,其本质是考察不同历史时间段用电模式的相似性。过长的基础数据时间跨度会降低用电模式稳定性分析的可行性和准确性,而在短期内基于日用电量数据评估的用电特征指标又受随机因素干扰,难以准确反映用电模式。为此,提出一种以过往几周日用电系数和日用电波动率为原始特征指标提取用电模式主成分,进而用两个历史时间段内用电模式主成分因子载荷的欧氏距离衡量用电模式稳定性的方法。针对某小区的算例结果表明,用所提方法判定为用电模式稳定、不稳定的用户组的用电量预测精度存在明显差异,且相似性距离与预测误差存在正相关性。算例分析表明,合适的历史数据时间跨度对提升方法的适用性和准确性至关重要,所提方法采用16周历史数据较为合理。
[Abstract]:Power mode stability analysis is the premise of power consumption prediction, and its essence is to investigate the similarity of power consumption modes in different historical periods. Long time span of basic data will reduce the feasibility and accuracy of power mode stability analysis. However, in the short term, the power consumption characteristic index based on daily electricity consumption data evaluation is disturbed by random factors, so it is difficult to accurately reflect the power consumption mode. This paper presents a method to extract the principal components of the power consumption mode based on the diurnal power consumption coefficient and the daily power fluctuation rate as the original characteristic indexes. Furthermore, the Euclidean distance of the principal component factor load in two historical periods is used to measure the stability of the power consumption mode. The results of an example of a small district show that the proposed method is considered to be stable. The prediction accuracy of power consumption of unstable user groups is obviously different, and the similarity distance is positively correlated with the prediction error. The analysis of example shows that the appropriate time span of historical data is very important to the applicability and accuracy of the method. It is reasonable to use 16 weeks historical data for the proposed method.
【作者单位】：电力传输与功率变换控制教育部重点实验室(上海交通大学);国网上海市电力公司市北供电公司;
【基金】：国家自然科学基金资助项目(51337005) 国家电网公司科技项目(5209141500QW)~~
【分类号】：TM712;TM715

【相似文献】