大用户用电行为分析及任务调度优化研究

发布时间：2018-06-27 12:23

本文选题：大用户 + 用电行为　；参考：《华北电力大学》2017年硕士论文

【摘要】：目前,随着国民经济的发展以及产业结构的调整,国家电网设备已经向大容量、高参数自控设备升级,导致了大用户即电压等级高、负荷大的用户数量明显增加。也就是说智能电网的要求越来越多,电力系统负荷越来越大,调度管理的作用越来越重要。再加上信息采集系统应用的扩展,用户用电负荷数据成海量态势增长。因此,对电网企业的电力信息化建设提出了更高的要求。如何处理不断增长的大用户用电负荷数据,进行快速有效地用电行为分析成为了重要课题。在此基础上,本文对如下问题开展了研究。首先,根据大用户用电负荷数据特点,选择模糊聚类算法进行用电负荷特性分析。为了解决传统FCM算法聚类效果一般、易陷入局部解的问题,本文利用免疫双态粒子群算法来改进FCM算法,设计出免疫双态粒子群模糊C均值聚类算法,该算法在全局收敛能力方面具有优势。其次,考虑到任务选择资源的不确定性对任务执行速度的影响,采用Min-Min启发式算法和吞吐量驱动的调度机制,依据任务的偏好类型,设计出吞吐量驱动最小代价模糊C均值聚类算法,该算法可以提高系统资源利用率和吞吐能力,保证系统负载均衡性。最后,结合这两种算法的优点,给出一种新的吞吐量驱动最小代价免疫双态粒子群模糊C均值聚类算法,并运用了Spark内存批处理技术,使该算法可以在云平台上并行执行,从而解决日益增长的大用户用电数据量与算法执行性能相矛盾的问题。为了验证本文设计的算法可以有效地分析用户用电行为,并且对任务调度有良好的优化效果,在实验室搭建的云集群上运行设计的算法进行验证。
[Abstract]:At present, with the development of the national economy and the adjustment of the industrial structure, the equipment of the State Grid has been upgraded to large capacity and high parameter automatic control equipment, which has led to the increase of the number of the large users, that is, the high voltage grade and the heavy load. In other words, the demand of smart grid is more and more, the load of power system is increasing, and the role of dispatching management is becoming more and more important. In addition, with the expansion of the application of information collection system, the power load data of users is growing in a huge amount. Therefore, higher requirements are put forward for electric power informatization construction of power grid enterprises. How to deal with the increasing load data of large users and how to analyze the power consumption behavior quickly and effectively has become an important subject. On this basis, this paper studies the following issues. Firstly, according to the characteristics of large user load data, fuzzy clustering algorithm is selected to analyze the power load characteristics. In order to solve the problem that the clustering effect of traditional FCM algorithm is general and easy to fall into local solution, the immune double state particle swarm optimization algorithm is used to improve the FCM algorithm and the immune double state particle swarm fuzzy C-means clustering algorithm is designed. The algorithm has advantages in global convergence ability. Secondly, considering the effect of uncertainty of task selection resources on task execution speed, Min-Min heuristic algorithm and throughput driven scheduling mechanism are used according to task preference type. A throughput driven minimum cost fuzzy C-means clustering algorithm is designed, which can improve system resource utilization and throughput capacity and ensure system load balance. Finally, combining the advantages of these two algorithms, a new throughput driven immune two-state particle swarm fuzzy C-means clustering algorithm is proposed, and Spark memory batch technology is used to make the algorithm run in parallel on the cloud platform. In order to solve the problem that the increasing amount of power consumption of large users contradicts the performance of the algorithm. In order to verify that the algorithm designed in this paper can effectively analyze the power consumption behavior of users and has a good optimization effect on task scheduling, the designed algorithm is run on the cloud cluster built in the laboratory to verify the proposed algorithm.
【学位授予单位】：华北电力大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TM73

【参考文献】