Data Center Thermal Control Using Task Redistribution and Pre-Cooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing HVAC control systems for data centers face challenges in optimizing energy usage and minimizing costs due to the difficulty in determining when and how to distribute thermal loads across multiple subplants, especially when considering electrical demand charges and real-time pricing.
Innovation Solution
A method and system that collect temperature data from servers and the data center, using predictive models to control the HVAC system, pre-cool the data center before high-activity periods, and redistribute tasks among servers to optimize energy usage and comply with temperature constraints, incorporating a controller that estimates thermal energy output and adjusts operations based on time-variant costs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If thermal load distribution across multiple subplants is optimized using real-time pricing and demand charges, then energy cost is reduced, but control system complexity increases
Solution Approach 1:
The control system is divided into multiple independent subplant controllers, each managing a specific subplant. This segmentation allows complex optimization to be distributed across simpler modular units, reducing overall system complexity while maintaining optimization capabilities through coordinated control of individual subplants based on real-time pricing and demand charges
Solution Approach 2:
The system performs preliminary actions by pre-cooling the data center before periods of high electrical demand or high energy costs. Thermal energy storage is charged in advance during low-cost periods, and predictive models forecast future thermal loads and energy prices, enabling proactive load distribution decisions that reduce energy costs without requiring complex real-time control during peak periods
2Adaptability or versatility
If thermal energy storage is integrated with multiple subplants, then flexibility and energy cost reduction are improved, but integration difficulty increases
Solution Approach 1:
The thermal energy storage system is designed with multi-functionality to serve multiple subplants simultaneously. A single thermal energy storage unit can charge from or discharge to any of the subplants, providing universal flexibility. The control system manages this universal resource through a unified optimization framework that automatically determines the most efficient charge/discharge paths based on real-time pricing and subplant thermal states, reducing integration difficulty through standardized interfaces and protocols
3Loss of energy
If HVAC system pre-cools the data center before high-activity periods, then energy cost is reduced by shifting load to low-cost times, but additional control complexity is introduced
Solution Approach 1:
The system performs preliminary cooling actions by pre-cooling the data center environment and charging thermal energy storage before periods of high electrical demand or elevated energy prices. Predictive models forecast future thermal loads based on historical patterns and weather data, enabling the control system to advance cooling operations to low-cost periods, thereby reducing peak-period energy costs without requiring complex real-time adjustments during high-demand periods
Solution Approach 2:
The control system incorporates feedback mechanisms that continuously monitor actual energy costs, thermal storage states, and data center temperatures. This feedback validates and refines predictive models, enabling the system to learn from past performance and improve pre-cooling decisions. The feedback loop adjusts control parameters based on actual versus predicted outcomes, reducing control complexity by using data-driven adjustments rather than complex theoretical calculations
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables efficient energy management by predicting thermal loads and adjusting operations to reduce energy consumption, aligning with cost-effective times, thereby minimizing overall energy costs and maintaining optimal data center temperatures.
Implementation Method 1
removing heat from air in the data center utilizing the HVAC system
Data Source
AI summary
A method includes measuring a plurality of temperatures corresponding to a plurality of servers located in a data center, determining a subset of the plurality of servers as high-temperature servers based on the plurality of temperatures, and reassigning tasks from at least a portion of the subset of the plurality of servers to one or more other servers of the plurality of servers.