智能电网超短期负荷预测方法研究

发布时间：2018-03-21 05:57

本文选题：智能电网　切入点：超短期负荷预测　出处：《华北电力大学(北京)》2017年硕士论文　论文类型：学位论文

【摘要】：随着智能电网的发展,越来越多的新能源接入其中,如太阳能、风能等,形成分布式电网模式。然而,这些新能源的发电量易受光照、风速等自然条件的影响,尤其随着新能源接入量的增加,其本身的波动性对智能电网的稳定性带来很大影响。在电网稳定性状态评估和电网实时动态无功电压优化控制等方面,超短期负荷预测具有重要的参考意义。超短期负荷预测具有预测时间短、实时性要求高等特点,目前正处于研究阶段。智能电网中大量的时序数据对于超短期负荷预测具有重要的参考价值,如何有效地利用智能电网中的时序数据,充分挖掘其中潜在信息的关联性进行超短期负荷预测,成为智能电网系统的一个热门研究方向。本文针对目前超短期负荷预测算法存在的稳定性差和忽略用户行为相似性等问题,提出了基于虚拟用户模型和预测区间的超短期预测模型;然后结合电力用户数据的数据流特点,提出了基于数据流聚类的超短期负荷预测方法,提高了预测速度。本文主要的研究有如下几个方面。首先,对现有的超短期负荷预测算法进行了综述,分析了现有预测算法的缺点。其次,针对现有的预测算法中未考虑到用户用电行为的相似性的问题,通过分析用户负荷曲线的特点,提出虚拟用户模型;再次,考虑到用户用电行为的随机性特点,引入预测区间以提高预测算法的稳定性,结合虚拟用户模型,提出了基于虚拟用户模型和预测区间的超短期预测模型;然后,根据智能电网中用户负荷数据的时序特性,采用数据流聚类技术对虚拟用户模型的超短期负荷预测算法进行改进,提高了算法的预测速度;最后,通过实验验证,本文提出的基于虚拟用户模型的超短期负荷预测算法的准确率要优于对比试验中的其它算法,并且引入数据流聚类分析技术后在预测精度可接受范围内,预测速度也得到了显著提升。
[Abstract]:With the development of smart grid, more and more new energy sources are connected to it, such as solar energy, wind energy and so on. However, the generation of these new energy sources is easily affected by natural conditions such as light, wind speed, etc. In particular, with the increase of new energy access, its own volatility has a great impact on the stability of smart grid. In the aspects of power grid stability evaluation and real-time dynamic reactive power and voltage optimization control, etc. Ultra-short-term load forecasting has important reference significance. Ultra-short-term load forecasting has the characteristics of short forecasting time, high real-time requirement and so on. At present, a lot of time series data in smart grid have important reference value for ultra-short-term load forecasting, how to utilize the time series data in smart grid effectively. Fully mining the correlation of potential information for ultra-short-term load forecasting, It has become a hot research direction in smart grid system. This paper aims at the problems of poor stability and neglecting the similarity of user behavior in the current ultra-short-term load forecasting algorithm. An ultra-short-term forecasting model based on virtual user model and prediction interval is proposed, and then a method of ultra-short-term load forecasting based on data stream clustering is proposed, which is based on the characteristics of data flow of power user data. The main research in this paper is as follows: firstly, the existing ultra-short-term load forecasting algorithms are reviewed, and the shortcomings of the existing forecasting algorithms are analyzed. In view of the problem that the existing prediction algorithms do not consider the similarity of the user's power consumption behavior, by analyzing the characteristics of the user load curve, a virtual user model is proposed. Thirdly, considering the randomness of the user's power consumption behavior, The prediction interval is introduced to improve the stability of the prediction algorithm. Combined with the virtual user model, the ultra-short-term forecasting model based on the virtual user model and the prediction interval is proposed. Then, according to the time series characteristics of the user load data in the smart grid, The data stream clustering technique is used to improve the ultra-short-term load forecasting algorithm of the virtual user model, which improves the forecasting speed of the algorithm. The accuracy of the proposed algorithm based on virtual user model is better than that of other algorithms in the contrast experiment, and the prediction accuracy is acceptable after the introduction of data stream clustering analysis technology. The predicted speed has also improved significantly.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM76;TM715

【参考文献】