Neural Network Predictive Control for Data Center Cooling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current computing environments in data centers face challenges in making dynamic adjustments to power and thermal management due to the computational complexity of existing modeling approaches, leading to potential hardware damage and data loss from excessive power draw or thermal events.
Innovation Solution
Implementing neural networks with computational fluids dynamics (NNCFD) systems that predict future states of temperature and fluid flow in data centers, enabling proactive adjustments to cooling and power distribution through machine learning capabilities integrated into control devices and systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional modeling approaches are used to determine environmental state, then measurement precision is improved, but device complexity increases making real-time control impossible
Solution Approach 1:
The patent replaces traditional computational fluid dynamics models with a neural network-based system. The neural network is trained offline using CFD simulations and sensor data, then deployed for real-time inference. This substitution transforms the complex mechanical/computational modeling approach into a lightweight machine learning model that can run on edge devices with limited computational resources, achieving both accuracy and real-time performance.
Solution Approach 2:
The neural network is trained in advance using extensive CFD simulations and historical sensor data to learn the complex relationships between environmental parameters. This preliminary training phase allows the model to capture intricate patterns without requiring complex computations during real-time operation. The pre-trained network can then make rapid predictions about future thermal and fluid states based on current sensor readings.
2Ease of operation
If purely reactive control is implemented, then ease of operation is improved, but reliability deteriorates due to inability to prevent thermal events
Solution Approach 1:
The system uses the trained neural network to predict future thermal and fluid states before they occur. By analyzing current sensor data and forecasting future conditions, the system can proactively adjust cooling systems and power distribution to prevent thermal events before they happen. This predictive capability maintains operational simplicity while significantly improving reliability through preventive rather than reactive control.
Solution Approach 2:
The neural network identifies emerging thermal conditions that could lead to harmful events and triggers preventive counter-actions. The system adjusts cooling rates and power distribution in advance to counteract potential thermal runaway or equipment damage, effectively applying anti-action before the harmful effect manifests. This approach protects hardware while maintaining simple automated operation.
Data Source
AI summary
Systems and methods for cooling a computer environment are disclosed. In at least one embodiment, one or more neural networks can be used to determine one or more temperature control settings associated with one or more servers.