Design Of Dynamic Power Management For Embedded System

With the rapid development of embedded computing technology,the embedded systems become increasingly powerful and fully functional,but more and more constrained by the relatively slow development of batteries power technologies,power has become another important design goals except performance and size on embedded system,become a important constraint factor to the development of embedded systems,subject to more and more attention.Dynamic power managemnet is the strategy adopted at the system level,aimed at chnging states of system components dynamically for power saving while satisfying user's performance requirement,For battery-powered embedded systems,the battery continued use time must be considered.Thus, this thesis does some reach about the dynamic power management on embedded systems.For considering the embedded linux operating system and hardware platform S3C2410 features, designed software architecture of dynamic power management,devided task into various levels and functional modules,this architecture reflects the separation ideas of mechanism and policy.take different power management approach for system and external equipment,support for system variety low-power state,support for external devices dynamic power management,and provides realization of framework and operation mechanism for dynamic power management strategies.Based on dynamic power management system architecture,power state of work has been partition,design and implement multi-state power management module for dynamic power management,and through test analysis effect of system saving power.Through research and improvement PM power management module on linux operating system,the device driver changes and transformation, a predictive shutdown method based on sliding window and exponential average used for the external module device.dynamic power management,and experiments compared improved algorithm pratical energy saving effect.For considering the existing algorithm in dynamic power management, a exponential average predictive shutdown method based on sliding window was proposed. This method utilized the sliding window, and also imported the dynamic adaptive factor, predict next idle period using previous information of idle period, The simulation experimental results show that this method can keep prediction error rate at low.Actual experimental results show that using dynamic power management algorithm of this thesis,the system continued work time and battery standby time than without power management have a marked increase,more significant saving power of system.