Hierarchical Data Center Power Control for Neural Load Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges in accommodating workload variations and energy efficiency, with existing control designs being complex and lacking scalability and technology reusability, and current AI/ML models requiring extensive training and tuning for each cluster, which is time-consuming and not universally applicable.
Innovation Solution
A data center system with a three-level control hierarchy that includes a load section with power flow optimizers using neural networks to predict power requirements, an intermediate section for power distribution, and a resource section for configuring power sources, integrating AI/ML models to manage power and thermal loads efficiently across the system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If separate control modules are used for cooling systems, power systems, and IT control, then each module can be designed independently, but the overall system integration becomes extremely complicated and lacks scalability
Solution Approach 1:
The control system is divided into three hierarchical levels: device level (individual equipment control), cluster level (group of devices with similar functions), and data center level (overall system coordination). Each level has its own controller that can operate independently but also coordinates with other levels, enabling modular design while maintaining system integration through standardized communication protocols.
Solution Approach 2:
The patent combines multiple control functions (cooling, power, IT operations) into a unified hierarchical control architecture where all modules operate under coordinated supervision. The cluster-level and data center-level controllers integrate information from various separate modules to achieve holistic system optimization without requiring complex point-to-point integration between individual modules.
2Measurement precision
If AI/ML models are trained specifically for each data center cluster, then the model accuracy for that cluster is improved, but the training time and computational resources increase significantly
Solution Approach 1:
The system performs preliminary data collection and preprocessing at the device and cluster levels before model training is needed. Historical operational data is continuously accumulated and pre-processed in advance, so when model training or retraining is required, the preparation work is already complete, significantly reducing the actual training time and computational burden.
3Productivity
If more data center servers are deployed to accommodate increasing workload requirements, then the processing capacity is improved, but the power consumption and energy efficiency challenges worsen
Solution Approach 1:
The hierarchical control system implements continuous feedback loops at each level where controllers monitor operational parameters (power consumption, workload, temperature) and adjust system operations accordingly. The data center-level controller aggregates feedback from all clusters to optimize overall energy efficiency, dynamically allocating workloads and adjusting operational parameters to maintain high productivity while minimizing total power consumption.
Data Source
Figure 1
Figure 2~3
Figure 4
AI summary
A data center system includes a load section having an array of electronic racks, a thermal management system, and a power flow optimizer. The power flow optimizer is configured to determine a load power requirement of the load section based on workload data of the electronic racks and thermal data of the thermal management system. The data center system further includes a resource section having a number of power sources to provide power to the load section. The resource section includes a resource controller to configure and select at least some of the power sources to provide power to the load section based on the load power requirement provided by the power flow optimizer. The power flow optimizer includes a power flow neural network (NN) model to predict, based on the thermal data and the load data, an amount of power that IT clusters and the thermal management system need.