ADRC Thermal Zone Control for Datacenter Cooling Fluctuations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The rapid growth of datacenters and increased computing infrastructure density have made it challenging to provide effective cooling, leading to higher cooling costs and thermal management issues, such as thermal throttling of processors and storage drives.
Innovation Solution
The implementation of an active disturbance rejection thermal control (ADRC) system that uses extended state observers to estimate temperature and disturbances, generating control signals for cooling elements to maintain target temperatures across multiple thermal zones, thereby improving thermal management efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If computing infrastructure density is increased to drive datacenter growth, then productivity and data processing capability are improved, but thermal management difficulty and cooling costs increase
Solution Approach 1:
The datacenter is divided into multiple thermal zones with dedicated ADRC controllers for each zone. This segmentation allows independent thermal management of different regions, enabling precise control of cooling resources and reducing overall thermal management complexity despite increased computing density.
Solution Approach 2:
The ADRC controller implements continuous feedback control by monitoring temperature measurements from thermal zones and adjusting cooling element outputs accordingly. The extended state observer estimates disturbances and temperature states in real-time, creating a closed-loop feedback system that automatically adapts to thermal changes caused by high-density computing workloads.
2Temperature
If cooling capacity is increased to manage thermal loads from high-density servers, then temperature control is improved, but energy consumption and cooling costs increase
Solution Approach 1:
The cooling system transitions from static to dynamic control through the ADRC controller, which continuously adjusts cooling element outputs based on real-time temperature measurements and disturbance estimates. This dynamic adaptation allows the system to provide precise temperature control only when and where needed, reducing unnecessary cooling energy consumption.
Solution Approach 2:
The system changes cooling parameters (output control signals) based on varying thermal conditions and workload disturbances. The extended state observer estimates disturbance parameters in real-time, allowing the controller to adapt cooling intensity dynamically, thereby maintaining temperature control precision while minimizing energy consumption during low-thermal-load periods.
3Device complexity
If conventional thermal control methods are used in high-density datacenters, then system simplicity is maintained, but thermal throttling and performance degradation occur
Solution Approach 1:
The extended state observer acts as an intermediary that estimates unmeasured thermal states and disturbances, providing this information to the ADRC controller. This intermediary component enables sophisticated thermal management without requiring direct measurement of all thermal parameters, maintaining system simplicity while preventing thermal throttling through proactive control.
Solution Approach 2:
The ADRC controller with extended state observer performs preliminary thermal management by estimating future thermal states and disturbances before they cause critical temperature rises. This proactive approach prevents thermal throttling by adjusting cooling outputs in advance, maintaining processor performance without requiring overly complex measurement and control infrastructure.
Data Source
AI summary
A system and method for active disturbance rejection based thermal control is configured to receive, at a first active disturbance rejection thermal control (ADRC) controller, a first temperature measurement from a first thermal zone. The ADRC controller generates a first output control signal for controlling a first cooling element, wherein the first output control signal is generated according a first estimated temperature and a first estimated disturbance calculated by a first extended state observer (ESO) of the first ADRC controller.


