Modeling Analysis And Intelligent Control Of Mobile Phone Power Consumption And Temperature

With the rapid development of mobile Internet technology,the smartphone market has been growing at a phenomenal rate,and smartphones have become an essential part of our daily life.With the design of high-performance chips,powerful functions of smartphones bring unprecedented convenience to human life.However,the designs of future high-performance chips of smartphones are hindered by severe energy drain and overheating problems.What is worse,user satisfaction has been severely limited by the smartphone battery life and surface temperature.Thus,smartphone designs must achieve a balance between performance,temperature,and energy management in a coordinated manner to maximize user satisfaction.In other words,designs must offer high performance with guarantees of high energy efficiency,along with temperature requirements.Studies of extending smartphone battery life and decreasing the smartphone temperature are necessarily needed to upgrade user satisfaction.Therefore,the study of methods for power and thermal management of smartphones has an essential academic significance and many practical application values.This thesis focuses on the power and thermal analyzing and management of smartphones.The main contents and contributions are as follows:Firstly,the power and thermal models are provided for the target smartphone.With these models,a Multi-com Ponent Power and Thermal Analysis Tool(MPPTAT),which is an informative,fine-grained real-time power and thermal analysis tool,is proposed.MPPTAT can analyze the power consumption and temperature at the granularity of device hardware components,single app,and thread level.Secondly,detailed characteristics are represented for emerging smartphone applications.Emerging Mobile Augmented Reality(MAR)apps face some critical challenges as the rich functionality increase concerns,including energy drain and heat dissipation.To understand the characteristics of MAR apps,a detailed characterization of Thread-Level Parallelism(TLP),utilization of big.LITTLE architecture,and microarchitecture behaviors are provided.Then the power and thermal analyzing are conducted using MPPTAT.Thirdly,continuous growing interests have been seen in bringing artificial intelligence capabilities on smartphones.However,the related work still faces several issues,such as constrained computation and memory resources,power drain,and thermal limitation.The architectural behaviors of some mainstream deep learning frameworks on mobile devices by performing a comprehensive characterization of performance,accuracy,energy efficiency,and thermal behaviors are explored.Then four model compression methods are chosen to analyze the related impact on the nodes amount,memory,execution time,model size,energy consumption,and thermal distribution.Fourthly,based on the above observations,a mobile Dynamic Thermal Energy Harvesting for Reusing(DTEHR)framework is proposed.The DTEHR exploits(1)Thermo Electric Generator(TEG)for waste heat harvesting,(2)Thermo Electric Cooler(TEC)for cooling,and(3)Micro-Super Capacitors(MSC)for generated energy storage.Different from conventional static TEG,a dynamic TEG module(3-dimensionally mounted between the rear case and chip metal)is proposed in the DTEHR framework to maximize power generation.Since hot spots only appear at the CPU and camera,the TEC module is mounted behind the camera and the CPU to cool hot spots below the human tolerable threshold.With TEC-based cooling,DTEHR reduces temperatures of the surface and internal components significantly.With dynamic TEG,DTEHR generates more than hundreds of times of power that TEC needs to cool down hot spots.Thus,extra-generated power can be stored into MSC to prolong battery life.Finally,a new concept of thermal-aware Satisfaction of smartphone User(t-So U),which is efficiently combined several contrasting factors into an integrated metric to justify their combined influence on user satisfaction,is provided.Then,a t-So U aware power and thermal management policy is developed.The t-So U aware management generates power with dynamic TEG module,and dynamically controls the CPU frequency and the TEC cooling module to maximize user satisfaction.Evaluation results show that the proposed t-So U aware management reduces energy consumption and surface temperature without user satisfaction decreasing.Therefore,the proposed t-So U aware management provides an available software and hardware-based solution for the future power and thermal management of smartphones.