分类算法在能耗分析系统中的应用场景研究及实现

发布时间：2018-07-15 11:00

【摘要】：随着移动互联网、物联网、云计算的高速发展,数据中心作为信息的重要载体,迎来了一波建设的浪潮。大规模、高密度的IT设备部署,大量的分布式系统的应用,供电和制冷的需求不断增加,所有这些使得机房的运行维护和管理复杂、能量消耗巨大。但在机房能耗分析和控制方面,通信业普遍存在采集范围程度不够、信息孤岛化及粗放式管理等问题。能耗分析系统可以针对收集的基站、机房能耗数据进行统一存储分析,为管理者提供决策支持,但是该系统目前仅能对数据进行简单的统计分析,远没有开发出能耗数据真正的价值。在这样的背景下,本课题从分类算法在能耗分析系统中应用场景的探索为切入点,使用分类算法挖掘能耗数据中潜在的价值。课题研究了分类算法在节能措施业务场景、异常能耗数据告警场景以及能耗模型场景中的应用,丰富并完善了原有的业务场景。重点研究了分类算法在节能措施业务场景中的应用,设计并实现了节能措施仿真平台。该平台由三个子系统组成:并行计算子系统,节能措施实施子系统和能耗数据仿真子系统。并行计算子系统提供计算支持,能耗数据仿真子系统提供能耗数据支持,节能措施实施子系统向用户推荐节能措施,并提供节能措施效果对比功能。当机房出现异常能耗告警时,基于分类算法实现的节能措施仿真平台可以向管理员推荐合适的节能措施。另一方面,管理员在真正在机房安装、配置、实施某节能措施前,可以通过节能措施仿真平台接近零成本模拟实施该节能措施,验证该节能措施的效果,大幅降低验证节能措施效果的成本。
[Abstract]:With the rapid development of mobile Internet, Internet of things, cloud computing, data center as an important carrier of information, ushered in a wave of construction wave. Large-scale, high-density IT equipment deployment, a large number of distributed system applications, the increasing demand for power supply and refrigeration, all these make the operation, maintenance and management of the computer room complex, energy consumption is huge. However, in the analysis and control of computer room energy consumption, there are some problems in the communication industry, such as insufficient acquisition scope, isolated information and extensive management, etc. The energy consumption analysis system can be used to store and analyze the energy consumption data of the base station and computer room in a unified way, which can provide decision support for the manager. But at present, the system can only carry out simple statistical analysis of the data. Far from developing the true value of energy consumption data. In this context, this paper uses the classification algorithm to mine the potential value of energy consumption data from the exploration of the application of classification algorithm in the energy consumption analysis system. This paper studies the application of classification algorithm in the business scene of energy saving measures, the alarm scenario of abnormal energy consumption data and the scenario of energy consumption model, which enriches and perfects the original business scenario. The application of classification algorithm in the business scene of energy saving measures is studied, and the simulation platform of energy saving measures is designed and implemented. The platform consists of three subsystems: parallel computing subsystem, energy saving implementation subsystem and energy consumption data simulation subsystem. Parallel computing subsystem provides computing support, energy consumption data simulation subsystem provides energy consumption data support, energy saving measures implementation subsystem recommends energy saving measures to users, and provides the function of comparing the effect of energy saving measures. When the abnormal energy consumption alarm appears in the computer room, the simulation platform based on the classification algorithm can recommend the appropriate energy-saving measures to the administrator. On the other hand, before the administrator really installs, configures, and implements some energy-saving measures in the computer room, he can simulate the energy-saving measures through the simulation platform of energy-saving measures at close to zero cost, and verify the effect of the energy-saving measures. Significantly reduce the cost of verifying the effectiveness of energy conservation measures.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP311.52

【参考文献】