Machine-Trained Model for Data Center Power Efficiency Change Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Data centers face challenges in accurately detecting and analyzing changes in power efficiency due to factors like outside air temperature, leading to incorrect identification of changes in power usage efficiency (PUE) metrics, which can result in inefficient energy management.
Innovation Solution
A method involving machine-trained models to analyze energy usage and associated characteristics over time periods, accounting for dependencies such as outside air temperature, to identify significant changes in power efficiency and provide feedback to operators, including the use of data preparation, dependency modeling, and change detection components to generate and update models based on actual energy usage deviations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional change detection methods are used to monitor power efficiency, then changes in PUE metrics can be identified, but false positives occur due to external factors like outside air temperature
Solution Approach 1:
The patent introduces machine-trained models as intermediary components that mediate between raw energy usage data and change detection decisions. These models account for external factors like outside air temperature by learning their influence patterns, thereby filtering out false positives while preserving true change detections. The model acts as a mediator that translates complex multi-factor relationships into accurate change detection signals.
Solution Approach 2:
The patent transforms the approach by changing from direct threshold-based PUE monitoring to model-based prediction where parameters are dynamically adjusted. The machine-trained models learn optimal parameter relationships between energy usage, external factors, and time, enabling adaptive detection that accounts for varying conditions without increasing false positives.
2Ease of manufacture
If simple threshold-based monitoring is used, then implementation is straightforward, but it cannot account for dependencies on external factors like temperature
Solution Approach 1:
The patent segments the monitoring system into distinct functional components: data collection module, machine-trained model evaluation module, and change detection module. This segmentation allows the complex analysis of external factors to be isolated in the model evaluation component, while maintaining simple interfaces for overall system implementation. Each segment handles specific tasks, making the overall system both comprehensive and implementable.
Solution Approach 2:
The machine-trained models serve as intermediaries that bridge simple monitoring interfaces and complex contextual analysis. These models internally process relationships between energy usage, outside air temperature, and other factors, presenting simplified change detection outputs to operators without requiring them to understand the complex underlying relationships.
3Measurement precision
If continuous monitoring of all factors is performed, then accurate change detection is achieved, but computational resources and time are consumed
Solution Approach 1:
The patent applies preliminary action by pre-training machine learning models offline using historical data that captures relationships between energy usage, external factors, and time. This pre-processing creates ready-to-use models that can quickly evaluate current conditions without performing heavy computations in real-time, thus achieving accurate change detection with minimal processing delay.
Solution Approach 2:
The system performs partial monitoring by focusing computational resources only on evaluating changes relative to model predictions rather than continuously analyzing all raw data streams. The machine-trained models capture the essential relationships, allowing the system to detect changes with sufficient accuracy without excessive processing of every data point.
Data Source
AI summary
In example implementations described herein, there are systems and methods for collecting first data for a first time period and second data for a second time period regarding energy usage for, and associated characteristics of, a datacenter. The method further includes generating, based on the first data for the first time period, a first machine-trained model modeling a relationship between the energy usage and the associated characteristics. For an identified change to the relationship between the energy usage and the associated characteristics based on a first prediction error being one of greater than a first value or less than a second value, the method may include displaying an indication of the identified change; collecting, based on the identified change, third data for a third time period; and generating a second machine-trained model based on the third data.