嵌入式系统电源管理的研究与实现
发布时间：2018-07-22 20:48
【摘要】：电子电路集成工艺的飞速发展使计算机系统体积不断缩小,性能不断地提升,同时移动通信技术的发展让这些计算机系统更加的便携,许多便携式计算机开始使用电池供电。我们知道高性能运算通常伴随着高功耗,而电池技术的严重滞后和人们环保意识的增加使得性能和功耗之间的问题越发凸显。电源管理技术的出现缓和了两者之间的矛盾,电源管理技术通过有效的电源分配降低系统的整体功耗。电源管理技术在桌上型计算机、服务器上十分常见,然而在嵌入式领域,由于嵌入式系统的开发通常是针对特殊的应用场合,电源管理技术发展相对缓慢,例如,通用电源管理软件比较少,也没有统一的标准规范。为了引入通用电源管理技术到嵌入式领域,本文研究了静态电源管理和动态电源管理。在静态电源管理方面,本文首先对现有静态电源管理的相关技术进行了分析,提出在嵌入式系统里实现静态电源管理的方法,并在样机上实现静态电源管理中的睡眠和休眠两个功能；在动态电源管理方面,本文首先分析和比较目前流行动态电源管理的策略,包括：贪婪策略,超时策略,预测策略,基于随机过程策略,然后指出了当前这几个策略中的存在若干问题,并由此提出动态自适应策略包括指数平均自适应算法和分类滑动窗口自适应算法。文中通过对静态电源管理睡眠和休眠两个功能在样机上的实现,证明了本文给出静态电源管理在嵌入式系统上应用的可行性和实用性,同时通过对动态电源管理的模拟实现的结果,分析和证明了本文中自适应策略的优势。
[Abstract]:With the rapid development of electronic circuit integration technology, the volume of computer system is shrinking and the performance of computer system is improving. At the same time, with the development of mobile communication technology, these computer systems become more portable, and many portable computers begin to use battery power supply. We know that high performance computing is usually accompanied by high power consumption, and the serious lag of battery technology and the increasing awareness of environmental protection make the problem between performance and power become more and more serious. The emergence of power management technology alleviates the contradiction between the two. Power management technology reduces the overall power consumption of the system through effective power allocation. Power management technology is very common in desktop computers and servers. However, in embedded field, the development of power management technology is relatively slow because embedded systems are usually developed for special applications. For example, General power management software is less, and there is no uniform standard. In order to introduce general power management technology into embedded field, static power management and dynamic power management are studied in this paper. In the aspect of static power management, this paper firstly analyzes the existing technology of static power management, and puts forward the method of realizing static power management in embedded system. In the aspect of dynamic power management, this paper first analyzes and compares the popular dynamic power management strategies, including: greedy strategy, time-out strategy, prediction strategy. Based on the stochastic process strategy, this paper points out some problems in these strategies, and then puts forward some dynamic adaptive strategies, including exponential average adaptive algorithm and classification sliding window adaptive algorithm. Through the realization of the static power management sleep and sleep functions on the prototype, the feasibility and practicability of the static power management in embedded system are proved in this paper. At the same time, through the simulation of dynamic power management, the advantages of adaptive strategy in this paper are analyzed and proved.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1;TN86
本文编号：2138528
[Abstract]:With the rapid development of electronic circuit integration technology, the volume of computer system is shrinking and the performance of computer system is improving. At the same time, with the development of mobile communication technology, these computer systems become more portable, and many portable computers begin to use battery power supply. We know that high performance computing is usually accompanied by high power consumption, and the serious lag of battery technology and the increasing awareness of environmental protection make the problem between performance and power become more and more serious. The emergence of power management technology alleviates the contradiction between the two. Power management technology reduces the overall power consumption of the system through effective power allocation. Power management technology is very common in desktop computers and servers. However, in embedded field, the development of power management technology is relatively slow because embedded systems are usually developed for special applications. For example, General power management software is less, and there is no uniform standard. In order to introduce general power management technology into embedded field, static power management and dynamic power management are studied in this paper. In the aspect of static power management, this paper firstly analyzes the existing technology of static power management, and puts forward the method of realizing static power management in embedded system. In the aspect of dynamic power management, this paper first analyzes and compares the popular dynamic power management strategies, including: greedy strategy, time-out strategy, prediction strategy. Based on the stochastic process strategy, this paper points out some problems in these strategies, and then puts forward some dynamic adaptive strategies, including exponential average adaptive algorithm and classification sliding window adaptive algorithm. Through the realization of the static power management sleep and sleep functions on the prototype, the feasibility and practicability of the static power management in embedded system are proved in this paper. At the same time, through the simulation of dynamic power management, the advantages of adaptive strategy in this paper are analyzed and proved.
【学位授予单位】：南京理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP368.1;TN86
【引证文献】
相关硕士学位论文 前2条
1 宋振孝;基于GINI指数分类的嵌入式CPU功耗预测优化方法研究[D];西北大学;2013年
2 马齐成;基于Android平台电源管理的研究与实现[D];华南理工大学;2013年
，本文编号：2138528
本文链接：https://www.wllwen.com/kejilunwen/jisuanjikexuelunwen/2138528.html
